Article
Computer Science, Information Systems
Shu-Rong Yan, Manwen Tian, Khalid A. A. Alattas, Ardashir Mohamadzadeh, Mohammad Hosein Sabzalian, Amir H. H. Mosavi
Summary: A neural network-based approach is designed for mid-term load forecasting, with the structure and hyperparameters tuned for the best accuracy one year ahead. The approach is practically applied in a region in Iran using real-world data sets of 10 years. The study investigates influential factors such as economic, weather, and social factors, and their impact on accuracy is numerically analyzed. The suggested approach also detects bad data and predicts the 24-hour load pattern, aiding in mid-term planning.
Article
Energy & Fuels
Joaquin Delgado Fernandez, Sergio Potenciano, Chul Min Lee, Alexander Rieger, Gilbert Fridgen
Summary: This paper investigates the challenges of using smart meter data for load forecasting and proposes a solution using a combination of federated learning and privacy preserving techniques. The results show that this combination can achieve high forecasting accuracy and near-complete privacy protection.
Article
Computer Science, Artificial Intelligence
Ashraful Haque, Saifur Rahman
Summary: This paper presents a method for short-term commercial building electrical load forecasting using a regularized deep neural network. Through detailed analysis and validation, the method demonstrates superior performance in load forecasting.
APPLIED SOFT COMPUTING
(2022)
Article
Thermodynamics
Niaz Bashiri Behmiri, Carlo Fezzi, Francesco Ravazzolo
Summary: One of the most controversial issues in mid-term load forecasting is how to treat weather. This article compares three approaches: excluding weather, assuming perfect knowledge of future weather, and including weather forecasts in load forecasting models. The results show that models including future temperature consistently outperform models excluding temperature, but predictions of future temperature weaken the results.
Article
Environmental Studies
Ali Can Ozdemir, Kurtulus Bulus, Kasim Zor
Summary: This study proposes the use of recurrent neural networks, specifically long short-term memory (LSTM) and gated recurrent unit (GRU) networks, based on deep learning algorithms to forecast nickel price variations. The results show that both networks are highly effective and successful, with low prediction error rates. Furthermore, the GRU networks outperform the LSTM networks in terms of computational time.
Article
Engineering, Electrical & Electronic
Yanzhu Liu, Shreya Dutta, Adams Wai Kin Kong, Chai Kiat Yeo
Summary: In current power systems, load forecasting is crucial for system planning and operations. This paper focuses on short-term load forecasting. The challenge arises from the complex multi-level seasonality of load series and the representation of load data in 1d numerical series. To address this, the study proposes a transformation of load data into 3d images and a recurrent neural network to model temporal trends. Experimental results demonstrate the effectiveness of the proposed method.
IEEE TRANSACTIONS ON POWER SYSTEMS
(2023)
Article
Thermodynamics
Ibrahim Anwar Ibrahim, M. J. Hossain
Summary: This study proposes an optimized deep-ensemble learning methodology for accurate and short-term load forecasting, which can help distributed network service providers plan and operate electrical networks and provide better quality and more reliable electricity services. The results showed that the Bi-LSTM-AWDO model outperforms other models in forecasting short-term load data for individual and aggregate households using actual smart meters' measurements.
ENERGY CONVERSION AND MANAGEMENT
(2023)
Article
Energy & Fuels
Alessandro Brusaferri, Matteo Matteucci, Stefano Spinelli, Andrea Vitali
Summary: This work presents a novel approach to improve the trustworthiness of neural network based load forecasting systems by integrating predictive distributions and uncertainty sources. Experimental results demonstrate significant performance improvements in load forecasting.
Article
Thermodynamics
Wenyu Zhang, Qian Chen, Jianyong Yan, Shuai Zhang, Jiyuan Xu
Summary: This study proposes a novel asynchronous deep reinforcement learning model for short-term load forecasting to address the challenges of high temporal correlation and high convergence instability. By introducing new methods to disrupt temporal correlation, adaptively judge training situation, and stabilize model training convergence by considering action trends, the proposed model achieves higher forecasting accuracy, less time cost, and more stable convergence compared to eleven baseline models.
Article
Energy & Fuels
V. Y. Kondaiah, B. Saravanan
Summary: In this paper, a modified deep residual network (deep-ResNet) is proposed to improve the precision of short-term load forecasting by adopting state-of-the-art deep learning techniques and overcome the issues of over-fitting and generalization. The concept of statistical correlational analysis is used to identify appropriate input features extraction ability and generalization capability, leading to improved accuracy of the model. Two utility datasets are used to evaluate the proposed model performance, and the results show promising and accurate prediction compared to existing models in the literature.
