Open Access
Open access
Energies, volume 16, issue 5, pages 2283

Electrical Load Forecasting Using LSTM, GRU, and RNN Algorithms

Mobarak Abumohsen 1
Amani Yousef Owda 1
Majdi Owda 2
1
 
Department of Natural, Engineering and Technology Sciences, Arab American University, Ramallah P600, Palestine
2
 
Faculty of Data Science, Arab American University, Ramallah P600, Palestine
Publication typeJournal Article
Publication date2023-02-27
Journal: Energies
scimago Q1
wos Q3
SJR0.651
CiteScore6.2
Impact factor3
ISSN19961073
Electrical and Electronic Engineering
Energy Engineering and Power Technology
Renewable Energy, Sustainability and the Environment
Building and Construction
Control and Optimization
Engineering (miscellaneous)
Energy (miscellaneous)
Abstract

Forecasting the electrical load is essential in power system design and growth. It is critical from both a technical and a financial standpoint as it improves the power system performance, reliability, safety, and stability as well as lowers operating costs. The main aim of this paper is to make forecasting models to accurately estimate the electrical load based on the measurements of current electrical loads of the electricity company. The importance of having forecasting models is in predicting the future electrical loads, which will lead to reducing costs and resources, as well as better electric load distribution for electric companies. In this paper, deep learning algorithms are used to forecast the electrical loads; namely: (1) Long Short-Term Memory (LSTM), (2) Gated Recurrent Units (GRU), and (3) Recurrent Neural Networks (RNN). The models were tested, and the GRU model achieved the best performance in terms of accuracy and the lowest error. Results show that the GRU model achieved an R-squared of 90.228%, Mean Square Error (MSE) of 0.00215, and Mean Absolute Error (MAE) of 0.03266.

Xia Y., Wang J., Wei D., Zhang Z.
2023-03-01 citations by CoLab: 14 Abstract  
In recent years, accurate electricity load forecasting has become increasingly essential for improving the management efficiency of power-generation systems. However, previously proposed hybrid models directly apply signal-processing technology in data preprocessing, resulting in poor efficiency in matching the electricity sequence characteristics. Moreover, most studies exhibit non-optimal performance in practical applications because they focus on forecasting accuracy and ignore forecasting stability. In this study, a combined framework that includes amodified noise processing strategy, multi-objective optimization algorithm, and deep neural network is proposed to solve the low prediction-accuracy problem in electricity load forecasting. The 30-minute real time data of electricity load from Queensland, Australia, are employed to verify the reliability of the proposed framework. The mean absolute percentage error values of the proposed framework in a multi-step prediction approached the values of MAPE1−step=0.79%, MAPE2−step=1.13%, and MAPE3−step=1.50% in Series 1, which significantly outperformed the existing contrast models.
Mansouri S.A., Rezaee Jordehi A., Marzband M., Tostado-Véliz M., Jurado F., Aguado J.A.
Applied Energy scimago Q1 wos Q1
2023-03-01 citations by CoLab: 120 Abstract  
The integrated exploitation of different energy infrastructures in the form of multi-energy systems (MESs) and the transformation of traditional prosumers into smart prosumers are two effective pathways to achieve net-zero emission energy systems in the near future. Managing different energy markets is one of the biggest challenges for the operators of MESs, since different carriers are traded in them simultaneously. Hence, this paper presents a hierarchical decentralized framework for the simultaneous management of electricity, heat and hydrogen markets among multi-energy microgrids (MEMGs) integrated with smart prosumers. The market strategy of MEMGs is deployed using a hierarchical framework and considering the programs requested by smart prosumers. A deep learning-based forecaster is utilized to predict uncertain parameters while a risk-averse information gap decision theory (IGDT)-based strategy controls the scheduling risk. A new prediction-based mechanism for designing dynamic demand response (DR) schemes compatible with smart prosumers’ behavior is introduced, and the results illustrate that this mechanism reduces the electricity and heat clearing prices in peak hours by 17.5% and 8.78%, respectively. Moreover, the results reveal that the introduced structure for hydrogen exchange through the transportation system has the ability to be implemented in competitive markets. Overall, the simulation results confirm that the proposed hierarchical model is able to optimally manage the competitive markets of electricity, heat and hydrogen by taking advantage of the potential of smart prosumers.
Huang N., Wang S., Wang R., Cai G., Liu Y., Dai Q.
