No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Multivariatc Modeling of Water Resources Time Series Using Artificial Neural Networks
Abstract
The artificial neural network (ANN) approach described in this paper for the synthesis of reservoir inflow series differs from the traditional approaches in synthetic hydrology in the sense that it belongs to a class of data-driven approaches as opposed to traditional model driven approaches. Most of the time series modelling procedures fall within the framework of multivariate autoregressive moving average (ARMA) models. Formal statistical modelling procedures suggest a four- stage iterative process, namely, model selection, model order identifi cation, parameter estimation and diagnostic checks. Although a number of statistical tools are already available to follow such a modelling process, it is not an easy task, especially if higher order vector ARMA models are used. This paper investigates the use of artificial neural networks in the field of synthetic inflow generation. The various steps involved in the development of a neural network and a multivariate auto regressive model for synthesis are presented. The application of both types of model for synthesizing monthly inflow records for two reservoir sites is explained. The performance of the neural network is compared with the statistical method of synthetic inflow generation. Modélisation multivariée de séries chronologiques hydrologiques grâce à l'utilisation de réseaux neuronaux artificiels
... Water resource allocation can balance supply and demand between different water use sectors among cities (Cai, 2008;Read et al., 2014;Hu et al., 2016a). Previous studies (Tabari and Soltani, 2013;Davijani et al., 2016;Yu and Lu, 2018;He et al., 2020) have made great efforts to allocate regional water resources for maximum socioeconomic benefits with minimum environmental impacts. ...
... The prediction of the volume of water availability and water demand forms the basis for water resource allocation (Wei et al., 2020). The most frequently used water supply and demand prediction methods include system dynamics (SD) (Forrester, 1987;Honti et al., 2019), time series prediction (Firat et al., 2009), and artificial neural networks (Raman and Sunilkumar, 1995;Al-Zahrani and Abo-Monasar, 2015). The SD model constructed through detailed parameter estimation and numerical equations can effectively reduce uncertainty and improve accuracy, and has become the most widely used in water resource management modeling (Neuwirth, This study focused on mitigating future water shortages caused by the spatially uneven distribution of regional water resources by a spatial equilibrium-based optimal water allocation. ...
... Compared with seeking the maximum socioeconomic benefits and the minimum environmental impacts, the proposed framework could obtain a compromise solution through the decision-making method. Read et al. (2014) underlined the importance of stability of allocation solution and suggested that selecting an inferior stable solution would be better than an unstable optimal solution. The selection of compromise solutions and the consideration of the principle of SEWA in this study could increase the stability of water allocation solutions somewhat (Hajkowicz and Collins, 2007). ...
Study region
Guangdong Province in China.
Study focus
Water shortages due to the spatially uneven distribution of water resources have become the main obstacle to the sustainable development of regional society and the economy. To alleviate this problem, this study developed a framework including prediction, optimization, and decision-making models to allocate available water resources among the different sectors of the cities in the region. The framework was advantageous in efficiently predicting future water demand and supply for multiple cities, quantitatively reflecting the level of the spatial equilibrium of water allocation (SEWA) through coupling coordination degree (CCD), and achieving a higher level of SEWA rather than just the equitable water distribution.
New hydrological insights for the region
The results indicated that: (i) by 2030, the deficit of water supply and demand of Guangdong Province would be further aggravated, with a water shortage rate of 4.18%; (ii) by optimal water allocation, the water shortage rate of Guangdong Province decreased to 1.56% and the level of SEWA improved significantly from moderate equilibrium to good equilibrium; and (iii) from 2018 to 2030, key water-saving sectors in different cities were identified, while the industrial sector had a higher water-saving intensity than other water use sectors. This study could provide references for integrated water allocation strategies to realize the coordinated development of socioeconomic and environmental systems in other regions of the world.
... It is becoming a research hotspot of hydrology to predict runoff with an in-depth learning algorithm. In 1995, Raman and Sunilkumar [92] applied machine learning to a mid-and long-term (>1 year) runoff forecast. They also constructed a mid-and long-term runoff forecast model based on CNN and used it to forecast runoffs flowing into the Mangalam Reservoir and Pothimdy Reservoir. ...
... It is becoming a research hotspot of hydrology to predict runoff with an in-depth learning algorithm. In 1995, Raman and Sunilkumar [92] applied machine learning to a mid-and long-term (>1 year) runoff forecast. They also constructed a mid-and longterm runoff forecast model based on CNN and used it to forecast runoffs flowing into the Mangalam Reservoir and Pothimdy Reservoir. ...
Inland river runoff variations in arid regions play a decisive role in maintaining regional ecological stability. Observation data of inland river runoff in arid regions have short time series and imperfect attributes due to limitations in the terrain environment and other factors. These shortages not only restrict the accurate simulation of inland river runoff in arid regions significantly, but also influence scientific evaluation and management of the water resources of a basin in arid regions. In recent years, research and applications of machine learning and in-depth learning technologies in the hydrological field have been developing gradually around the world. However, the simulation accuracy is low, and it often has over-fitting phenomenon in previous studies due to influences of complicated characteristics such as “unsteady runoff”. Fortunately, the circulation layer of Long-Short Term Memory (LSTM) can explore time series information of runoffs deeply to avoid long-term dependence problems. In this study, the LSTM algorithm was introduced and improved based on the in-depth learning theory of artificial intelligence and relevant meteorological factors that were monitored by coupling runoffs. The runoff data of the Yarkant River was chosen for training and test of the LSTM model. The results demonstrated that Mean Absolute Error (MAE) and Root Mean Square error (RMSE) of the LSTM model were 3.633 and 7.337, respectively. This indicates that the prediction effect and accuracy of the LSTM model were significantly better than those of the convolution neural network (CNN), Decision Tree Regressor (DTR) and Random Forest (RF). Comparison of accuracy of different models made the research reliable. Hence, time series data was converted into a problem of supervised learning through LSTM in the present study. The improved LSTM model solved prediction difficulties in runoff data to some extent and it applied to hydrological simulation in arid regions under several climate scenarios. It not only decreased runoff prediction uncertainty brought by heterogeneity of climate models and increased inland river runoff prediction accuracy in arid regions, but also provided references to basin water resource management in arid regions. In particular, the LSTM model provides an effective solution to runoff simulation in regions with limited data.
... This model consists of neurons, which are units responsible for processing information. In this model, the neurons receive one or more inputs and determine the vector output based on a predetermined nonlinear function (Raman and Sunilkumar 1995). In general, no need of the prior knowledge for implementation of the model and no definition of predetermined limitation or structure to form a relationship between inputs and output are the unique features of ANN model (French et al. 1992). ...
A key issue for effective management and operating of dam reservoirs is predicting the water inflow values into dam reservoir. To address this subject, here, genetic programming (GP) is used by proposing two cases. In the first case, water inflow values are predicted separately for each month. However, in the second case, these values are predicted simultaneously for all months. Furthermore, for each case, two approaches are proposed here. In the first approach, the hybrid method, called the ANN-NGSA-II method, is proposed to find proper input data sets. However, in the second approach, the useful input data sets are found automatically using the GP method. For comparison purpose, the ANN and SARIMA models are also used, to predict the water inflow values. As a case study, in this research, the Zayandehroud dam reservoir is selected. The results indicate that the ANN model outperforms both results of the GP and SARIMA methods. In other words, correlation coefficient (R²), Nash Sutcliffe (NS), and root means square error (RMSE) values of ANN are 0.97, (0.88), 0.954 (0.87), and 17.19 (30.54) million cubic meters, respectively, for training (test) data set.
... Then, nonparametric models, such as the moving block bootstrap method (Srinivas and Srinivasan 2005) and k-nearest neighbor method (Prairie et al. 2007), are applied to the stochastic simulation of hydrological series. In 1995, Raman and Sunilkumar (1995) first introduced the artificial neural network (ANN) to generate monthly streamflow series for two reservoirs. The results demonstrated that ANN is better than the AR(2) model at keeping streamflow series statistically stable. ...
The streamflow process is an crucial information resource for the joint optimal operation of reservoirs. As the length and representativeness of historical streamflow samples are insufficient for practice projects, streamflow stochastic generation approaches are usually used to expand the streamflow series. For the joint operation and management of the multi-reservoir system, the multisite streamflow stochastic generation (MSSG) with high-dimensional temporal-spatial correlation poses a challenge. This paper develops the generative adversarial network as a novel MSSG model. In contrast to the existing literature on MSSG, which solely focuses on a specific case study and provides a comparatively one-sided assessment, this paper evaluates multiple characteristics of streamflow at various time scales from three MSSG models in two instances. Specifically, three MSSG models, namely the seasonal autoregression (SAR) model coupled with the master station method, the Copula model coupled with the master station method, and the deep convolutions generative adversarial network (DCGAN) model, are employed to generate monthly, ten-daily, and daily streamflow series of the two-reservoir and eight-reservoir systems. This study aims to examine the performance of three models and provide recommendations for implementing MSSG approaches in practice. Results show that: (1) the priority should be given to the maximum iterations on the DCGAN model at a large time scale, while at a smaller time scale, the training of the model is directly linked to the setting of batch size; (2) the Copula model is capable for better retaining statistical characteristics of streamflow series for similarity; (3) the SAR model excels in simulating the extremes of streamflow; and (4) the DCGAN model possesses a significant advantage in capturing the temporal-spatial higher-order correlation, especially in systems comprising more than two reservoirs and with small time scales (e.g., daily streamflow). Furthermore, this study presents comprehensive and multi-scale recommendations for selecting MSSG approaches, thereby providing a theoretical foundation and practical value for MSSG in diverse scenarios.