FRONTIERS IN ENERGY RESEARCH
(2022)
Article
Automation & Control Systems
Sana Arastehfar, Mohammadjavad Matinkia, Mohammad Reza Jabbarpour
Summary: This study introduces a novel neural network architecture combining Graph Convolutional Networks and Long Short-Term Memory networks for Short-Term Load Forecasting problem. The model captures spatial information from users without prior knowledge of their geographic location and does not rely on additional environmental variables, showing significant improvement compared to baseline models.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Green & Sustainable Science & Technology
Elissaios Sarmas, Evangelos Spiliotis, Efstathios Stamatopoulos, Vangelis Marinakis, Haris Doukas
Summary: This paper proposes a meta-learning method to improve short-term deterministic forecasts of PV systems by blending the base forecasts of multiple DL models. Results indicate that different base models perform best at different PV plants, and meta-learning can improve accuracy by up to 5% over the most accurate base model per plant.
Article
Engineering, Electrical & Electronic
Weixuan Lin, Di Wu, Benoit Boulet
Summary: This paper proposes a graph neural network based framework to address short-term residential load forecasting, which can effectively capture hidden spatial dependencies between different houses. Experimental results demonstrate that the framework can significantly improve the accuracy of residential load forecasting.
IEEE TRANSACTIONS ON SMART GRID
(2021)
Article
Computer Science, Information Systems
K. U. Jaseena, Binsu C. Kovoor
Summary: Weather forecasting is the practice of predicting the state of the atmosphere based on different weather parameters. Accurate weather forecasts are crucial in various fields. With the advancement of atmospheric observing systems and the increasing volume of weather data, deep learning techniques are being used to improve weather prediction. This paper provides a comprehensive review of weather forecasting approaches and discusses potential future research directions.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Ziyu Sheng, Zeyu An, Huiwei Wang, Guo Chen, Kun Tian
Summary: As the energy system becomes more complex and flexible, accurate load forecasting is crucial for power scheduling, load switching, and infrastructure development. This paper proposes a neural network framework that combines modified deep residual network (DRN) and long short-term memory (LSTM) recurrent neural network (RNN) to address the short-term load forecasting (STLF) problem. The proposed model utilizes the strengths of DRN for avoiding vanishing gradient and LSTM for capturing nonlinear patterns, and incorporates dimension weighted units to improve performance in terms of depth, time, and feature dimension. The model is evaluated using public datasets and shows high accuracy, robustness, and generalization capability compared to existing mainstream models.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Grzegorz Dudek, Pawel Pelka
Summary: Pattern similarity-based frameworks are widely used in classification and regression problems, with variations like nearest-neighbor, fuzzy neighborhood, kernel regression, and general regression neural network models. In this study, the proposed models showed high performance in forecasting monthly electricity demand, outperforming classical statistical models and machine learning models. Among the variants, a hybrid approach combining similarity-based methods with exponential smoothing yielded the most accurate results.
APPLIED SOFT COMPUTING
(2021)
Article
Energy & Fuels
Grzegorz Dudek
Summary: Weighting individual errors of training samples in the loss function makes the learning process more sensitive to the neighborhood of the test pattern, thereby improving forecasting accuracy.
Editorial Material
Chemistry, Multidisciplinary
Grzegorz Dudek
APPLIED SCIENCES-BASEL
(2022)
Article
Computer Science, Artificial Intelligence
Grzegorz Dudek, Pawel Pelka, Slawek Smyl
Summary: The study presents a hybrid and hierarchical deep learning model for midterm load forecasting, combining ETS, LSTM, and ensembling. The model demonstrates high performance in load prediction and competes with both classical and state-of-the-art machine learning models.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Energy & Fuels
Grzegorz Dudek
Summary: The study focuses on using random forest (RF) for short-term load forecasting (STLF), with a focus on data representation and training modes. Experimental results show that the optimal RF model performs well on four STLF problems, outperforming statistical and machine learning models in accuracy.
Article
Energy & Fuels
Pawel Pelka
Summary: This article presents a statistical solution (ARIMA, ETS, and Prophet) for predicting monthly power demand, which estimates the relationship between historical and future demand patterns. The time series of energy demand exhibits seasonal fluctuations, long-term trends, instability, and random noise. To simplify the prediction issue, the monthly load time series is represented by an annual cycle pattern, which standardizes the data and filters out the trends. A simulation study conducted on the monthly electricity load time series for 35 European countries confirmed the high accuracy of the proposed models.