2023-02-01 citations by CoLab: 52 Abstract  
• The constructed similar-weighted spatial-temporal graph can directly reflect the coupling relationship between loads of each bus and has certain interpretability. • Mining coupling associations between specific bus loads through SCL. And the spatial aggregation features are generated to realize the enhancement of full-domain node features. • The aggregated features are constructed as temporal series and input to GRUL to mine their temporal domain features. Comparative experiments show that the proposed method greatly reduces the worst load forecast evaluation metrics. Moreover, it has better robustness when containing abnormal load data. Existing short-term bus load forecasting methods mostly use temporal domain features, such as historical loads, to forecast and do not fully consider the influence of unstructured spatial–temporal coupling correlations among multiple bus loads in a wide spatial area on the forecasting results. In this paper, a wide-area multiple bus loads forecasting model is proposed. First, take the rapid computation of the maximal information coefficient (RapidMIC) value between bus loads as edge features, and combine the multiple node feature set to construct a similar-weighted spatial–temporal graph. The constructed similar-weighted spatial–temporal graph is not constrained by geography and grid topology, and can directly reflect the degree of spatial–temporal coupling correlation between bus loads. Second, the neighboring node features of each node in the graph are extracted by the spatial convolution layer (SCL) to achieve full-domain node feature enhancement. Finally, the features extracted were formed into temporal series and input to the gated recurrent unit layer (GRUL) to achieve a wide-area multiple buses short-term load forecast. The research was carried out by selecting the measured load data (15 busbars). The corresponding symmetric mean absolute percentage error (SMAPE) and mean absolute error (MAE) are 3.19 % to 5.89 % and 1.83 megawatts (MW) to 9.8 MW, respectively. Compared with the SMAPE of other methods, the worst evaluation metrics in the test set are improved by 1.82 % to 5.94 %, and it has better robustness in the scene with abnormal load data.
Liu C., Tseng C., Huang T., Yang J., Huang K.
Energy and Buildings scimago Q1 wos Q1
2023-01-01 citations by CoLab: 22 Abstract  
Buildings are one of the largest energy-consuming sectors in the world. Accurate forecasting of building electricity loads can bring significant environmental and economic benefits by reducing electricity use and the corresponding greenhouse gas emissions. Deep learning has witnessed great success in the past decade, so this work proposes a deep learning model to predict the electrical loads of a commercial building. We perform a data exploration on the training set, and the results show that the electricity load is relevant to the temperature. Thus, this work proposes a deep learning model based on multi-task learning (MTL) architecture to predict the hourly electricity load. In our proposed architecture, the main task is to predict the electricity load, while the auxiliary task is to predict the outdoor temperature. The auxiliary task can provide additional regularization to the model to prevent overfitting. Furthermore, the proposed model comprises task-shared and task-specific layers, giving a base to learn cross-task and task-specific representations, respectively. We conduct experiments to assess our proposed model and compare it with other alternatives. The experimental results show that our proposed model can significantly outperform the comparison methods. Furthermore, the analysis shows that our proposed model can benefit from a simple ensemble technique to improve the prediction performance. To validate the generalization of the proposed model, we perform a robustness analysis on three additional datasets. We demonstrate that the proposed MTL approach can provide superior predictive accuracy and robustness.
Ebadzadeh M.M., Nasiri H.
2022-12-30 citations by CoLab: 7 Abstract  
<p>Financial time series prediction, a growing research topic, has attracted considerable interest from scholars, and several approaches have been developed. Among them, decomposition-based methods have achieved promising results. Most decomposition-based methods approximate a single function, which is insufficient for obtaining accurate results. Moreover, most existing researches have concentrated on one-step-ahead forecasting that prevents stock market investors from arriving at the best decisions for the future. This study proposes two novel methods for multi-step-ahead stock price prediction based on the issues outlined. DCT-MFRFNN, a method based on discrete cosine transform (DCT) and multi-functional recurrent fuzzy neural network (MFRFNN), uses DCT to reduce fluctuations in the time series and simplify its structure and MFRFNN to predict the stock price. VMD-MFRFNN, an approach based on variational mode decomposition (VMD) and MFRFNN, brings together their advantages. VMD-MFRFNN consists of two phases. The input signal is decomposed to several IMFs using VMD in the decomposition phase. In the prediction and reconstruction phase, each of the IMFs is given to a separate MFRFNN for prediction, and predicted signals are summed to reconstruct the output. Three financial time series, including Hang Seng Index (HSI), Shanghai Stock Exchange (SSE), and Standard & Poor's 500 Index (SPX), are used for the evaluation of the proposed methods. Experimental results indicate that VMD-MFRFNN surpasses other state-of-the-art methods. VMD-MFRFNN, on average, shows 35.93%, 24.88%, and 34.59% decreases in RMSE from the second-best model for HSI, SSE, and SPX, respectively. Also, DCT-MFRFNN outperforms MFRFNN in all experiments.</p>
Nasiri H., Ebadzadeh M.M.