... Numerous studies have been dedicated to streamflow forecasting using time-series models, with notable examples including the autoregressive integrated moving average (ARIMA) model [18], as well as MA and AR models [19]. While these models often assume a linear relationship between inputs and outputs, the actual relationship is typically characterized by a higher degree of nonlinearity. ...
Accurate monthly streamflow prediction is crucial for effective flood mitigation and water resource management. The present study proposes an innovative similarity-derived model (SDM), developed based on the observation that similar monthly streamflow patterns recur across different years under comparable hydrological and climate conditions. The model is applied to the Lancang River Basin in China. The model performance is compared with the commonly used support vector machine (SVM) and Mean methods. Evaluation measures such as RMSE, MAPE, and NSE confirm that SDM6 with a reference period of six months achieves the best performance, improving the Mean model by 79.9 m3/s in RMSE, 6.07% in MAPE, and 8.62% in NSE, and the SVM by 53.65 m3/s, 0.24%, and 5.53%, respectively.
... Neural network models can find a relationship between input samples and can group samples similarly to cluster analysis. Neural networks have been applied in many areas of water resources such as the development of the rainfall-runoff model, stream flow forecasting, ground water modeling, etc. Neural network models provide better results when compared with other conceptual SAC-SMA (Sacramento Soil Moisture Accounting) models (Hsu et al. 1995), autoregressive models (Raman & Sunilkumar 1995), ARMAX models (Fernando & Jayawardena 1998), multiple regression models (Thirumalaiah & Deo 2000), linear and non-linear regressive models (Elshorbagy et al. 2000) and conceptual models (Salas 1993). Asadnia et al. (2014); Kalteh (2008); Kumar et al. (2008); Nourani et al. (2014); Nourani et al. (2011);Solaimani (2009);Sudheer et al. (2002) used the neural network model for rainfall-runoff studies. ...
In this study, rainfall–runoff (R–R) models were developed by assembling Particle Swarm Optimization (PSO) with the Feed Forward Neural Network (FFNN) and the Wavelet Neural Network (WNN). Performances of the model were compared with the wavelet ensembled neural network (WNN) and the conventional FFNN. The data from 1981 to 2005 were used for calibration and from 2006 to 2014 for validation of the models. Different combinations of rainfall and runoff were considered as inputs to the PSO–FFNN model. The fitness value and computational time of all the combinations were computed. Input combination was selected based on the lowest fitness value and lowest computational time. Four R–R models (FFNN, WNN, PSO–FFNN and PSO–WNN) were developed with the best input combination. The performance of the models was evaluated using statistical parameters (Nash–Sutcliffe Efficiency (NSE), D and root mean square error (RMSE)) and parameters vary in the range of (0.86–0.90), (0.95–0.97) and (68.87–84.37), respectively. After comparing the performance parameters and computational time of all four models, it was found that the PSO–FFNN model gave better values of NSE (0.89), D (0.97), RMSE (68.87) and less computational time (125.42 s) than other models. Thus, the PSO–FFNN model was better than the other three models (FFNN, WNN and PSO–WNN).
HIGHLIGHTS
The goals of this research are selection of input combination based on fitness value and computational time.;
Four R-R models (FFNN, WNN, PSO-FFNN and PSO-WNN) were developed with the best input combination. The performance of the models was evaluated using statistical parameters (NSE, D and RMSE).;
... Fiziksel tabanlı bu modellerin karmaşık yapılarından dolayı yağış akış ilişkisinin fiziksel yönünün dikkate alınmadığın kapalı kutu modeli geliştirilmiştir. Kapalı kutu modellerinde yağış ile akarsu çıkışında akımın görülmesi arasındaki gecikmeden yararlanılmakta ve genel olarak stokastik zaman serilerini kullanan çalışmalar (Raman and Sunilkumar, 1995;Karabörk ve Kahya, 1999;Keskin ve Taylan, 2007), klasik istatistiksel yöntemler kullanılarak yapılan AR, ARMA ve regresyon analizi çalışmaları (Alp ve Cığızoğlu, 2005;Raman ve Sunilkumar, 1995) ve son yıllarda daha pratik uygulamalar sunan veriye dayalı bir teknik olan Yapay Sinir Ağları (YSA) çalışmaları (Shamseldin, 1997;Tokar ve Johnson, 1999;Chang ve Chen, 2001;Öztopal ve diğerleri, 2001;Jayawardena ve Fernando, 2001;Sivakumar ve diğerleri, 2002;Kisi, 2004;Demirpençe, 2005;Yurdusev ve diğerleri, 2008;Okkan ve Mollamahmutoğlu, 2010;Okkan ve Dalkılıç, 2012;Kızılaslan ve diğerleri, 2014;Altunkaynak ve Başakın, 2018) olarak belirtilebilir. ...
Su kaynaklarını korumak ve verimli bir biçimde kullanabilmek için mevcut su kaynaklarının gelecekteki miktarlarının iyi tahmin etmek gerekmektedir. Akarsuların akımlarının doğru tahmini, baraj haznelerinin kapasitesinin belirlenmesi, taşkın koruma yapılarının tasarımı, kuraklık ve havza yönetimi, doğal hayatın korunması gibi birçok toprak ve su kaynakları yönetimi projeleri için oldukça önemlidir. Bu çalışmada, ileri beslemeli geri yayılımlı yapay sinir ağı modeli geliştirerek, Fırat Havzası'nın başlıca su kaynaklarından olan Karasu Nehrine ait günlük akımlara uygulanmıştır. Bu amaç çerçevesinde, DSİ tarafından işletilen D21A001 numaralı Karasu Karagöze akım gözlem istasyonuna ait 01 Ekim 2000-30 Eylül 2015 tarihleri arasında gözlenen 16 yıllık 5844 adet günlük akım verisi kullanılmıştır. Tahminler yapılırken 6 adet senaryo geliştirilmiş ve ağın eğitiminde sırasıyla Levenberg-Marquardt ve Bayesian Regularization algoritmaları kullanılmış ve her iki yöntemden benzer sonuçlar elde edilmiştir. Senaryolar değerlendirilirken determinasyon katsayısı (R 2), Ortalama Karesel Hata (MSE) ve Ortalama Mutlak Hata (RMSE) değerleri göz önünde bulundurulmuştur. Kurulan YSA modelinin Karasu Karagöze akımlarını başarıyla temsil ettiği ve bu çalışmanın taşkın ve su tutulması çalışmalarına ışık tutacağı düşünülmektedir. ABSTRACT In order to protect water resources and to use them efficiently, it is necessary to accurately estimate available water resources in the future. Accurate estimating of river flowrates is very important for many soil and water resource management projects such that determination of the capacity of the dam reservoirs, design of flood protection structures, drought and watershed management and conservation of natural life. In this study, a multi-layer perception Artificial Neural Network (ANN) trained with the backpropagation algorithm is adopted to make daily river flow forecasts on the Karasu River which is the one of main water resource for Fırat Basin. For this purpose, 5844 daily flow data which is observed between 01 October 2000 and 30 September 2015 of D21A001 Karasu Karagöze gauging station operated by DSİ has been used. While the estimations were made, 6 scenarios were developed and Levenberg-Marquardt and Bayesian Regularization algorithms were used for training of the ANN respectively, and similar results were obtained for both methods. While evaluating the scenarios, the coefficient of determination (R2), the mean square error (MSE) and the mean absolute error (RMSE) values were taken into consideration. It is thought that the ANN model has successfully represented Karasu Karagöze flows and this study will be helpful to future studies in flood and water storage capacity.
... Many data-driven models including the autoregressive (AR), and autoregressive moving average (ARMA) models (Raman and Sunilkumar 1995;Liu et al. 2015) have been applied in hydrological forecasting since 1970. However, these models assume linearity of the input-output relation of the streamflow time series data while this relation is indeed largely nonlinear. ...
Streamflow estimation is highly significant for water resource management. In this work, we improve the accuracy and stability of streamflow estimation through a novel hybrid decompose-ensemble model that employs variational mode decomposition (VMD) and back-propagation neural networks (BPNN). First, the latest decomposition algorithm, namely, VMD, was used to extract multiscale features that were subsequently learned and ensembled by the BPNN model to obtain the final estimate streamflow results. The historical daily streamflow series of Laoyukou and Wushan hydrological stations in China were analysed by VMD-BPNN, by a single GBRT and BPNN model, ensemble empirical mode decomposition (EEMD) models. The results confirmed that the VMD outperformed a single-estimation model without any decomposition and EEMD-based models; moreover, ensemble estimations using the BPNN model development technique were consistently better than a general summation method. The VMD-BPNN model’s estimation performance was superior to that of five other models at the Wushan station (GBRT, BPNN, EEMD-BPNN-SUM, VMD-BPNN-SUM, and EEMD-BPNN) using evaluation criteria of the root-mean-square error (RMSE = 2.62 m³/s), the Nash–Sutcliffe efficiency coefficient (NSE = 0. 9792) and the mean absolute error (MAE = 1.38 m³/s). The proposed model also had a better performance in estimating higher-magnitude flows with a low criterion for MAE. Therefore, the hybrid VMD-BPNN model could be applied as a promising approach for short-term streamflow estimating.
