Figure - uploaded by Dai Xilei
Content may be subject to copyright.
Strengths and weaknesses of common machine learning models
Source publication
This study carries out a literature review on studies using machine learning (ML) models to predict occupancy and window-opening behaviour and their application in smart buildings. For occupant number/level prediction, the indoor CO2 concentration is an important predictor variable. To further improve the model performance, it is recommended to use...
Contexts in source publication
Context 1
... that, the machine learning models were trained by the paired data. In Table 1, we briefly summarize the strengths and weaknesses of popular machine learning models to help readers understand and choose the correct models. Among these models, the support vector machine, decision tree, random forest, and kNN are well implemented in the Python package [33], which can be easily used by researchers once the data are ready. ...Context 2
... can be seen that the artificial (recurrent) neural network (ANN/RNN) and logistics regression are the most popular models for the identification/prediction of occupancy and window-opening behaviour, respectively. The summary of previous studies that used machine learning models to identify/predict the occupancy and window-opening behaviour can be seen in Tables S1 and S2. The measurement durations of these studies are summarized in Table S3. ...Context 3
... that, the machine learning models 84 were trained by the paired data. In Table 1, we briefly summarize the principles, strengths and weaknesses 85 of popular machine learning models to help readers understand and choose the correct models. Among these 86 models, the support vector machine, decision tree, random forest, and kNN are well implemented in the ...Context 4
... for classification problems, the sample 127 sizes may be imbalanced among classes, which means the number of observations available for different classes occupancy and window-opening behaviour, respectively. The summary of previous studies that used machine 137 learning models to identify/predict the occupancy and window-opening behaviour can be seen in Table S1 138 and S2. The measurement durations of these studies are summarized in The studies on occupancy can be roughly classified into three parts: real time identification, future time- number of occupants or detecting whether an occupant exists inside a given space of a building based on 145 instant monitoring parameters; the future time-step prediction mainly focuses on predicting the number of 146 occupants at a subsequent time; and the occupancy profile learning tries to generalize a few typical occupancy 147 profiles according to the historical occupancy profiles. ...Similar publications
Multivariate time series prediction has important applications in the domain of energy-efficient building technology. With the buildings consuming large amounts of electrical energy, it is critical to reducing energy consumption and economic costs while ensuring a better quality of urban living standards. As sensor-actuator rich, smart buildings ar...
Reducing the energy consumption of the heating, ventilation, and air conditioning (HVAC) systems while ensuring users’ comfort is of both academic and practical significance. However, the-state-of-the-art of the optimization model of the HVAC system is that either the thermal dynamic model is simplified as a linear model, or the optimization model...
Double Skin Façade (DSF) systems have become an alternative to the environmental and energy savings issues. DSF offers thermal buffer areas that can provide benefits to the conditioned spaces in the form of improved comforts and energy savings. There are many studies conducted to resolve issues about the heat captured inside DSF. Various window con...
This paper aims to determine cost optimality between heating, ventilation and air conditioning (HVAC) systems operating with air to water heat pumps (AWHP) and water to water heat pumps (WWHP). The analysis is performed for a certain number of heat pump units with fixed and variable capacity made by four manufacturers available on European market....
End-user awareness achieved through heat metering and accounting is a well-known possibility to optimize
the energy consumption in buildings. However, to the best of the authors’ knowledge, there is no
field research, showing how the heat cost allocation influences the operation of a heating system and
which kind of additional possibilities to decr...
Citations
... The data-driven occupancy prediction modeling concerning time series can be categorized as temporal occupancy prediction. For temporal resolution, occupancy prediction models can be categorized into three categories: real-time estimation, future prediction, and occupancy profile modeling [23]. The temporal-based occupancy prediction could be short-term or long-term with respect to the prediction horizon. ...