Neurocomputing scimago Q1 wos Q1
2022-10-01 citations by CoLab: 62 Abstract  
Chaotic time series prediction, a challenging research topic in dynamic system modeling, has drawn great attention from researchers around the world. In recent years extensive researches have been done on developing chaotic time series prediction methods, and various models have been proposed. Among them, recurrent fuzzy neural networks (RFNNs) have shown significant potential in this area. Most of the proposed RFNNs learn a single function, but when dealing with chaotic time series, different outputs may be generated for a specific input based on the system’s state. So, a network is required that can learn multiple functions simultaneously. Based on this concept, a novel multi-functional recurrent fuzzy neural network (MFRFNN) is proposed in this paper. MFRFNN consists of two fuzzy neural networks with Takagi-Sugeno-Kang fuzzy rules, one is used to produce the output, and the other to determine the system’s state. There is a feedback loop between these two networks, which makes MFRFNN capable of learning and memorizing historical information of past observations. Employing the states allows the proposed network to learn multiple functions simultaneously. Moreover, a new learning algorithm, which employs the particle swarm optimization algorithm, is developed to train the networks’ weights. The effectiveness of MFRFNN is validated using the Lorenz and Rossler chaotic time series and four real-world datasets, including Box–Jenkins gas furnace, wind speed prediction, Google stock price prediction, and air quality index prediction. Based on the root mean square error, the proposed method shows a decrease of 35.12 % , 13.95 % , and 49.62 % from the second best methods in the Lorenz time series, Box–Jenkins gas furnace, and wind speed prediction dataset, respectively.
Azarpour A., Mohammadzadeh O., Rezaei N., Zendehboudi S.
2022-10-01 citations by CoLab: 110 Abstract  
• A comprehensive study on renewable energy resources in North America is conducted. • A detailed evaluation of renewable energy potentials is provided. • Important steps for better use of renewable energy resources are summarized. • Emerging technologies for development and management of renewable energy are addressed. • Theoretical and practical challenges in development of renewable energies are highlighted. Energy is an important drive for the economic development and social growth. The energy demand has remarkably escalated due to the technological developments in various governmental, industrial, and municipal activities. The fast growing rate of the energy demand as well as the fuel prices, along with the necessity to control the greenhouse gas emissions are the leading driving forces for effective exploitation of the renewable energy resources. In fact, satisfying the growing global energy demands and mitigating the climate change are of the utmost importance in the modern days. In this review paper, a detailed overview of the development and evolution of the renewable and sustainable energy sources is provided in terms of their types, characteristics, applications, production processes, advantages and disadvantages. Moreover, we discuss the impacts of the renewable energies on health and environment, along with the relevant policies and regulations. Further, the emerging technologies and theoretical and practical challenges in the development of the renewable energies are analyzed. Particularly, this review provides information on renewable and sustainable energies' status and prospects in North America. Central large-scale technology and distributed configurations of the renewable energies might be improved in the North American countries to effectively utilize the renewable energies for producing electricity, biomass-based fuels, and heating in both local/rural and industrial divisions. In addition to employment opportunities stemmed from manufacturing of the renewable energies, its significant impact on the economic development can be enhanced in the US, Canada, and Mexico through development of the renewable energy sources.
Zhou F., Zhou H., Li Z., Zhao K.