Editorial Material
Energy & Fuels
Grzegorz Dudek, Pawel Piotrowski, Dariusz Baczynski
Article
Computer Science, Artificial Intelligence
Grzegorz Dudek
Summary: The decomposition of a time series is crucial for understanding its nature and facilitating analysis and forecasting of various components. However, existing methods neglect the variance of the time series. This work proposes a seasonal-trend-dispersion decomposition method to address heteroscedasticity and extract multiple components effectively.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Editorial Material
Chemistry, Multidisciplinary
Grzegorz Dudek
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Artificial Intelligence
Slawek Smyl, Grzegorz Dudek, Pawel Pelka
Summary: This article proposes a novel hybrid hierarchical deep-learning model that can handle complex time series with multiple seasonality and produce both point forecasts and predictive intervals. The model combines exponential smoothing and a recurrent neural network, which can extract the main components of each time series and effectively model short and long-term dependencies. Empirical study shows that the proposed model has high expressive power and outperforms statistical and state-of-the-art machine learning models in terms of accuracy.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Slawek Smyl, Grzegorz Dudek, Pawel Pelka
Summary: This paper proposes a hybrid forecasting model ES-dRNN with a mechanism for dynamic attention to improve the accuracy of short-term load forecasting. Experimental results show that the proposed model outperforms traditional statistical models and state-of-the-art machine learning forecasting models in terms of prediction accuracy.
2022 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN)
(2022)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Grzegorz Dudek
Summary: In this work, the authors propose an ensemble learning method based on randomized neural networks for forecasting complex time series. By unifying the tasks for all ensemble members, the authors simplify ensemble learning and achieve improved forecasting accuracy. Experimental results confirm the effectiveness of this method in forecasting time series with multiple seasonality.
COMPUTATIONAL SCIENCE - ICCS 2022, PT I
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Grzegorz Dudek
Summary: This study contributes to the development of neural forecasting models with novel randomization-based learning methods, which improve fitting abilities and allow for forecasting time series with multiple seasonality. The proposed models show promising performance in terms of forecasting accuracy, training speed, simplicity, and robustness in dealing with nonstationarity and multiple seasonality in time series.
ADVANCES IN COMPUTATIONAL INTELLIGENCE (IWANN 2021), PT II
(2021)
Article
Energy & Fuels
Shitong Fang, Houfan Du, Tao Yan, Keyu Chen, Zhiyuan Li, Xiaoqing Ma, Zhihui Lai, Shengxi Zhou
Summary: This paper proposes a new type of nonlinear VIV energy harvester (ANVEH) that compensates for the decrease in peak energy output at low wind speeds by introducing an auxiliary structure. Theoretical and experimental results show that ANVEH performs better than traditional nonlinear VIV energy harvesters under various system parameter variations.
Article
Energy & Fuels
Wei Jiang, Shuo Zhang, Teng Wang, Yufei Zhang, Aimin Sha, Jingjing Xiao, Dongdong Yuan
Summary: A standardized method was developed to evaluate the availability of solar energy resources in road areas, which combined the Analytic Hierarchy Process (AHP) and the Geographic Information System (GIS). By analyzing critical factors and using a multi-indicator evaluation method, the method accurately evaluated the utilization of solar energy resources and guided the optimal location selection for road photovoltaic (PV) projects. The results provided guidance for the application of road PV projects and site selection for route corridors worldwide, promoting the integration of transportation and energy.
Article
Energy & Fuels
Chang Liu, Jacob A. Wrubel, Elliot Padgett, Guido Bender
Summary: The study investigates the effects of coating defects on the performance of the anode porous transport layer (PTL) in water electrolyzers. The results show that an increasing fraction of uncoated regions on the PTL leads to decreased cell performance, with continuous uncoated regions having a more severe impact compared to multiple thin uncoated strips.
Article
Energy & Fuels
Marcos Tostado-Veliz, Xiaolong Jin, Rohit Bhakar, Francisco Jurado
Summary: In this paper, a coordinated charging price mechanism for clusters of parking lots is proposed. The research shows that enabling vehicle-to-grid characteristics can bring significant economic benefits for users and the cluster coordinator, and vehicle-to-grid impacts noticeably on the risk-averse character of the uncertainty-aware strategies. The developed pricing mechanism can reduce the cost for users, avoiding to directly translate the energy cost to charging points.
Article
Energy & Fuels
Duan Kang
Summary: Building an energy superpower is a key strategy for China and a long-term goal for other countries. This study proposes an evaluation system and index for measuring energy superpower, and finds that China has significantly improved its ranking over the past 21 years, surpassing other countries.