Accurate occupancy prediction in smart buildings is crucial for optimizing energy management, improving occupant comfort, and effectively controlling building systems, particularly for short- and long-term horizons. Recently, deep learning-based occupancy prediction methods have gained considerable attention. However, the full potential of these methods remains under explored in terms of model architecture variations and prediction horizons. This study introduces cascaded LSTM and cascaded Bi-LSTM models for multi-horizon predictions from 10 minutes to 24 hours, integrating a modified activation function, additional input features, and optimized hyper-parameters using OPTUNA. Traditional performance metrics and various other analyses were conducted to compare the models. Both models performed well for short- and long-term predictions, with minimal differences in the results. Nevertheless, analysis focusing on non-zero data errors (accounting for approximately 11% of occupied periods) and occupancy-wise errors showed a significant performance gap between the two models. The cascaded Bi-LSTM model demonstrated consistent performance across various prediction horizons and occupancy variations, with accuracy approximately 10-15% higher than the cascaded LSTM model, highlighting its superior capability in capturing complex dataset dynamics through a bidirectional process. This study highlights the importance of additional input features, data feature analysis, and multi-perspective result analysis to select the most suitable model for occupancy prediction, validated with pre- and post-modeling feature importance analysis.
... However, they primarily capture the mapping rather than directly illustrating the underlying connections. It is suggested to test the performance of ANN in specific criteria as it might not be as good as traditional model techniques in transferability (Dai, Liu, and Zhang 2020b). The Bernoulli process predicts the probability of interaction under each state. ...
The occupant behavior has a significant impact on the building performance. This review offers researchers and practitioners a clear understanding of the evolution of the occupant behavior model and its involvement in building optimization design, providing guidance for future research. Studies are selected if they are peer-reviewed conference proceedings and journal papers published from 2005 to 2023. By analyzing 203 core papers from the Web of Science, a bibliometric study is conducted using CiteSpace. The findings indicate that research on occupant behavior has advanced significantly since 2005, with China being the most active country. The interaction between occupant and building performance, development of occupant behavior model, building optimization framework, and optimized strategies in buildings are the four fundamental topics. Finally, the current challenges and future opportunities are discussed. Future research should focus on conducting real-case quantitative studies, expanding performance simulation, developing comprehensive techniques, improving user-system interaction, and exploring feasible social practices. However, limitations arise due to deficiencies in the thoroughness and depth of analyses resulting from sample constraints. To obtain more reliable results, a broader and more systematic approach to collecting publications is needed.
... Artificial neural networks are especially effective for short-term predictions, while regression models are preferred for long-term forecasting. [17] Previous research has demonstrated that decision trees can achieve an accuracy of over 94% in predicting room occupancy using only CO2 data as data basis. [7]. ...
Detection Prediction CO2 Building automation A B S T R A C T The paper aims to improve heating energy efficiency in academic rooms by developing a method to identify and predict room occupancy with limited measurement data. To enable the reduction of room temperatures when unoccupied, a scalable method is required that can operate with limited measurement data while delivering reliable results. This study employs a weakly supervised learning approach for this purpose. The initial step involves labelling the measurement data using a rule-based classifier, which may result in classification errors in the dataset. Subsequently, decision trees are trained and generalised to reclassify the measurement data. The method leads to an improvement so that 84 % instead of 67 % of the occupancies are correctly recognised. Using the Minimum Description Length method, another decision tree is trained to predict room occupancy based on the improved data set. The resulting model achieves an accuracy of 99 % in predicting room occupancy given the time of day and day of the week. Compared to other methods of room occupancy detection, this approach requires minimal measurement points and no manual data preparation, making it cost-effective and scalable. As a result, it can significantly enhance the energy efficiency of buildings and contribute to achieving climate targets.
... Machine learning algorithms, both supervised and unsupervised, have been successfully employed to estimate occupancy and obtain occupancy patterns [8]. Supervised learning algorithms are used to learn either occupancy count or presence, while unsupervised learning algorithms are used to find occupancy patterns [9]. ...
... Supervised learning algorithms are used to learn either occupancy count or presence, while unsupervised learning algorithms are used to find occupancy patterns [9]. Some common supervised algorithms for occupancy estimation include artificial neural networks (ANN), k-nearest neighbors (kNN), and decision trees (DT) [8]. Clustering or data mining algorithms are common unsupervised learning algorithms used for occupancy pattern learning [4]. ...