Energies scimago Q1 wos Q3 Open Access
2022-07-25 citations by CoLab: 14 PDF Abstract  
The electricity load forecasting plays a pivotal role in the operation of power utility companies precise forecasting and is crucial to mitigate the challenges of supply and demand in the smart grid. More recently, the hybrid models combining signal decomposition and artificial neural networks have received popularity due to their applicability to reduce the difficulty of prediction. However, the commonly used decomposition algorithms and recurrent neural network-based models still confront some dilemmas such as boundary effects, time consumption, etc. Therefore, a hybrid prediction model combining variational mode decomposition (VMD), a temporal convolutional network (TCN), and an error correction strategy is proposed. To address the difficulty in determining the decomposition number and penalty factor for VMD decomposition, the idea of weighted permutation entropy is introduced. The decomposition hyperparameters are optimized by using a comprehensive indicator that takes account of the complexity and amplitude of the subsequences. Besides, a temporal convolutional network is adopted to carry out feature extraction and load prediction for each subsequence, with the primary forecasting results obtained by combining the prediction of each TCN model. In order to further improve the accuracy of prediction for the model, an error correction strategy is applied according to the prediction error of the train set. The Global Energy Competition 2014 dataset is employed to demonstrate the effectiveness and practicality of the proposed hybrid model. The experimental results show that the prediction performance of the proposed hybrid model outperforms the contrast models. The accuracy achieves 0.274%, 0.326%, and 0.405 for 6-steps, 12-steps, and 24 steps ahead forecasting, respectively, in terms of the mean absolute percentage error.
Yohanandhan R.V., Elavarasan R.M., Pugazhendhi R., Premkumar M., Mihet-Popa L., Zhao J., Terzija V.
2022-03-01 citations by CoLab: 43 Abstract  
• History of cyberattacks in energy sector is presented. • Necessity of CPPS testbeds in energy sector are discussed. • A detailed overview on outlook of future CPPS testbeds are provided. • Communication, Control, Computing, and Educational point of view is covered. The Cyber-Physical Power System (CPPS) is a new type of system in which the traditional energy system is integrated into the information network with control systems, communication networks, and computational units. CPPS is the foundation of a long-term transformation of energy management that will fundamentally alter the perspective and application pattern of traditional energy research. The strong interactions between systems in a CPPS introduce new challenges in maintaining high supply security, as new factors can affect the overall security of the power system. Such factors include cybersecurity, the behaviour and constraints of neighbouring energy systems, and the dynamics of interactions between the various systems. Integrating computing, communications, and control at all levels of electrical energy generation, transmission, distribution, utilisation, and storage in CPPS increases vulnerabilities and complicates security. Security studies must reflect the characteristics of the CPPS infrastructure in actual testing environments that support the interfacing of real-world hardware devices. CPPS testbeds are effective in this context because they provide testing capabilities for evaluating the synergistic relationship between physical and virtual components in controlled environments. Security-oriented CPPS testbeds are invaluable for performing cybersecurity and cyberattack analyses, identifying system threats and vulnerabilities in various layers of CPPS, implementing intrusion detection and prevention algorithms, and evaluating the efficacy of mitigation techniques without imposing excessive economic burdens or posing safety risks. This paper provides a comprehensive review of CPPS testbeds from the standpoint of the physical power system layer, cyber system layer, and cyber-physical fusion layer. Following a brief description of the importance of testbeds for cybersecurity research in CPPS, and a brief classification of existing CPPS testbeds for cyberattacks, and cybersecurity analysis in CPPS is presented. A detailed overview of the outlook for future CPPS testbeds in terms of communication, control, computing, and educational perspective is provided. Finally, the work is concluded with future research directions for developing a secure electric power grid.
Yuan B., He B., Yan J., Jiang J., Wei Z., Shen X.
2022-02-01 citations by CoLab: 6 Abstract  
Abstract Accurate electricity consumption forecasting can improve the efficiency of grid dispatching and effectively guarantee the stable operation of the power system. Electricity consumption forecasting is important for the analysis of customer-side electricity consumption behavior, but the instability of electricity consumption sequences poses difficulties for forecasting. Therefore, an improved combination of integrated empirical modal decomposition (EMD) and long short-term memory network (LSTM) is proposed for customer-side electricity consumption forecasting. This paper starts from the idea of blind source separation and independent prediction, and firstly decomposes the original electricity consumption data into several inherent mode functions (IMFs) with different frequencies and amplitudes by empirical modal decomposition (EMD), and then uses LSTM to extract features and make temporal prediction for each IMF component one by one with machine learning intelligent algorithm., and finally obtains end-user-side short-term electricity consumption prediction results by accumulating multiple target prediction results. Compared with direct forecasting, the proposed EMD-LSTM independent forecasting model is able to identify the characteristics of each frequency component of electricity consumption data, and its error is reduced by about 15% on average, thus achieving the goal of improving the accuracy of load forecasting in short-term electricity consumption forecasting scenarios.