... Based on the human brain neural network, an intelligent computational system was proposed, which called the ANN model. This model comprises a combination of information processing units called neurons in which one or more inputs are received, and the outputs are determined using the predetermined non-linear function (Raman and Sunilkumar 1995). Nowadays, the ANN model solves many problems due to unique features (Ferench et al. 1992). ...
One of the most important and effective works of water resource planning and management is determining the specific, applicable, regulated operating policies of the Zayandehroud dam reservoir, as a case study, in which it should be user-friendly and straightforward for the operator. For this purpose, different methods have been proposed in which each of them has its limitations. Due to the unique capabilities of the genetic programming (GP) model, here, this method is used to determine the operating rule curves and policies of the dam reservoir. For this purpose, here, two cases are proposed in which, in the first case, each month is individually simulated and modeled. However, in the second case, all months are simulated simultaneously. A second case is proposed here to determine simple and more applicable operation rule curves. In addition, two approaches are suggested for each case in which in the first approach, the influential input variables are selected by presenting the hybrid method. In the proposed hybrid method, the artificial neural network (ANN) model is equipped with non-dominated sorting genetic (NSGA-II) algorithm leading to a hybrid method named the ANN-NSGA-II method. However, in the second approach, the influential input variables are selected automatically using the GP method. Here, the hybrid method is proposed and used to overcome the limitations of existing usual method. In other words, it is proposed to reduce the number of influential input variables of data-driven methods and select the effective ones. The obtained results of all proposed cases and approaches are presented and compared with the standard operation policy method, stochastic dynamic programming, ANN model, and NLP method. Comparison of the results shows the acceptable performance of the proposed cases and approaches. In other words, the best-obtained values of (stability index) SI index and water deficit (objective function value) are 49.3% and 32, respectively.
... It develops a memory which is able to associate a great number of input datasets which results a set of outputs. ANN can provide the accurate solutions to many of the complex models, which can be treated as a global approximation model (Cancelliere et al. 2002;Dogan et al. 2007;Karunanithi et al. 1994;Raman and Sunilkumar 1995). ANN model is the most excellent for the dynamic nonlinear modeling system (Sajikumar and Thandaveswara 1999). ...
This book addresses the various challenges in achieving sustainable groundwater development, management, and planning in semi-arid regions, with a focus on India, and discusses advanced remote sensing and GIS techniques for the estimation and management of groundwater resources. The book is timely as there is a need for a better understanding of the various tools and methods required to efficiently and sustainably meet the growing demand for clean surface and groundwater in developing countries, and how these tools can be combined with other strategies in a multi-disciplinary fashion to achieve this goal in water-scarce regions. To wit, the book combines remote sensing and GIS techniques, runoff modeling, aquifer mapping, land use and land cover analyses, evapotranspiration estimation, crop coefficients, and water policy approaches. This will be of use to academics, policymakers, social scientists, and professionals involved in the various aspects of sustainable groundwater development, planning, and management.
... Moreover, accuracy of the newly employed AI technique as well as multiple AI techniques is compared with that of the original or proved AI technique to adjudge capability and comment on reliability of the former in forecasting streamflows. Few of the earlier studies made a comparison between ANN and stochastic time series approaches in making streamflow forecasts (e.g., Raman and Sunilkumar, 1995;Thirumalaiah and Deo, 2000;Kişi, 2003Kişi, , 2005. Later on, with start of the ANN applications in streamflow forecasting, performance of the ANN was compared with that of conventional ARIMA techniques (e.g., Jain et al., 1999;Castellano-Méndeza et al., 2004;Valipour et al., 2013). ...
Reliable and realistic streamflow forecasting is very important in hydrology, hydraulic, and water resources engineering as it can directly affect the dams operation and performance, groundwater recharge/exploitation, sediment conveyance capability of river, watershed management, etc. However, an accurate streamflow forecasting is not an easy task due to the high uncertainty associated with climate conditions and complexity of collecting and handling both spatial and non-spatial data. Therefore, hydrologists from all over the world have developed and adopted several types of data-driven techniques ranging from traditional stochastic time-series modeling to modern hybrid artificial intelligence models for future prediction of streamflow. In literature, studies dealing with streamflow forecasting used a variety of techniques having dissimilar concepts and characteristics, and streamflow datasets at different time scale such as daily, monthly, seasonal and yearly etc. This chapter first describes and classifies available data-driven techniques used in streamflow forecasting into suitable groups depending upon their characteristics. Then, growth of the salient data-driven models both single and hybrid such as time-series models, artificial neural network models, and other artificial intelligence models is discussed with their applications and comparisons as reported in studies on streamflow forecasting over time. Thereafter, current approaches used in the recent five-year streamflow-forecasting studies are briefly summarized. Also, challenges experienced by the researchers in applying data-driven techniques for streamflow forecasting are addressed. It is concluded that a vast scope exists for improving streamflow forecasts using emerging and modern tools and combining them with location-specific and in-depth knowledge of the physical processes occurring in the hydrologic system.
... It develops a memory which is able to associate a great number of input datasets which results a set of outputs. ANN can provide the accurate solutions to many of the complex models, which can be treated as a global approximation model (Cancelliere et al. 2002;Dogan et al. 2007;Karunanithi et al. 1994;Raman and Sunilkumar 1995). ANN model is the most excellent for the dynamic nonlinear modeling system (Sajikumar and Thandaveswara 1999). ...
... It develops a memory which is able to associate a great number of input datasets which results a set of outputs. ANN can provide the accurate solutions to many of the complex models, which can be treated as a global approximation model (Cancelliere et al. 2002;Dogan et al. 2007;Karunanithi et al. 1994;Raman and Sunilkumar 1995). ANN model is the most excellent for the dynamic nonlinear modeling system (Sajikumar and Thandaveswara 1999). ...
This chapter describes Artificial Neural Network optimization technique and Wavelet Support Vector Machine (WA-SVR) model for predicting the groundwater level in the given terrain. Ground water forecasting is significantly needed for groundwater management. The most important purpose of applying the neural network (artificial) was to extract the probability of various algorithms. Now-a-days, Artificial Neural Networks (ANN) are the most widely accepted modeling technique which can approximate the relationship (nonlinear) between input datasets and output results without considering physical processes and the corresponding equations of the system. Due to this capability, an ANN model is much faster than a physically based model which it approximates. Prediction of accurate groundwater head helps in the practical and best possible usage of the water resources. The performance of the models was evaluated by using two performance measures, the correlation coefficient (R) and Nash–Sutcliffe Efficiency Index (Ef). Model accuracy has been improved using different network architectures and training algorithms. After improving the model accuracy, it has been noticed that the best results can be accomplished with a typical feedforward neural network (FFNN) trained with the Levenberg–Marquardt (LM) algorithm obtained for simulation of ground water levels. The results revealed that ANN model technique was appropriate for predicting the groundwater heads. The study conclusively confirms the ability of ANNs to give the precise estimation of the head value with fair accuracy. From the results, it has been concluded that the ANNs have simulated and predicted the water heads in the river under acceptable residuals, whereas WA-SVR model’s results are more accurate. Study concludes that wavelet decomposition-based SVR is found superior in comparison of the ANN model.
... It develops a memory which is able to associate a great number of input datasets which results a set of outputs. ANN can provide the accurate solutions to many of the complex models, which can be treated as a global approximation model (Cancelliere et al. 2002;Dogan et al. 2007;Karunanithi et al. 1994;Raman and Sunilkumar 1995). ANN model is the most excellent for the dynamic nonlinear modeling system (Sajikumar and Thandaveswara 1999). ...
In this chapter we discuss groundwater development and planning using rainwater harvesting, which can be helpful for crops in semi-arid regions. Rainwater harvesting activities have been performed on drainage lines for stored rainwater in drainage and to recharge poor aquifer zones in a given area. Before rainwater harvesting structures were installed. Drought, water changes, and availability then resulted in the need of the micro-shed area. One rainwater reaping of existing CNB (permanent rainwater reaping structures) at the semi-arid micro-shed in Barshitakli Taluka, Akola district of Maharashtra, India. A total stretch of 5000 m of drainage line (nala) was deepened and widened by necessity in order to have great rainwater storage capacity, and ten permanent structures (CNBs) were restructured. Because of this project, rainwater has been stored in permanent structures (CNBs) during the rainy season after that dry spells period is long that time farmers can be utilized the water to crop growth development. During the rabi season farmers can use rainwater stored in drainage lines and CNB structures. This is very helpful and directly impacts groundwater development and crop growth in semi-arid regions. Thanks to long-term storage, harvested water in CNBs are available for use as a form of protective irrigation for different crops during kharif and rabi seasons. In 2016, some wells were selected near rainwater harvesting structures in semi-arid watersheds.
... Among them, comparisons between ARIMA and ANN are common. Raman and Sunilkumar (1995) used ARIMA and ANN to predict reservoir inflow. The trial-anderror method was used to select the "optimal" model structure. ...
The performance comparison studies of the autoregressive integrated moving average model (ARIMA) and the artificial neural network (ANN) were mostly carried out between the selected model structures through trial-and-error, strongly influenced by model structure uncertainty. This research aims to make up for this inadequacy. First, a surface water quality prediction case study including eight monitoring sites in China was introduced. Second, the ARIMA and ANN’s performance was compared statistically between 6912 Seasonal ARIMA (SARIMA) and 110,592 feedforward ANN with different model structures, based on the mean square error (MSE) distributions depicted by boxplots. In a statistical view, the ANN models obtained a significantly lower median value and a more concentrated distribution of validation MSEs, which indicated lighter overfitting and better generalization ability. Furthermore, the optimal SARIMA models’ performance is inferior to even the median of the ANN models in the case study. In contrast with the previous comparisons among selected models, the statistical comparison in this study shows lower uncertainty.