... Intelligent glazing systems allow you to regulate how much light and heat from the sun reaches the structure. The structures that regulate Smart windows come with continuous shade control that controls light levels through daylight hours, in addition to active as well as passive window tinting that reacts to variations in temperature or sunshine [24]. With more research focusing on its application in adaptable substrates for higher efficiencies at lower costs and more durable devices, a method used in [25] shows an acceptable rate of 22.3% produced in goods made of cheap building materials. ...
In order to enhance a building's functionality and energy efficiency, smart construction technology combines controlled reliable controllers and programs with connected sensors, smart energy products, and data analytics software to track ambient data and resident energy usage patterns. The integration of intelligent technologies and controllers in the construction sector is steadily expanding as a result of their growing significance in the fabrication, business-related; it and scholarly domains. This investigation uses a systematic strategy to review the related published literature between 2013 and 2023.Prior to investigating the rating of publications based on quality evaluation and yearly trend publication was done. The study further classifies the literature based on integration technologies and application. The finding reveals that most of the literature features with occupants an interface to schedule, monitor, and adjust energy consumption profiles. These developments also enable utilities to engage in an exchange with the grid by means of demand response strategies and automatic feedback interruption features.Potential study directions are also examined in the paper, particularly with regard to seamless integration, privacy, and security enhancements. It is also proposed that continuous surveillance of data and machine learning are two ways to enhance the intelligence of smart building technology.
... Over the last decade, the number of ML applications dedicated to energy management for building environment has increased significantly, as evidenced by the high number of scientific studies and reviews in the literature [11,[24][25][26][27][28][29][30][31][32][33][34][35][36][37][38]. The main topics investigated are prediction of load and power, fault detection and diagnosis (FDD), and occupancy-related applications. ...
... For example, in energy-related applications, it is possible to use occupancy data to regulate thermostats [86,87], control HVAC systems and air-exchange rate [62,[88][89][90][91][92], control lighting systems [86,87,93], and generally control electrically powered systems [94]. To give a reference value, energy consumption could be reduced by up to 23% by using occupancy predictions to control HVAC systems and by up to 20% beyond that value by using activity recognition [34,95]. ...
... The vertical arrow represents the rise in accuracy associated with more complex "black box" models, while the horizontal arrow represents the intrinsic interpretability of simpler ones. consumption could be reduced by up to 23% by using occupancy predictions to control HVAC systems and by up to 20% beyond that value by using activity recognition [34,95]. ...
Machine learning (ML) algorithms are now part of everyday life, as many technological devices use these algorithms. The spectrum of uses is wide, but it is evident that ML represents a revolution that may change almost every human activity. However, as for all innovations, it comes with challenges. One of the most critical of these challenges is providing users with an understanding of how models’ output is related to input data. This is called “interpretability”, and it is focused on explaining what feature influences a model’s output. Some algorithms have a simple and easy-to-understand relationship between input and output, while other models are “black boxes” that return an output without giving the user information as to what influenced it. The lack of this knowledge creates a truthfulness issue when the output is inspected by a human, especially when the operator is not a data scientist. The Building and Construction sector is starting to face this innovation, and its scientific community is working to define best practices and models. This work is intended for developing a deep analysis to determine how interpretable ML models could be among the most promising future technologies for the energy management in built environments.
... The quest for techniques that can effectively conserve energy while upholding clean and pleasurable indoor environments is of paramount importance and has garnered significant attention from both society and academia [1,2,5]. The continuous evolution of computing and machine learning technologies has significantly contributed to optimizing human comfort and health, as well as reducing energy consumption [6,7]. ...
This study focuses on gathering environmental data concerning the indoor climate within a dormitory, encompassing variables such as air temperature, relative humidity, CO2 concentration, fine dust concentration, illuminance, and total volatile organic compounds. Subsequently, an anomaly detection long short-term memory model (LSTM) model, utilizing a two-stacked LSTM model, was developed and trained to enhance indoor environment control. The study demonstrated that the trained model effectively identified anomalies within eight environmental variables. Graphical representations illustrate the model’s accuracy in anomaly detection. The trained model has the capacity to monitor indoor environmental data collected and transmitted using an Internet-of-Things sensor. In the event of an anomaly domain prediction, it proactively alerts the building manager, facilitating timely indoor environment control. Furthermore, the model can be seamlessly integrated into indoor environment control systems to actively detect anomalies, thereby contributing to the automation of control processes.