Veloso B., Gama J., Malheiro B., Vinagre J.
Information Fusion scimago Q1 wos Q1
2021-12-01 citations by CoLab: 28 Abstract  
The number of Internet of Things devices generating data streams is expected to grow exponentially with the support of emergent technologies such as 5G networks. Therefore, the online processing of these data streams requires the design and development of suitable machine learning algorithms, able to learn online, as data is generated. Like their batch-learning counterparts, stream-based learning algorithms require careful hyperparameter settings. However, this problem is exacerbated in online learning settings, especially with the occurrence of concept drifts, which frequently require the reconfiguration of hyperparameters. In this article, we present SSPT , an extension of the Self Parameter Tuning ( SPT ) optimisation algorithm for data streams. We apply the Nelder–Mead algorithm to dynamically-sized samples, converging to optimal settings in a single pass over data while using a relatively small number of hyperparameter configurations. In addition, our proposal automatically readjusts hyperparameters when concept drift occurs. To assess the effectiveness of SSPT , the algorithm is evaluated with three different machine learning problems: recommendation, regression, and classification. Experiments with well-known data sets show that the proposed algorithm can outperform previous hyperparameter tuning efforts by human experts. Results also show that SSPT converges significantly faster and presents at least similar accuracy when compared with the previous double-pass version of the SPT algorithm. • Stream-based hyper-parameter optimisation algorithm. • Single pass data exploration mode. • Extended evaluation with three different machine learning tasks (recommendation, regression and classification). • Systematic literature review on automatic hyper-parameter tuning for AutoML.
Oprea S., Bâra A., Puican F.C., Radu I.C.
Sustainability scimago Q1 wos Q2 Open Access
2021-10-02 citations by CoLab: 46 PDF Abstract  
When analyzing smart metering data, both reading errors and frauds can be identified. The purpose of this analysis is to alert the utility companies to suspicious consumption behavior that could be further investigated with on-site inspections or other methods. The use of Machine Learning (ML) algorithms to analyze consumption readings can lead to the identification of malfunctions, cyberattacks interrupting measurements, or physical tampering with smart meters. Fraud detection is one of the classical anomaly detection examples, as it is not easy to label consumption or transactional data. Furthermore, frauds differ in nature, and learning is not always possible. In this paper, we analyze large datasets of readings provided by smart meters installed in a trial study in Ireland by applying a hybrid approach. More precisely, we propose an unsupervised ML technique to detect anomalous values in the time series, establish a threshold for the percentage of anomalous readings from the total readings, and then label that time series as suspicious or not. Initially, we propose two types of algorithms for anomaly detection for unlabeled data: Spectral Residual-Convolutional Neural Network (SR-CNN) and an anomaly trained model based on martingales for determining variations in time-series data streams. Then, the Two-Class Boosted Decision Tree and Fisher Linear Discriminant analysis are applied on the previously processed dataset. By training the model, we obtain the required capabilities of detecting suspicious consumers proved by an accuracy of 90%, precision score of 0.875, and F1 score of 0.894.
Javed U., Ijaz K., Jawad M., Ansari E.A., Shabbir N., Kütt L., Husev O.
Energies scimago Q1 wos Q3 Open Access
2021-09-03 citations by CoLab: 29 PDF Abstract  
Power system planning in numerous electric utilities merely relies on the conventional statistical methodologies, such as ARIMA for short-term electrical load forecasting, which is incapable of determining the non-linearities induced by the non-linear seasonal data, which affect the electrical load. This research work presents a comprehensive overview of modern linear and non-linear parametric modeling techniques for short-term electrical load forecasting to ensure stable and reliable power system operations by mitigating non-linearities in electrical load data. Based on the findings of exploratory data analysis, the temporal and climatic factors are identified as the potential input features in these modeling techniques. The real-time electrical load and meteorological data of the city of Lahore in Pakistan are considered to analyze the reliability of different state-of-the-art linear and non-linear parametric methodologies. Based on performance indices, such as Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and Mean Absolute Error (MAE), the qualitative and quantitative comparisons have been conferred among these scientific rationales. The experimental results reveal that the ANN–LM with a single hidden layer performs relatively better in terms of performance indices compared to OE, ARX, ARMAX, SVM, ANN–PSO, KNN, ANN–LM with two hidden layers and bootstrap aggregation models.