... ANN, for example, has a variety of ductility functions including simulation prediction and nonlinear evaluation. ANN originated in the 1980s and can be described as a large-scale nonlinear system connected by a large number of artificial neurons [24]. ANN is a physical abstraction, simplification, and simulation of the human brain. ...
This work aimed to assess the water quality of the Tuojiang River Basin in recent years to provide a better understanding of its current pollution situation, and the potential pollution risks and causes. Water quality parameters such as dissolved oxygen (DO), ammonia–nitrogen (NH3-N), total phosphorus (TP), the permanganate index (CODMn), five-day biochemical oxygen demand (BOD5), pH, and concentrations of various heavy metals were measured in the Tuojiang River, according to the national standards of the People’s Republic of China. Samples were collected between 2012 to 2018 at 11 national monitoring sites in the Tuojiang River Basin. The overall water pollution situation was evaluated with back propagation artificial neural network (BP-ANN) analysis. The pollution causes were analyzed considering both industrial wastewater discharge in the upper reaches and the current pollution situation. We found potential risks of excessive NH3-N, TP, Cd, Hg, and Pb concentrations in the Tuojiang River Basin. Moreover, corresponding water pollution control suggestions were given.
... Over the few last decades, several studies have explicitly explored the merits and disadvantages of physically-based models and evaluated their prediction performance with that of the emerging artificial intelligence models. Some examples include the studies on support vector machine (SVM) (Yoon et al. 2011;Yu et al. 2006), and artificial neural networks (ANN) (Raman and Sunilkumar 1995;Seo and Kim 2016) as commonly used data-driven methods. Very few studies, however, have focussed on soil moisture prediction, with examples being those on applying artificial neural network (ANN) (Huang et al. 2010), extreme learning machine (ELM) (Prasad et al. 2018) and multivariate relevance vector machines (Zaman and McKee 2014). ...
Future soil moisture (SM) estimation is a practically useful task for eco-hydrologists, agriculturists, and stakeholders in environment health monitoring to generate comprehensive understanding of hydro-physical and soil dynamic system. This paper demonstrates the capability of a hybridised long short-term memory (LSTM) predictive framework to emulate SM under global warming scenarios. The proposed model is developed by integrating Boruta-random forest (BRF) feature selection and capturing significant antecedent memory of SM behaviour were applied to estimate the future SM using Coupled Model Intercomparison Phase-5 (CMIP5) repository. The BRF is adapted to extract pertinent features in hydro-meteorological variables intrinsically related to SM, and therefore, is used to construct a hybridised deep learning (i.e., BRF-LSTM) model. To establish the viability of deep learning model for SM estimation until 2100, five stations closely matched to the global climate model grid are selected in Australia's Murray Darling Basin. The performance skill of BRF-LSTM model is compared against standalone models (i.e., LSTM, SVR, and MARS). The results showed that the hybrid deep learning model (i.e., BRF-LSTM) with a feature selection capability could significantly outperform the standalone models for both warming simulations. The proposed hybrid model also demonstrated superiority in SM estimation with over 95% of all predictive errors lying below 0.02 mm, and low relative root means square error (≈ 1.06% for RCP4.5 and ≈ 1.888% for RCP8.5) to outperform all the benchmark models. This study demonstrates the capability of LSTM algorithm coupled with BRF feature selection to simulate future soil moisture under climate change, and so, can be successfully implemented in hydrology, agriculture, soil use management and environmental management.
... Menurut French et al. (1992) dalam makalah Raman et al. (1995) ada beberapa keuntungan lain dalam menggunakan ANN antara lain : ...
Abstrak. Analisa ketersediaan air memerlukan data debit yang relatif cukup panjang dan kontinyu. Jika data debit yang tersedia tidak cukup panjang dan kontinyu maka diperlukan model hubungan hujan-limpasan. Salah satu model yang tersedia, dimana parameter-parameter yang membentuk suatu persamaan yang memiliki arti fisik yang mengilustrasikan air permukaan dan air tanah pada suatu sungai dikenal dengan NRECA (Non Recorded Catchment Area). Pemodelan yang lain, diantaranya adalah Artificial Neural Network (ANN) yang merupakan elemen-elemen yang saling berhubungan dengan simpul-simpul dan dianalogikan sebagai neuron. Neural network memiliki lapisan input dan output yang saling berhubungan dan akan memiliki pola arsitektur berbeda-beda tergantung dari kendala dalam suatu permasalahan. Data hujan bulanan merupakan input sedangkan data debit bulanan merupakan variabel output. Perbedaan hasil simulasi debit bulanan yang dihasilkan dengan model ANN atau Nreca dan hasil debit pengamgatan akan ditentukan berdasarkan error mutlak rata-rata yang disebut KAR (Kesalahan Absolut Rata-rata). Studi kasus yang digunakan adalah data hujan bulanan dan debit di Cikapundung-Gandok selama kurun waktu 30 tahun.Untuk NRECA, semakin besar error yang diperoleh KAR, semakin besar deviasi debit yang terjadi. Dalam studi ini dapat disimpulkan bahwa secara umum NRECA memberikan hasil yang lebih baik kecuali untuk kondisi low flow.Abstract. The analysis of water availability needs continuous and long discharge data. If the data is not long and continuous thus the rainfall-runoff model is needed. One of the available models, where parameters forming equation which have physical meaning illustrate ground water and surface runoff in the river, is NRECA (Non Recorded Catchments Area). Another model, an artificial neural network (ANN) is a computing system made up of a highly interconnected set of simple information processing elements, analogous to a neuron, called units. A neural network has an input layer, a hidden layer and an output layer. Each layer is made up of several nodes, and layers are interconnected by sets of correlation weights. The pattern of connectivity and the number of processing units in each layer may vary within some constraints. The input used is monthly rainfall whereas the output is monthly discharge. The difference between simulated discharge produced by ANN or NRECA and observed discharge is determined by mean absolute error, namely KAR (Kesalahan Absolut Rata-rata). For case study, 30 year monthly rainfall and discharge at Cikapundung-Gandok are used. For NRECA, the more error on KAR, the more deviation of discharge will be, particularly which is under average for dependable discharge as well as mean annual minimum discharge. It is concluded that in general NRECA is better but especially for low flow ANN model is leading.
... In addition, it is assumed that all measurement data at the SBJ in the frequency range below 0.2 Hz can be modelled by a non-linear model (NAR network). For the investigations the non-parametric method Frequency Domain Decomposition [6] and the multilayer perceptron corresponding to Ref. [4,7,17] are applied. With the NAR network different measurement data and time-related dependencies can be linked via the non-linear regression model and tested for changes. ...
Suction bucket foundations for offshore wind turbines (OWT) have considerable advantages compared to conventional foundation types: Due to the installation process with dead weight and applied negative pressure, noise from pile driving can be completely avoided. In addition, the installation process of the whole substructure, consisting of the buckets connected to the jacket, can be carried out in one work step, which increases efficiency. A prototype of the suction bucket jacket was installed in the wind park ‘Borkum Riffgrund 1’ (North Sea) in August 2014. Due to the pre-installed and comprehensive measuring system, it was possible to monitor all installation and operating phases. The data analysis of a storm event show an amplitude and frequency-dependent behaviour of the soil stiffness and the suction bucket foundation without wind turbine. In the frequency range of the first and second eigenfrequency (0.2 Hz < f < 5 Hz), the system behaves linearly. Here, the Frequency Domain Decomposition is used for identification and monitoring. For the lower frequency band (0.05 Hz < f < 0.2 Hz) where higher forces and displacements occur, a non-linear multilayer perceptron is chosen to model the non-linear relations between measurements. By applying two mathematical models for the relevant frequency ranges, all the information from the measurement data can be used for system identification and novelty detection under varying environmental conditions.
... These studies proposed that the EANN was more effective than a Single ANN (SANN) or other existing physical approaches. Moreover, cross-validation techniques have been widely used for different hydrologic variables to assess the estimates obtained from hydrological models [23][24][25][26]. ...
The present work aimed to examine the feasibility of using artificial neural network (ANN) based models to obtain accurate estimates of nitrate loads in river basins, which is an important parameter for water quality management. Both Single ANN (SANN) and Ensemble ANN (EANN) models were used to obtain the load estimations for five river basins in the Midwest United States. These basins included the Cuyahoga, Raisin, Sandusky, Muskingum, and Vermilion basins in Michigan and Ohio. Further, canonical correlation analysis (CCA) was applied to the ANN models to improve the performance. The k-fold cross-validation method was then utilized to evaluate the proposed models based on two statistical indices, namely, the rRMSE and rBAIS, and the estimates were compared for four different k values (k = 3, 5, 7, and 10). According to the results, the EANN model seemed to produce better load estimations than the SANN model, and the CCA based EANN model tended to produce the best estimates among all of the proposed models in this study. The box plot data for the rRMSE index were also investigated, and the plot results indicated that increasing values of k tended to generate better estimates. Thus, the use of k = 10 is recommended for load estimations since this value was associated with better performances and less biased estimates.
... Authors of the work [22] presented an eight-step procedure to design a neural network forecasting model for financial and economic time series. In [47], the authors proposed the use of FFNN for inflows synthesis. FFNN offers a viable alternative also for multivariate modeling of water resources time series. ...