... These requirements culminate in the development of smart equipment and solutions that, when implemented within Smart Buildings, foster a cycle of continuous improvement. Quanto ao uso da tecnologia, verifica-se o desenvolvimento de tecnologias abrangendo inteligência artificial e outras técnicas para previsão da ocupação, tais como refletidos nos trabalhos de [7], [8] e [9]. Entretanto, [1] indicou que implementações em campo de smart buildings residenciais têm sido pouco estudadas e aplicadas, apesar de este setor ser responsável por 30% do consumo mundial de energia, conforme [10]. ...
O conceito de Edifícios Inteligentes (Smart Buildings) centra-se em atender às necessidades do usuário e otimizar, através da tecnologia, o equilíbrio entre essas e o desempenho da edificação. Embora os sistemas de controle inteligentes tenham potencial significativo de impacto no desempenho dos edifícios, há escassez de estudos e aplicações em edifícios residenciais. Este estudo objetivou realizar revisão sistemática da literatura para mapear parâmetros relacionados ao desempenho de Smart Buildings residenciais e sua interconexão com necessidades do usuário, propondo fluxo de trabalho sistemático focado no desenvolvimento nesse domínio. A metodologia envolveu processo de categorização documental, utilizando critérios predefinidos para seleção dos estudos e construção de fluxograma estruturado. Concluiu-se que o conceito de desempenho se dá principalmente sob a perspectiva do consumo energético, prioritariamente nos sistemas de aquecimento, ventilação e ar-condicionado (AVAc) restritos, em sua maioria, à previsão da ocupação baseada em simulações e bases de dados variadas e segmentadas. O fluxograma desenvolvido abrange quatro etapas e propõe início a partir de categorização crítica sobre os aspectos que são influenciados pelo usuário para definir requisitos de desempenho baseados em dados estruturados. Estes requisitos, proporcionam desenvolvimento de equipamentos e soluções inteligentes, que aplicadas em edificações inteligentes, incentivam ciclo de melhoria contínua.
... A detailed review conducted by [16] highlighted the dire need for adaptation to occupancy variation; the studies reviewed showed the primary sources of inefficiency in the building system, namely, irregular and partial occupancy. Prior research has demonstrated significant reductions in building energy usage by aligning HVAC systems with actual occupancy patterns [17,18]. For example, the simulation results presented in [19], showed that there was the potential to achieve energy savings ranging from 11% to 34% across different climatic regions while maintaining occupant comfort levels. ...
Buildings are responsible for approximately 40% of the world’s energy consumption and 36% of the total carbon dioxide emissions. Building occupancy is essential, enabling occupant-centric control for zero emissions and decarbonization. Although existing machine learning and deep learning methods for building occupancy prediction have made notable progress, their analyses remain limited when applied to complex real-world scenarios. Moreover, there is a high expectation for Transformer algorithms to predict building occupancy accurately. Therefore, this paper presents an occupancy prediction Transformer network (OPTnet). We fused and fed multi-sensor data (building occupancy, indoor environmental conditions, HVAC operations) into a Transformer model to forecast the future occupancy presence in multiple zones. We performed experimental analyses and compared it to different occupancy prediction methods (e.g., decision tree, long short-term memory networks, multi-layer perceptron) and diverse time horizons (1, 2, 3, 5, 10, 20, 30 min). Performance metrics (e.g., accuracy and mean squared error) were employed to evaluate the effectiveness of the prediction algorithms. Our OPTnet method achieved superior performance on our experimental two-week data compared to existing methods. The improved performance indicates its potential to enhance HVAC control systems and energy optimization strategies.
... In an adaptive behavior-based control strategy, environmental data and the occupants' behavioral patterns are considered at the same time [137]. In most studies, the adaptive behavior of the occupant opening or closing the window was detected via FIR sensors, while environmental parameters, such as temperature, humidity, and CO 2 concentration, were collected via environmental parameters. ...
