Figure 5 - uploaded by Roberto Célio Limão de Oliveira
Content may be subject to copyright.
Rules obtained for very high price (Class 5).
Source publication
This paper proposes a short-term energy price classification model using decision tree. The proposed model does not predict the exact value of future electricity price, but the class to which it belongs, established with respect to pre-specified threshold. This strategy is proposed since for some applications, the exact value of future prices is no...
Context in source publication
Context 1
... class 3, 98.9% for class 4 and 100% for class 5. Class 5 contains the very high prices also called spikes. Generally speaking, a price spike is an abnormal market price at time t and is significantly different from the price at previous time t -1. Despite class 5 has a small number of instances, the classifier presented excellent accuracy (100% Fig. 5 shows the decision tree obtained with the combined model CART+C5.0 to North region. Due to space limitation, only rules associated with very high price (Class 5) will be presented. From this can be stayed that energy prices becomes very high in North region when Load from 5 weeks ago is below 4148MW, Load from 10 weeks ago is bigger ...
Similar publications
Classical decision trees such as C4.5 and CART partition the feature space using axis-parallel splits. Oblique decision trees use the oblique splits based on linear combinations of features to potentially simplify the boundary structure. Although oblique decision trees have higher generalization accuracy, most oblique split methods are not directly...
Citations
... A novel predictive method for energy price classification was proposed in [71]. The proposed method does not identify the exact value of electricity price, but it classifies the values to a specific threshold. ...
Energy management is an indispensable part of today’s electrical systems and Smart Grid (SG) paradigms, especially with the high penetration of Renewable Energy Sources (RES). Thus, significant attention is paid from academia and industry to foster the synergy between these two paradigms as a means to accelerate the transition to a more diverse generation portfolio that includes an unprecedented amount of RES such as Photovoltaic (often shortened as PV) energy. Due to the ever-growing electricity consumption, state-of-the-art Artificial Intelligence (AI)-based techniques play a central role in providing necessary system flexibility to deal with the bulk integration of the PV energy for power-and-energy-efficient computing. AI lies at the core of forecasting methods to enhance the power delivery service between the grid-connected PV stations and end-consumers. In other words, the futuristic power grid infrastructure should rely on accurate PV Power Forecasting (PVPF) methods as a cornerstone of achieving unit commitment and stable energy supply. Nevertheless, designing effective energy management systems is complex because it involves designing components and hardware-software interfaces across the computing stack. So ubiquitous and complex are energy management mechanisms requiring a high level of scalability and generalization potential to gratify the load needs and cope with the meteorological factors' stochastic nature in tandem with optimal power grid stability. In an effort to break this stalemate, this research aims to explore the potential AI techniques for energy management between the energy-mix on the supply side and the load side. Consequently, the present work proposes efficient techniques to tackle the instability and intermittency of PV power production and its significant impact on the load demand. First, a comprehensive overview of SG and energy management has been conducted. Next, various ML models-based PVPF have been introduced and applied to real-world scenarios to ensure an uninterrupted power supply. Afterward, innovative load forecasting methods have been proposed to cope with the volatile PV energy supply and accommodate the stochasticity of the customers' demand with efficient energy management strategies. Finally, SG stability methods have been proposed to predict the grid's state. This will aid in shaping the best strategies for preventive maintenance and risk hedging policies. This research thesis applies data science methods to the SG paradigm for effective dynamic control and management.
... A novel predictive method for energy price classification was proposed in [50]. The proposed method does not identify the exact value of electricity price but it classifies the values to a specific threshold. ...
This chapter addresses the status of artificial intelligence (AI) as a central element in smart grid (SG) while focusing on the recent progress of research on machine learning techniques to pave the future work in the SG area. The SG framework provides a performance toolkit based on information and communication technologies. The SG paradigm supervises and promotes grid operations at a high level of expertise. AI systems have attributed to the realization of sustainable development goals including the vital integration of renewable energy sources (RES). The scarcity of conventional energy resources in the near future and their increasing threats to the environment extremely require the transition toward RES. The bulk penetration of RES into the electrical grid leads to unstable and volatile power generation. The chapter presents the commonly applied AI methods to the SG system and the key elements of the evaluation procedure.
... Specifically, three feature selection techniques and four classifiers, i.e., decision tree, Bayes, multilayer perceptron, and k-nearest neighbors, were considered (Huang et al. (2012)). A different approach using decision trees has being used for predicting future electricity prices in the Brazilian electricity market (Filho et al. (2015)).This model does not predict the exact value, but the class to which it belongs respect to a pre-specified threshold. ...
Abstract— Forecasting of electricity prices using multiple features of explanatory variables can become a challenging task which may affect the forecast accuracy or understandability of models. Our study evaluates various feature selection approaches compared to a novel hybrid feature selection approach for supporting the forecasting process with artificial neural networks. The benefits using feature selection algorithms are discussed and improvement of forecasts accuracy of marginal capacity prices on primary control reserve market by different methods are demonstrated. All identified feature selection approaches work better and support the generation of higher performance forecasts than the forecast without feature selection. The proposed hybrid approach supports the most accurate forecast generation compared to the other featured selection approaches.
In this paper, an ensemble learning model, namely the random forest (RF) model, is used to predict both the exact values as well as the class labels of 24 hourly prices in the California Independent System Operator (CAISO)'s day-ahead electricity market. The focus is on predicting the prices for the Pacific Gas and Company (PG&E) default load aggregation point (DLAP). Several effective features, such as the historical hourly prices at different locations, calender data, and new ancillary service requirements are engineered and the model is trained in order to capture the best relations between the features and the target electricity price variables. Insightful case studies are implemented on the CAISO market data from January 1, 2014 to February 28, 2016. It is observed that the proposed data mining approach provides promising results in both predicting the exact value and in classifying the prices as low, medium and high.