In this paper, a novel algorithm called opposition-based coral reefs optimization (OCRO) is introduced. The algorithm is built as an improvement for coral reefs optimization (CRO) using opposition-based learning (OBL). For efficient modeling as the main part of this work, a novel time series forecasting model called OCRO-multi-layer neural network (MLNN) is proposed to explore hidden relationships in the non-linear time series data. The model thus combines OCRO with MLNN for data processing, which enables reducing the model complexity by faster convergence than the traditional back-propagation algorithm. For validation of the proposed model, three real-world datasets are used, including Internet traffic collected from a private internet service provider (ISP) with distributed centers in 11 European cities, WorldCup 98 contains request numbers to the server in football world cup season in 1998, and Google cluster log dataset gathered from its data center. Through the carried out experiments, we demonstrated that with both univariate and multivariate data, the proposed prediction model gains good performance in accuracy, run time and model stability aspects as compared with other modern learning techniques like recurrent neural network (RNN) and long short-term memory (LSTM). In addition, with used real datasets, we intend to concentrate on applying OCRO-MLNN to distributed systems in order to enable the proactive resource allocation capability for e-infrastructures (e.g. clouds services, Internet of Things systems, or blockchain networks).
Related materials and implementation: https://github.com/chasebk/code_ocro_mlnn
... ANNs can be trained (i.e. supervised learning) to perform various operations including, but not restricted to, recognition tasks (Cheng et al., 2010;Pozzi, Gamba, & Giacoma, 2010), clustering (Chon, Park, Moon, & Cha, 1996) or nonlinear statistical modelling with predictive purposes (Raman & Sunilkumar, 1995;Tu, 1996). ...
It is generally accepted that comparative studies on animal communication can provide insights into the coevolution of social life, vocal communication, cognitive capacities and notably the emergence of some human language features. Recent studies suggested that non-human primates possess combinatorial abilities that may allow a diversification of vocal repertoires or a richer communication in spite of limited articulatory capacities. However, the functions of combined calls and the information that receivers can extract remain poorly understood. This thesis investigated call combination systems in two species of guenons: Campbell’s monkey (Cercopithecus Campbelli) and Diana monkey (Cercopithecus Diana). Firstly, I studied the combinatorial structure and relevance to receivers of combined calls in of both species using playback experiments. Results confirmed the presence of a suffixation mechanism reducing the emergency of danger signaled by calls of male Campbell’s monkeys. Also, they showed that combined calls of females Diana monkeys convey linearly information via their two units, which signal respectively caller’s emotional state and identity. Secondly, focusing on the context associated with the emission of simple and combined female Campbell’s monkey calls, results revealed flexible use of combination reflecting the immediate need to remain cryptic (i.e. simple calls) or to signal caller’s identity (i.e. combined calls). Finally, I compared females’ communication systems of both species to identify their similarities and differences. As predicted by their close phylogenetic relatedness, their repertoires are mostly based on homologous structures. However, the females differ strongly in their use of those structures. In particular, the great number of calls combined by Diana monkeys increases considerably their vocal repertoire compared to Campbell’s monkeys. Given that the combinations are non-random, meaningful to receivers and used flexibly with the context, I propose a parallel with a rudimentary form of semantic morphosyntax and discuss more generally the possible existence of similar capacities in other non-human animals.
... У [23] је применом вишеслојних неуронских мрежа без повратних веза вршена предикција годишњих протока. Упоредна анализа добијених протока који су предвиђани помоћу двадесет различи-68 • 31 st INTERNATIONAL CONGRESS ON PROCESS INDUSTRY тих неуронских мрежа приказани су у [24]. Увођењем хибридних модела [25] и комбинујући анализе помоћу SWAT (Soil and Water Assessment Tool) и вештачких неуронских мрежа, [26], показује се тежња ка унапређењу поузданости предвиђања дневних и недељних протока. ...
The development of renewable energy systems requires the use of sophisticated techniques for an accurate estimation of the available energy potential and for effective control and optimization of systems operation. The common feature of artificial intelligence methods is that they employ computer systems to perform tasks which require intelligent behavior, such as learning, reasoning, problem solving and decision making under uncertainty. This can be particularly beneficial in modeling, analysis, optimization and prediction of the performance and control of renewable energy systems and more efficient energy use. These systems are highly nonlinear, complex and dynamic, where the underlying physical relationships are not fully understood and the available data are often noisy and/or incomplete. Multi-parameter and multicriteria aspects of the design of these systems are not easily handled using analytical methods, physical
models or numerical methods. Artificial intelligence techniques may provide a promising and reliable
alternative, or a complement, to the traditional process-based or statistical approaches used in the energy efficiency and renewable energy systems. They enable to study these systems without any knowledge of the exact relations governing their operation, and once trained, allow performing as complex tasks as prediction, modeling, identification, optimization, forecasting and control. Artificial neural networks and support vector machine, as commonly used artificial intelligence methodologies, and their application in the energy efficiency and renewable energy systems are presented.
... Od tada pa do danas UNM-e su pronašle veliku primjenu u području hidrologije, a naročito u izradi modela koji opisuju odnos oborina i otjecanja. U tome su pridonijeli Halff i suradnici upotrebom UNM-a u predviđanju hidrograma, a Rayman i Sunil Kumar [11] predstavljaju model procjene mjesečnih oborina. Od tada pa do danas upotreba UNM-a u području hidrologije je sve učestalija. ...
Detailed steps of a hydrological model implementation methodology, as based on application of the artificial neural network (ANN) in small catchments, is presented in this paper. The motivation for this paper arises from the established lack of precise steps and procedures in the existing ANN-based methodologies for development of hydrological models. The implementation of hydrological discharge model based on ANN in a small catchment of Slani potok served as a basis for the development of detailed procedural steps for this methodology.
... In the last decade, many studies have investigated the advantages and disadvantages of various physically based models and evaluated their prediction performance with that of emerging data-driven, machine learning methods. The machine learning methods include Multiple Linear Regression (MLR) (Sahoo and Jha, 2013), Support Vector Regression (SVR) (Yu et al., 2006;Yoon et al., 2011;Belayneh et al., 2014;Mirzavand, 2015), and Artificial Neural Networks (ANN) (Raman and Sunilkumar, 1995;Daliakopoulos, 2005;Sarangi et al., 2006;Napolitano et al., 2011;Parchami-Araghi et al., 2013;Seo et al., 2015;Chang et al., 2016). These statistical methods explore the spatial and temporal patterns hidden in historical data without using a physical model because the latter always requires a large number of physical parameters and a deep understanding of the physical processes of the modeling domain. ...
Predicting water table depth over the long-term in agricultural areas presents great challenges because these areas have complex and heterogeneous hydrogeological characteristics, boundary conditions, and human activities; also, nonlinear interactions occur among these factors. Therefore, a new time series model based on Long Short-Term Memory (LSTM), was developed in this study as an alternative to computationally expensive physical models. The proposed model is composed of an LSTM layer with another fully connected layer on top of it, with a dropout method applied in the first LSTM layer. In this study, the proposed model was applied and evaluated in five sub-areas of Hetao Irrigation District in arid northwestern China using data of 14 years (2000-2013). The proposed model uses monthly water diversion, evaporation, precipitation, temperature, and time as input data to predict water table depth. A simple but effective standardization method was employed to pre-process data to ensure data on the same scale. 14 years of data are separated into two sets: training set (2000-2011) and validation set (2012-2013) in the experiment. As expected, the proposed model achieves higher R² scores (0.789-0.952) in water table depth prediction, when compared with the results of traditional feed-forward neural network (FFNN), which only reaches relatively low R² scores (0.004-0.495), proving that the proposed model can preserve and learn previous information well. Furthermore, the validity of the dropout method and the proposed model’s architecture are discussed. Through experimentation, the results show that the dropout method can prevent overfitting significantly. In addition, comparisons between the R² scores of the proposed model and Double-LSTM model (R² scores range from 0.170 to 0.864), further prove that the proposed model’s architecture is reasonable and can contribute to a strong learning ability on time series data. Thus, one can conclude that the proposed model can serve as an alternative approach predicting water table depth, especially in areas where hydrogeological data are difficult to obtain.
In this study an Artificial Neural Network for the simulation of flood phenomena in a natural area was developed. Then this network was implemented in the urban area of a Greek city (Amyntaio, Florina). The neural networks have many advantages: non-linearity, adaptability, input-output mapping, indicative response, damage resistance, possibility of implementation with VLSI (Very Large Scale Integration) technology, content related information and analysis and design uniformity. With neural networks, mathematical simulation of the considered phenomenon is not attempted, but the extraction of quantitative conclusions for specific data, based on similar cases. With the development and implementation of this network all the points that are in risk for flood are identified. The results showed that the help of an Artificial Neural Network in these cases is crucial for the future decisions in cases of flood phenomena.