Gong L., Yu M., Jiang S., Cutsuridis V., Pearson S.
Sensors scimago Q1 wos Q2 Open Access
2021-07-01 citations by CoLab: 69 PDF Abstract  
Currently, greenhouses are widely applied for plant growth, and environmental parameters can also be controlled in the modern greenhouse to guarantee the maximum crop yield. In order to optimally control greenhouses’ environmental parameters, one indispensable requirement is to accurately predict crop yields based on given environmental parameter settings. In addition, crop yield forecasting in greenhouses plays an important role in greenhouse farming planning and management, which allows cultivators and farmers to utilize the yield prediction results to make knowledgeable management and financial decisions. It is thus important to accurately predict the crop yield in a greenhouse considering the benefits that can be brought by accurate greenhouse crop yield prediction. In this work, we have developed a new greenhouse crop yield prediction technique, by combining two state-of-the-arts networks for temporal sequence processing—temporal convolutional network (TCN) and recurrent neural network (RNN). Comprehensive evaluations of the proposed algorithm have been made on multiple datasets obtained from multiple real greenhouse sites for tomato growing. Based on a statistical analysis of the root mean square errors (RMSEs) between the predicted and actual crop yields, it is shown that the proposed approach achieves more accurate yield prediction performance than both traditional machine learning methods and other classical deep neural networks. Moreover, the experimental study also shows that the historical yield information is the most important factor for accurately predicting future crop yields.
Lei L., Chen W., Wu B., Chen C., Liu W.
Energy and Buildings scimago Q1 wos Q1
2021-06-01 citations by CoLab: 117 Abstract  
• The combination of rough set theory and deep learning algorithms is analyzed. • Rough set theory is used to reduce the influencing factors of building energy consumption. • Deep learning is used to extract features of building energy consumption data. • An accuracy comparison of several prediction models based on rough sets and different neural networks is made. The efficient and accurate prediction of building energy consumption can improve the management of power systems. In this paper, the rough set theory was used to reduce the redundant influencing factors of building energy consumption and find the critical factors of building energy consumption. These key factors were then used as the input of a deep neural network with a “deep” architecture and powerful capabilities in extracting features. Building energy consumption is output of the deep neural network. This study collected data from 100 civil public buildings for rough set reduction, and then collected data from a laboratory building of a university in Dalian for nearly a year to train and test deep neural networks. The test included both the short-term and medium-term predictions of building energy consumption. The prediction results of the deep neural network were compared with that of the back propagation neural network, Elman neural network and fuzzy neural network. The results show that the integrated rough set and deep neural network was the most accurate. The method proposed in this study could provide a practical and accurate solution for building energy consumption prediction.
Hou Y., Ma C., Li X., Sun Y., Yu H., Fang Z.
Energies scimago Q1 wos Q3 Open Access
2025-01-31 citations by CoLab: 0 PDF Abstract  
Simple load forecasting and overload prediction models, such as LSTM and XGBoost, are unable to handle the increasing amount of data in power systems. Recently, various foundation models (FMs) for time series analysis have been proposed, which can be scaled up for large time series variables and datasets across domains. However, the simple pre-training setting makes FMs unsuitable for complex downstream tasks. Effectively handling real-world tasks depends on additional data, i.e., covariates, and prior knowledge. Incorporating these through structural modifications to FMs is not feasible, as it would disrupt the pre-trained weights. To address this issue, this paper proposes a frequency domain mixer, i.e., FreqMixer, framework for enhancing the task-specific analytical capabilities of FMs. FreqMixer is an auxiliary network for the backbone FMs that takes covariates as input. It has the same number of layers as the backbone and communicates with it at each layer, allowing the incorporation of prior knowledge without altering the backbone’s structure. Through experiments, FreqMixer demonstrates high efficiency and performance, reducing MAPE by 23.65%, recall by 87%, and precision by 72% in transformer load forecasting during the Spring Festival while improving precision by 192.09% and accuracy by 14% in corresponding overload prediction, all while processing data from over 160 transformers with just 1M additional parameters.
Szostak B., Doroz R., Marker M.