Due to the influence of human regulation and storage factors, the runoff series monitored at the hydro-power stations often show the characteristics of non-periodicity which increases the difficulty of forecasting. The prediction model based on the neural network can avoid the interference of the non-periodicity by focusing on the relationship between rainfall input and runoff output. However, the physical correlation of the rainfall-runoff and the complexity of the neural network still flaw the subdivision research. In this paper, an improved convolutional neural network (CNN) was innovatively constructed to model runoff prediction, which contains effective layers design and adaptive activation function. The long-term and irregular observation data collected by the Zhexi reservoir were used for training and validation. In addition, the models based on traditional artificial neural networks and ordinary CNN were applied to the forecast simulation for contrast. Evaluation results using real data indicated that the improved CNN model performs better in these acyclic series, with over 0.9 correlation coefficient values and under 185 root means square error values during the validation, meanwhile averting the gradient vanishing and negative discharge problems occurring in other models. Numerous indicators and plots prove the excellent effect and reliability of the model forecast. Considering the robustness and validity of the neural network, this research and verification are of significance to non-periodic reservoir inflow prediction.
Assessment and Estimation of Flood Discharge for given return period is utmost important for planning and Design of Hydraulic Structures in the project site. But as it involves various parameters such as temperature, relative humidity, rate of evaporation, discharge, precipitation, topographical usage etc. which are highly unsteady, thus making it one of the most difficult task to have an accurate predictability. Through Artificial Neural Network, we have estimated flood discharge of next day. We used 3 set of input data viz. rain gauge, temperature and discharge for Phulgaon area (Pune District) Maharashtra. Through black box technique, all the input variables will be given set of weights and bias, and after training of those data sets, various network architecture would be created and among them best one would be selected comprising of lowest possible errors and best predicted output. These evaluations would be based on RMSE (Root Mean Square Error) and Hydrograph analysis, thus making it effective for pre or post management of disaster as well.
Nowadays, modelling, automation and control are widely used for Water Resource Recovery Facilities (WRRF) upgrading and optimization. Influent generator (IG) models are used to provide relevant input time series for dynamic WRRF simulations used in these applications. Current IG models found in literature are calibrated on the basis of a single performance criterion, such as the mean percentage error or the root mean square error. This results in the IG being adequate on average but with a lack of representativeness of, for instance, the observed temporal variability of the dataset. However, adequately capturing influent variability may be important for certain types of WRRF optimization, e.g., reaction to peak loads, control system performance evaluation, etc. Therefore, in this study, a data-driven IG model is developed based on the long short-term memory (LSTM) recurrent neural network and is optimized by a multi-objective genetic algorithm for both mean percentage error and variability. Hence, the influent generator model is able to generate a time series with a probability distribution that better represents reality, thus giving a better influent description for WRRF design and operation. To further increase the variability of the generated time series and in this way approximate the true variability better, the model is extended with a random walk process. HIGHLIGHTS
A data-driven influent generator based on the long short-term memory (LSTM) is proposed.;
To represent both the average behaviour and variability of the influent, the IG is trained by a multi-objective genetic algorithm.;
The model is extended with a random walk stochastic process with a probability distribution that better represents reality, giving a better influent description for model-based WRRF design and operation.;
Modelling of rainfall-runoff is considered one of the prerequisite of hydrological processes for various applications involving conservation and management of water resources. In this study, two techniques that is Multi-Layer Perceptron (MLP) neural network, which is well known efficient Artificial Neural Network (ANN), and Multi-Linear Regression (MLR) were applied for modelling daily rainfall-runoff and results obtained were compared. In order to simulate the processes, time series monsoon data of ten years (2000-2009) of rainfall and runoff at Bino watershed in Almora and Pauri Garhwal districts of Uttarakhand, India were used. In addition, Gamma Test (GT) was used for identifying the best input combinations for rainfall-runoff modelling. Performance of models was evaluated qualitatively as well as quantitatively employing statistical indices viz. correlation coefficient (r), root mean square error (RMSE) and coefficient of efficiency (CE), both for training as well as testing. Different MLP based ANN models were developed with the change of number of neurons and hidden layers and best model among them was selected based on performance indices. The same inputs were used to develop MLR model. The r, RMSE and CE values of best performing MLP model were found to be 0.95, 1.27 (mm) and 0.88, respectively during training while their corresponding values during testing were determined to be 0.92, 0.96 (mm) and 0.80. The comparison of both MLP and MLR models reveals that MLP based ANN is superior in performance for rainfall-runoff modelling and able to predict the daily runoff with good accuracy for the study area.
River-flood forecasting is among the most important feasible non-structural approaches used in reducing economic losses and alleviating human sufferings. In spite of uncertainty in the forecasting of natural disasters, the current prevailing methods developed in many parts of the world in the recent history has made good progress to a great extent. The advancement is attributed mainly due to the availability of high-resolution weather data and the use of sophisticated computer modelling algorithms. However, it is desirable to conduct exploratory review studies to further improving the current state of affairs. The present paper reviews briefly the river-flood forecasting methods currently used worldwide with a specific focus in the context of the Kelantan River in Malaysia. Flooding in Malaysia is recurrent covering a large inhabited area compared with other natural disasters. Some of the popularly used methods in the literature such as statistical methods machine learning and methods based on chaos theory have been reviewed, The paper will also attempt to explore the future direction for research and development that might be useful specifically for dealing with the recurrent rivers flooding in Malaysia. A reasonably acceptable prediction of river streamflow is significantly important in disaster management and water resources management.
Due to the influence of human regulation and storage factors, the runoff series monitored at the hydropower stations often show the characteristics of non-periodicity, which makes runoff prediction simulation difficult. This paper attempts to construct an improved one-dimensional convolutional neural network (CNN) model for runoff prediction simulation. The improved CNN model consists of two convolution layers and a full connection layer and uses LeakyRelu as the activation function. Based on the historical rainfall and runoff data of the ZheXi reservoir in Hunan Province, this paper uses the improved CNN model to simulate runoff prediction and compares the results with the traditional ANN model and the traditional CNN model. The results show that the improved CNN model is superior to the traditional ANN model and the traditional CNN model. It proves that the improved CNN model is suitable for the non-periodic runoff prediction simulation, and it can avoid the data problems such as gradient disappearance that may occur in the traditional neural network model.
This study proposes a density prediction model for reservoir sediment deposition using artificial neural networks (ANNs). To compute the reservoir capacity loss, it is necessary to estimate the weight per unit submerged sediment volume as the sediment transported in a river is measured in gravimetric terms as sediment load. Understanding the sensitivity of the estimated density to relate catchment sediment yield with the reservoir deposition rate, ANNs are utilized to precisely compute submerged sediment density. A dataset with 262 field observed densities for the reservoirs which always remain submerged is prepared and used. Three input variables, sand, silt, and clay proportion are selected. Then, the method of training and validation of the ANNs for the density prediction is presented. The model results show that the ANN model is flexible and robust to capture the complex physical process and is better than the Lara Pemberton empirical relationship. The proposed trained network, having the best predictive capability, is given as a MATLAB code.
In recent years, the Artificial Intelligence and Data Mining (AI&DM) models have become popular tools in assisting various aspects of reservoir operation. However, the practical uses are still rarely reported. Comparison experiment of many AI&DM models over a large number of reservoir cases is particularly valuable to help reservoir operators first examine the usefulness and transferability of different AI&DM models, and then identify the most stable and reliable AI&DM model in assist of various decision-making processes. In this study, a total of 12 AI&DM models with different parameterizations and simulation scenarios are comprehensively tested out and compared in simulating the controlled reservoir outflows of 33 reservoir cases over the Upper Colorado Region, United States. Results show that the Random Forecast and the Long-Short-Term-Memory model could consistently derive the best statistical performances than other models under the baseline simulation scenario. The employed AI&DM models could obtain satisfactory statistical interquartile ranges (25-75%) between [0.6 to 0.9], [0.3 to 0.8], and [0.2 to 0.8], for CORR, NSE, and KGE measurements, respectively, and [1.5 to 6.5], [-15 to 20], and [0.5 to 8.5] for the normalized RMSE, PBIAS and RSR measurements, respectively. Results also show Multi-Layer Perceptron model and Extreme Gradient Boosting Tree Algorithm produced more stable and superior performances than other models under more complex input scenarios. We also found that the performances of different AI&DM models are closely relevant to the reservoir elevations, sizes, and functionalities. Discussions were made about the sensitivity of AI&DM models’ parameterizations and the key advantages of AI&DM models over the rule-based reservoir models. We further identify that the main advantage of AI&DM models is the flexibility in designing input structures, whereas the rule-based simulation model is rather limited. Future studies were suggested regarding the best way reservoir operators and researchers could use, select, and apply different AI&DM models in simulating reservoir releases under different natural and modeling environments. This comparison study also serves as a reference and a piece of groundwork for further promoting the practical uses of AI&DM models in assisting reservoir operation.
Climate change continues to change the course of streamflow regimes. River basins around the world are experiencing frequent floods and/or droughts due to climate change. To carry out robust water management, climate models are increasingly being used to simulate the low and high flow regimes. However, due to inherent parametric differences among the models, climate models often produce different predictions. The model selection has been used to select the best performing climate model. Besides model selection, model averaging has also been adopted in order to take advantage of the relative strengths of the respective climate models. The climate models’ streamflow is averaged in different paradigms such as Bayesian frameworks and machine learning paradigms. The objective of this chapter is to synthesize the recent application of the machine learning paradigm in averaging climate model–based streamflow simulation. The machine learning paradigms considered include the neural networks and the support vector regression models.