Applied Sciences (Switzerland) scimago Q2 wos Q2 Open Access
2025-01-22 citations by CoLab: 0 PDF Abstract  
Accurate weather prediction and electrical load modeling are critical for optimizing energy systems and mitigating environmental impacts. This study explores the integration of the novel Mean Background Method and Background Estimation Method with Explainable Artificial Intelligence (XAI) with the aim to enhance the evaluation and understanding of time-series models in these domains. The electrical load or temperature predictions are regression-based problems. Some XAI methods, such as SHAP, require using the base value of the model as the background to provide an explanation. However, in contextualized situations, the default base value is not always the best choice. The selection of the background can significantly affect the corresponding Shapley values. This paper presents two innovative XAI methods designed to provide robust context-aware explanations for regression and time-series problems, addressing critical gaps in model interpretability. They can be used to improve background selection to make more conscious decisions and improve the understanding of predictions made by models that use time-series data.
Jing J., Di H., Wang T., Jiang N., Xiang Z.
Energy Informatics scimago Q2 Open Access
2025-01-08 citations by CoLab: 1 PDF Abstract  
AbstractThis study seeks to enhance the accuracy and economic efficiency of power system load forecasting (PSLF) by leveraging Artificial Neural Networks. A predictive model based on a Residual Connection Bidirectional Long Short Term Memory Attention mechanism (RBiLSTM-AM) is proposed. In this model, normalized power load time series data is used as input, with the Bidirectional Long and Short Term Memory network capturing the bidirectional dependencies of the time series and the residual connections preventing gradient vanishing. Subsequently, an attention mechanism is applied to capture the influence of significant time steps, thereby improving prediction accuracy. Based on the load forecasting, a Particle Swarm Optimization (PSO) algorithm is employed to quickly determine the optimal scheduling strategy, ensuring the economic efficiency and safety of the power system. Results show that the proposed RBiLSTM-AM achieves an accuracy of 96.68%, precision of 91.56%, recall of 90.51%, and an F1-score of 91.37%, significantly outperforming other models (e.g., the Recurrent Neural Network model, which has an accuracy of 69.94%). In terms of error metrics, the RBiLSTM-AM model reduces the root mean square error to 123.70 kW, mean absolute error to 104.44 kW, and mean absolute percentage error (MAPE) to 5.62%, all of which are lower than those of other models. Economic cost analysis further demonstrates that the PSO scheduling strategy achieves significantly lower costs at most time points compared to the Genetic Algorithm (GA) and Simulated Annealing (SA) strategies, with the cost being 689.17 USD in the first hour and 2214.03 USD in the fourth hour, both lower than those of GA and SA. Therefore, the proposed RBiLSTM-AM model and PSO scheduling strategy demonstrate significant accuracy and economic benefits in PSLF, providing effective technical support for optimizing power system scheduling.
Hussein E.E., Zerouali B., Bailek N., Derdour A., Ghoneim S.S., Santos C.A., Hashim M.A.
Water (Switzerland) scimago Q1 wos Q2 Open Access
2024-12-29 citations by CoLab: 0 PDF Abstract  
Irrigation water quality is crucial for sustainable agriculture and environmental health, influencing crop productivity and ecosystem balance globally. This study evaluates the performance of multiple deep learning models in classifying the Irrigation Water Quality Index (IWQI), addressing the challenge of accurate water quality prediction by examining the impact of increasing input complexity, particularly through chemical ions and derived quality indices. The models tested include convolutional neural networks (CNN), CNN-Long Short-Term Memory networks (CNN-LSTM), CNN-bidirectional Long Short-Term Memory networks (CNN-BiLSTM), and CNN-bidirectional Gated Recurrent Unit networks (CNN-BiGRUs). Feature selection via SHapley Additive exPlanations (SHAP) provided insights into individual feature contributions to the model predictions. The objectives were to compare the performance of 16 models and identify the most effective approach for accurate IWQI classification. This study utilized data from 166 wells in Algeria’s Naama region, with 70% of the data for training and 30% for testing. Results indicate that the CNN-BiLSTM model outperformed others, achieving an accuracy of 0.94 and an area under the curve (AUC) of 0.994. While CNN models effectively capture spatial features, they struggle with temporal dependencies—a limitation addressed by LSTM and BiGRU layers, which were further enhanced through bidirectional processing in the CNN-BiLSTM model. Feature importance analysis revealed that the quality index (qi) qi-Na was the most significant predictor in both Model 15 (0.68) and Model 16 (0.67). The quality index qi-EC showed a slight decrease in importance, from 0.19 to 0.18 between the models, while qi-SAR and qi-Cl maintained similar importance levels. Notably, Model 16 included qi-HCO3 with a minor importance score of 0.02. Overall, these findings underscore the critical role of sodium levels in water quality predictions and suggest areas for enhancing model performance. Despite the computational demands of the CNN-BiLSTM model, the results contribute to the development of robust models for effective water quality management, thereby promoting agricultural sustainability.
Chang Q., Yuan T., Li H., Chen Y., Wang X., Gao S., Ren H., Zhao X., Wang L.
Actuators scimago Q2 wos Q2 Open Access
2024-12-11 citations by CoLab: 0 PDF Abstract  
The growing diversity and number of industrial robots make energy consumption prediction and optimization increasingly essential. Current data-driven approaches, particularly those based on multi-layer perception (MLP), have shown feasibility but typically overlook the variability or unknown nature of load-related parameters in real-world applications. This paper presents a KAN-LSTM model designed to accurately predict energy consumption under unknown load conditions, alongside a particle swarm optimization (PSO) algorithm for minimizing energy use. First, an industrial robot dynamics and energy consumption model is established. Then, the KAN-LSTM model is trained on datasets from the AUBO-E5 robot, with its predictions compared to alternative network models. Finally, PSO is applied to optimize energy consumption. Experimental results indicate that the KAN-LSTM model achieves high prediction accuracy (95.7–97.1%) and offers substantial energy optimization potential (53.1–64.7%). Optimized industrial robots are particularly suitable for tasks such as picking and palletizing in the courier industry, saving operational costs and increasing the sustainability of automated systems in logistics environments.
Haidar D., Mouatassim S., Benabbou R., Benhra J.
2024-12-06 citations by CoLab: 1 Abstract  
There's a pressing need to democratise DL algorithms while leveraging their performance. This chapter proposes a customisable and efficient Automated Machine Learning (AutoML) forecasting framework to deal with volatile and complex time series using Hyperparameter Optimization (HPO) techniques in combination with ANN, LSTM, GRU, Bi-LSTM and Bi-GRU. The forecasting framework uses hyperband and random search in a high-dimensional hyperparameter space to demonstrate the models' performance without requiring sophisticated pre-processing steps, thereby providing a milestone to design DL models after a comparative analysis of specific recurrent models. After finding optimal hyperparameter combinations for each model, we study the correlation and the variance between the performance and specific hyperparameter combinations using statistical tests, data visualisation tools, and SHAP. The results discussed improvements of the forecasting framework after elaborating on the relationship between the models' performance, the dataset's size, its inherent noise and the hyperparameter selection.
Li J., Zhang X., Hu Q., Zhang F., Gaida O., Chen L.
Sustainability scimago Q1 wos Q2 Open Access
2024-12-06 citations by CoLab: 0 PDF Abstract  
Factory aquaculture faces a difficult situation due to its high running costs, with one of the main contributing factors being the high energy consumption of aquaculture workshops. Accurately predicting the power load of recirculating aquaculture systems (RAS) is critical to optimizing energy use, reducing energy consumption, and promoting the sustainable development of factory aquaculture. Adequate data can improve the accuracy of the prediction model. However, there are often missing and abnormal data in actual data detection. To solve this problem, this study uses a time-series convolutional network–temporal sequence generation adversarial network (TCN-TimeGAN) to synthesize multivariate RAS data and train a long short-term memory (LSTM) network on the original and generated data to predict future electricity loads. The experimental results show that the data generated based on the improved TCN-TimeGAN provide more comprehensive coverage of the original data distribution, with a lower discriminative score (0.2419) and a lower predictive score (0.0668) than the conventional TimeGAN. Using the generated data for prediction, the R2 reached 0.86, which represents a 19% improvement over the ARIMA model. Meanwhile, compared to LSTM and GRU without data augmentation, the mean absolute error (MAE) was reduced by 1.24 and 1.58, respectively. The model has good prediction performance and generalization ability, which benefits the RAS energy saving, production planning, and the long term sustainability of factory aquaculture.

Top-30

Journals

1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9

Publishers

5
10
15
20
25
30
35
40
45
5
10
15
20
25
30
35
40
45
  • We do not take into account publications without a DOI.
  • Statistics recalculated only for publications connected to researchers, organizations and labs registered on the platform.
  • Statistics recalculated weekly.

Are you a researcher?

Create a profile to get free access to personal recommendations for colleagues and new articles.
Share
Cite this
GOST | RIS | BibTex | MLA
Found error?