Flood forecasting is needed for developing appropriate measures to control flood risk, mitigate flood hazard, evacuate people from flood hazard areas, determine insurance premiums, and manage environmental and water resources systems. Data-driven techniques/models have gained a significant attention for flood forecasting in recent years, of which artificial neural network (ANN) models are perhaps the most popularly used models of the modern era. The most important step in flood forecasting, based on ANN modeling, is the selection of influential input variables. This chapter first discusses a new analytical approach of input selection based on the copula entropy (CE) method, which employs the nonlinear measure of statistical dependence and makes no assumptions about the functional form. Then, application of the new CE method is demonstrated in flood forecasting through two case studies in the Yangtze River basin, China. The distribution of flood forecasting uncertainties is also discussed, and results indicated that the assumption of normal distribution and its use are not always justified. Therefore, the copula-based method is described for simulating flood forecasting uncertainties. The new CE method for flood forecasting and simulating uncertainties presented in this chapter is of great significance for controlling flood risk. The flowchart of this chapter is given in Fig. 13.1.
Sand and Dust Storms (SDS) are known as one of the most common environmental problems in arid and semi-arid regions of the world. This phenomenon is harmful to human health as well as economically. Over the past two decades, SDS have been increasing on a local, regional and even global scale. The Euphrates Basin is recognized as one of the most active SDS sources in the world. The first step in managing this environmental phenomenon, is to identifying dust storm sources. The aim of this study is to mapping the potential sources of SDS in the Euphrates basin by using Multi-Layer Perceptron Neural Network. In the first step, long-term time series data related to key environmental parameters affecting the occurrence of SDS including: soil moisture, soil temperature, soil texture, land surface temperature, wind speed, precipitation, evapotranspiration, dusty months, land use, pressure due to population, elevation and slope were identified and used as artificial neural network model inputs. Using the visual interpretation of 2500 MODIS images in natural color composite, 190 SDS events were identified and introduced to the neural network as training points. 70% of points (133 points) and 30% of points (57 points) were used for training, testing and validation of model, respectively. After running the model, the estimated mean squared error (MSE) was equal to 0.1, which indicating acceptable accuracy of the neural network model in susceptibility mapping of SDS sources. The results show that, 147000 km2 of the basin is prone to the formation of SDS sources, which mainly include low rainfall, dry and barren areas of the basin.
The gradient descent (GD) and Levenberg–Marquardt (LM) algorithms are commonly adopted methods for training artificial neural network (ANN) models for modeling various earth system and environmental processes. The performance of these algorithms is sensitive to the initialization of their parameters. The initialization of the algorithm’s parameters for modeling different physical processes also varies process to process. However, there is a minority that tried to verify the sensitivity of the parameters of the algorithm than the sensitivity of the input data to the model. This work investigates the sensitivity of the popular ANN training algorithms to initial weights while modeling one of the earth system processes, i.e., the rainfall-runoff (RR) process. A novel methodology consisting of basic statistics for assessment of sensitivity of ANN parameters is proposed. The rainfall and flow data derived from three contrasting catchments are employed to establish the conclusions of this study. The results indicate that the RR model trained by LM algorithm is more robust in achieving performance with less variance irrespective of the existence of randomness in initialization of parameters than that of the GD trained models.
Günümüzde suya olan talebin artması, mevcut su kaynaklarının korunması ve projelendirilmesi konusunu çok önemli hale getirmiştir. Projelendirme aşamasının en uygun şekilde yapılabilmesi için veri toplama, analiz etme ve gelecek verilerin tahmini gereklidir. Gelecek verilerin tahmin edilebilmesi için pek çok yöntem geliştirilmiştir. Bu çalışmada Yeşilırmak Nehri üzerinde bulunan 3 adet istasyonun gelecek dönemdeki akış miktarlarının yapay sinir ağları ile tahmin edilmesi amaçlanmıştır. Yapay sinir ağları (YSA) ile zaman serisi analizi yapılırken geriye dönük olarak kaç dönemlik verinin kullanılması gerektiği problemin başarısını doğrudan etkilemektedir. Geriye dönük olarak bakılan veri sayısı (zaman gecikmesi) bir optimizasyon problemi olarak ele alınabilmektedir. Bu aşamada Karınca Aslanı Algoritması (KAA) kullanılarak, zaman gecikmesi boyutunun optimize edilmesi gerçekleştirilmiştir. Zaman gecikmesine ek olarak yapay sinir ağlarında kullanılan gizli katman sayısı da modele eklenerek optimize edilmiştir. Elde edilen sonuçlara bakıldığında; YSA ile birlikte sezgisel optimizasyon tekniklerinin kullanılmasının modelin başarısını artırdığı sonucuna varılmıştır. Abstract Today, the increase in demand for water has made it very important to preserve and design the existing water resources. Data collection, analysis and forecasting of future data are required for optimum project planning. Many methods have been developed to predict future data. In this study, it is aimed to estimate the flow rates of 3 stations on Yeşilırmak River by artificial neural networks. The success of the problem directly affects how many periods of data should be used retrospectively when performing time series analysis with artificial neural networks. The number of data retrospectively (time delay) can be considered as an optimization problem. At this stage, time lag size optimization was performed by using Ant lion algorithm. In addition to time delay, the number of hidden layers used in artificial neural networks has been optimized by adding to the model. When the results obtained are examined; It was concluded that the use of heuristic optimization techniques together with ANN increased the success of the model.
This paper presents the results from a study on the application of an artificial neural network (ANN) model for regional flood frequency analysis (RFFA). The study was conducted using stream flow data from 88 gauging stations across New South Wales (NSW) in Australia. Five different models consisting of three to eight predictor variables (i.e., annual rainfall, drainage area, fraction forested area, potential evapotranspiration, rainfall intensity, river slope, shape factor and stream density) were tested. The results show that an ANN model with a higher number of predictor variables does not always improve the performance of RFFA models. For example, the model with three predictor variables performs considerably better than the models using a higher number of predictor variables, except for the one which contains all the eight predictor variables. The model with three predictor variables exhibits smaller median relative error values for 2- and 20-year return periods compared to the model containing eight predictor variables. However, for 5-, 10-, 50- and 100-year return periods, the model with eight predictor variables shows smaller median relative error values. The proposed ANN modelling framework can be adapted to other regions in Australia and abroad.
A compreensão de processos hidrológicos é de grande importância na gestão de recursos hídricos e em projetos de obras hidráulicas. A modelagem hidrológica é utilizada como ferramenta de observância do comportamento de variáveis hidrológicas. Este artigo apresenta uma análise de modelos RNA (Redes Neurais Artificiais), tipo MLP (Multilayers Perceptron) na previsão de níveis de água mensais e de condições de navegabilidade, com antecedência de vinte e quatro meses, na bacia do rio Tapajós, PA. Foi usada uma MLP com quatro neurônios na camada de entrada, doze camadas escondidas e uma camada de saída, onde os pesos iniciais foram gerados aleatoriamente. Como funções de transferência foram usadas funções sigmoides na camada escondida e a função linear na camada de saída. Para treino e validação foram utilizados cento e oitenta padrões de entrada, agrupados em blocos de doze meses, com intervalos bianuais, para o período de 1976 a 2005. Para teste, trinta e seis padrões de entrada de níveis de água mensal, para o período de 2006 a 2008. O modelo RNA apresentou desempenho satisfatório com R² médio de 0,971 e erro quadrático médio RMSE de 0,008.
The multi-layer perceptron is a type of Statistical Neural Network (SNN) model that is more precise than the linear regression model. However, it fails to attain high precision due to the use of homogeneous Transfer Functions (TFs) which do not appropriately link the input layer to the output layer. Therefore, an alternative SNN model using heterogeneous TFs to overcome the limitations of homogeneous TFs was developed. An Adjusted Network Information Criterion (ANIC) for testing the adequacy of the SNN models was also derived. An Heterogeneous SNN (HETSNN) model was derived by the convolution of two Homogeneous SNN (HOMSNN) models: y 1 = αX+ H h=1 β h g 1 (I i=0 γ hi x i)+ e i and y 2 = αX + H h=1 β h g 1 (I i=0 γ hi x i) + e i , where y 1 and y 2 are the dependent variables, X" is a matrix of independent variables, α,β, and γ are the parameters of the network, e i is the noise normally distributed with mean 0 and variance σ 2 (e i ∼ N (0, σ 2)), g 1 (.)andg 2 (.) are the transfer functions, h = 1, 2, , H are the number of hidden units, and i = 0, 1, , I are the number of input units. One TF was used in the HOMSNN model while a convolution of two TFs was used to derive the HETSNN. Two sets of meteorological data [Amravati Hydrology Project at Manasgaon station, India from 1990 to 2004 and Nigeria Meteorological (NIMET) station, Ibadan from 1971 to 2004] were used to investigate the fit of the derived model. Thirteen sub-samples: 10, 20, 40, 60, 80, 100, 125, 150, 175, 200, 250, 300, and 400 generated from the NIMET data were used to investigate the asymptotic behaviour of the models and ANIC using the Kolmogorov-Smirnov one sample normality test. Network Information Criterion (NIC) and the derived ANIC, using Kullback's symmetric divergence, were computed for determining the adequacy of the i two models. Variance-ratio tests were carried out on the significance of the HETSNN model. The derived HETSNN model was y i = αX+ H h=1 β h g 1 (I i=0 γ hi x i)g 2 (I i=0 γ hi x i)+ u i u j ; i = 1, 2, ..., n, where e i e j ∼ N (0, σ 2). The HETSNN approached zero asymptotically faster than HOMSNN. The asymptotic behaviour of the models showed that HETSNN improved steadily over HOMSNN at an average rate of 3.0% to 15.0%. The rates of model adequacy for both HETSNN and HOMSNN using NIC were respectively 72.9% and 58.1%. Consequently, using ANIC, the rates were 67.7% and 65.0% respectively. This indicated that ANIC was an unbiased information criterion for the two models. The ANIC decayed to zero much slower than NIC with increasing sample size. The asymptotic behaviour of the ANIC also showed that as the sample size increased, ANIC approximated the standard normal distribution, N (0, 1). There was a significant difference (p < 0.05) between the variances of HET-SNN and HOMSNN. The heterogeneous statistical neural network model linked the input layer to the output layer more appropriately. The derived adjusted network information criterion performed better in model selection when compared with the network information criterion. Keywords: Transfer functions, heterogeneous statistical neural network, adjusted network information criterion. Word count: 458 ii Dedication This study is dedicated, first and foremost, to God, who knows my beginning and my glorious destiny, which no man can change. Also, it is dedicated to my wife, Mrs. J. E. Udomboso, my children, for understanding with me in all my ordeals, and finally to my supervisors, brethren in the Lord, friends and colleagues who desired and prayed for my progress out of conscience that is true, honest, and void of offence. iii
Artificial neural networks (ANNs) have proved to be an efficient alternative to traditional methods for hydrological modeling. One of the most important steps in the ANN development is the determination of significant input variables. This chapter proposes a new method based on the copula entropy (CE) theory to identify the inputs of an ANN model. Two tests are carried out for verifying the accuracy and performance of the CE method. The CE theory-based input determination methodology is applied to identifying suitable inputs of flood forecast models for case studies. Different kinds of ANN models, including multi-layer feed-forward networks (MLF), radial basis function (RBF) and general regression neural network (GRNN), are also discussed. Results indicate that the CE method could effectively identify inputs for the ANN with smallest RMSE values for training and validation data, and the ANN models coupled with the CE method perform better than the traditional linear correlation analysis (LCA) method.
A neural network is developed to forecast rainfall intensity fields in space and time; it is a three-layer learning network with input, hidden, and output layers. Training is conducted using back propagation where the input and output rainfall fields are presented to the neural network as a series of learning sets. After training is complete, the neural network is used to forecast rainfall intensity fields with a lead time of 1 h using only the current field as input. Rainfall fields are generated using a space-time mathematical rainfall simulation model, and forecasted fields are compared with the perfectly known model-produced fields. Results indicate that a neural network is capable of learning the complex relationship describing the space-time evolution of rainfall such as that inherent in a complex rainfall simulation model. One hour ahead forecasts are produced, and comparisons with true mean areal intensities and percent areal coverage indicate that in most cases the method performs well when applied to the events used in training. The neural network is used to forecast a series of events not included in the training data and is shown to perform well when a relatively large number of hidden nodes are utilized. Performance of the neural network is compared with two other methods of short-term forecasting, persistence and nowcasting.
The need to preserve both long-term and short-term variance and covariance properties of hydrologic processes has led to the development of numerous stochastic models during the past few years; Among these are the models first proposed by Mandelbrot and Wallis for preserving the Hurst phenomenon. All these models use sequential generation schemes, and some require the process to be stationary without seasonal variations. The model presented herein preserves seasonal variations and does not necessarily generate data sequentially, although the autoregressive models of any order are special cases of this model. If an annual series has been generated by this or any other model, this model may be used to generate a parallel series of seasonal events from the given series of annual events. It may further be used to generate monthly, weekly, daily, and hourly events.
The method described in this paper for the synthesis of streamflows
differs from the traditional approaches in synthetic hydrology in the
sense that it utilizes the information contained in or among the groups
of data in a streamflow record. The existense of such groups in
geophysical records, including hydrologic records, is well emphasized by
Hurst (1951). Further, in the proposed method, based on concepts of
pattern recognition, neither a basic structure nor any preconceived
model is imposed on the data; rather the data are allowed to speak for
themselves in a most `democratic' way. The preliminary details of the
method were provided in an earlier paper by Panu et al. (1978). The
intent of this paper is to describe a procedure whereby it is possible
to specify explicitly multivariate probability distribution for the
intrapattern structure and first-order Markovian dependence for the
interpattern structure in the feature prediction model (Panu et al.,
1978). The various steps involved in the construction and operation of
the model for streamflow synthesis are presented. The application of the
model for synthesizing monthly streamflow records of three Canadian
rivers exhibiting biannual cycles is explained. Statistical and
hydrological tests show that these synthetic realizations possess
relevant properties that are comparable with the corresponding
properties contained in the historical record. This article should be
read in conjunction with the previous publication by Panu et al. (1978).
The modification in this paper of the familiar Wilson-Hilferty
transformation has exactly the same mean, variance, skew, and lower
bound as the Pearson type 3 variate that it is intended to approximate.
Although the original Wilson-Hilferty approximation is adequate only at
skews of less than about 3.0, the new one remains satisfactory
throughout the range of hydrologic interest.
A part of the streamflow construction and simulation process is the
verification that a stochastic streamflow model reproduces those
statistics which by design it should reproduce; the use of unbiased
statistics is shown to be advantageous in this exercise. In addition, if
a model reproduces statistics not used in the model-parameter estimation
process, its credibility is further enhanced. These concepts are
illustrated, and the performance of a wide range of monthly streamflow
models are compared using data for the upper Delaware River basin in New
York State.
The application of neural-network computers to pattern-recognition tasks is discussed in an introduction for advanced students. Chapters are devoted to the nature of the pattern-recognition task, the Bayesian approach to the estimation of class membership, the fuzzy-set approach, patterns with nonnumeric feature values, learning discriminants and the generalized perceptron, recognition and recall on the basis of partial cues, associative memories, self-organizing nets, the functional-link net, fuzzy logic in the linking of symbolic and subsymbolic processing, and adaptive pattern recognition and its applications. Also included are C-language programs for (1) a generalized delta-rule net for supervised learning and (2) unsupervised learning based on the discovery of clustered structure. 183 refs.
Alternative approaches suggested for modeling multiseries of water resources systems are reviewed and compared. Most approaches fall within the general framework of multivariate ARMA models. Formal modeling procedures suggest a three-stage iterative process, namely: model identification, parameter estimation and diagnostic checks. Although a number of statistical tools are already available to follow such modeling process, in general, it is not an easy task, especially if high order vector ARMA models are used. However, simpler ARMA models such as the contemporaneous and the transfer-function models may be sufficient for most applications in water resources. Two examples of modeling bivariate and trivariate streamflow series are included. Alternative modeling procedures are used and compared by using data generation techniques. The results obtained suggest that low order models, as well as contemporaneous ARMA models, reproduce quite well the main statistical characteristics of the time series analyzed. It is assumed that the same conclusions apply for most water resources time series.
This paper presents a neural network approach to multivariate time-series analysis. Real world observations of flour prices in three cities have been used as a benchmark in our experiments. Feedforward connectionist networks have been designed to model flour prices over the period from August 1972 to November 1980 for the cities of Buffalo, Minneapolis, and Kansas City. Remarkable success has been achieved in training the networks to learn the price curve for each of these cities and in making accurate price predictions. Our results show that the neural network approach is a leading contender with the statistical modeling approaches.
Interest in the study of neural networks has grown remarkably in the last several years. This effort has been characterized in a variety of ways: as the study of brain-style computation, connectionist architectures, parallel distributed-processing systems, neuromorphic computation, artificial neural systems. The common theme to these efforts has been an interest in looking at the brain as a model of a parallel computational device very different from that of a traditional serial computer.
Incluye bibliografía e índice
Artificial neural net models have been studied for many years in the hope of achieving human-like performance in the fields of speech and image recognition. These models are composed of many nonlinear computational elements operating in parallel and arranged in patterns reminiscent of biological neural nets. Computational elements or nodes are connected via weights that are typically adapted during use to improve performance. There has been a recent resurgence in the field of artificial neural nets caused by new net topologies and algorithms, analog VLSI implementation techniques, and the belief that massive parallelism is essential for high performance speech and image recognition. This paper provides an introduction to the field of artificial neural nets by reviewing six important neural net models that can be used for pattern classification. These nets are highly parallel building blocks that illustrate neural net components and design principles and can be used to construct more complex systems. In addition to describing these nets, a major emphasis is placed on exploring how some existing classification and clustering algorithms can be performed using simple neuron-like components. Single-layer nets can implement algorithms required by Gaussian maximum-likelihood classifiers and optimum minimum-error classifiers for binary patterns corrupted by noise. More generally, the decision regions required by any classification algorithm can be generated in a straightforward manner by three-layer feed-forward nets.
Applied Modelling of Hydrologie Time Series Approaches to multivariate modelling of water resources time series
- J D Salas
- J W Deulleur
- V Yevjevich
- W L Lane
- J D Salas
- Q B Tabios
- P Bartolfni
Salas, J. D., Deulleur, J. W., Yevjevich, V. & Lane, W. L. (1980) Applied Modelling of Hydrologie Time Series. Water Resources Publ., Littleton, Colorado, USA. Salas, J. D., Tabios, Q. B. & Bartolfni, P. (1985) Approaches to multivariate modelling of water resources time series. Wat. Resour. Bull. 21(4), 683-708.
Approaches to multivariate modelling of water resources time series
- J D Salas
- Q B Tabios
- P Bartolfni
Salas, J. D., Tabios, Q. B. & Bartolfni, P. (1985) Approaches to multivariate modelling of water resources
time series. Wat. Resour. Bull. 21(4), 683-708.