Article
Energy & Fuels
Fucheng Deng, Yifei Wang, Xiaosen Li, Gang Li, Yi Wang, Bin Huang
Summary: This study investigated the synergistic blockage mechanism of sand and hydrate in gravel filling layer and the evolution of permeability in the layer. Experimental models and modified permeability models were established to analyze the effects of sand particles and hydrate formation on permeability. The study provided valuable insights for the safe and efficient exploitation of hydrate reservoirs.
Article
Energy & Fuels
Hao Wang, Xiwen Chen, Natan Vital, Edward Duffy, Abolfazl Razi
Summary: This study proposes a HVAC energy optimization model based on deep reinforcement learning algorithm. It achieves 37% energy savings and ensures thermal comfort for open office buildings. The model has a low complexity, uses a few controllable factors, and has a short training time with good generalizability.
Article
Energy & Fuels
Moyue Cong, Yongzhuo Gao, Weidong Wang, Long He, Xiwang Mao, Yi Long, Wei Dong
Summary: This study introduces a multi-strategy ultra-wideband energy harvesting device that achieves high power output without the need for external power input. By utilizing asymmetry, stagger array, magnetic coupling, and nonlinearity strategies, the device maintains a stable output voltage and high power density output at non-resonant frequencies. Temperature and humidity monitoring are performed using Bluetooth sensors to adaptively assess the device.
Article
Energy & Fuels
Tianshu Dong, Xiudong Duan, Yuanyuan Huang, Danji Huang, Yingdong Luo, Ziyu Liu, Xiaomeng Ai, Jiakun Fang, Chaolong Song
Summary: Electrochemical water splitting is crucial for hydrogen production, and improving the hydrogen separation rate from the electrode is essential for enhancing water electrolyzer performance. However, issues such as air bubble adhesion to the electrode plate hinder the process. Therefore, a methodology to investigate the two-phase flow within the electrolyzer is in high demand. This study proposes using a microfluidic system as a simulator for the electrolyzer and optimizing the two-phase flow by manipulating the micro-structure of the flow.
Article
Energy & Fuels
Shuo Han, Yifan Yuan, Mengjiao He, Ziwen Zhao, Beibei Xu, Diyi Chen, Jakub Jurasz
Summary: Giving full play to the flexibility of hydropower and integrating more variable renewable energy is of great significance for accelerating the transformation of China's power energy system. This study proposes a novel day-ahead scheduling model that considers the flexibility limited by irregular vibration zones (VZs) and the probability of flexibility shortage in a hydropower-variable renewable energy hybrid generation system. The model is applied to a real hydropower station and effectively improves the flexibility supply capacity of hydropower, especially during heavy load demand in flood season.
Article
Energy & Fuels
Zhen Wang, Kangqi Fan, Shizhong Zhao, Shuxin Wu, Xuan Zhang, Kangjia Zhai, Zhiqi Li, Hua He
Summary: This study developed a high-performance rotary energy harvester (AI-REH) inspired by archery, which efficiently accumulates and releases ultralow-frequency vibration energy. By utilizing a magnetic coupling strategy and an accumulator spring, the AI-REH achieves significantly accelerated rotor speeds and enhanced electric outputs.
Article
Energy & Fuels
Yi Yang, Qianyi Xing, Kang Wang, Caihong Li, Jianzhou Wang, Xiaojia Huang
Summary: In this study, a novel hybrid Quantile Regression (QR) model is proposed for Probabilistic Load Forecasting (PLF). The model integrates causal dilated convolution, residual connection, and Bidirectional Long Short-Term Memory (BiLSTM) for multi-scale feature extraction. In addition, a Combined Probabilistic Load Forecasting System (CPLFS) is proposed to overcome the inherent flaws of relying on a single model. Simulation results show that the hybrid QR outperforms traditional models and CPLFS exceeds the best benchmarks in terms of prediction accuracy and stability.
Article
Energy & Fuels
Wen-Jiang Zou, Young-Bae Kim, Seunghun Jung
Summary: This paper proposes a dynamic prediction model for capacity fade in vanadium redox flow batteries (VRFBs). The model accurately predicts changes in electrolyte volume and capacity fade, enhancing the competitiveness of VRFBs in energy storage applications.
Article
Energy & Fuels
Yuechao Ma, Shengtie Wang, Guangchen Liu, Guizhen Tian, Jianwei Zhang, Ruiming Liu
Summary: This paper focuses on the balance of state of charge (SOC) among multiple battery energy storage units (MBESUs) and bus voltage balance in an islanded bipolar DC microgrid. A SOC automatic balancing strategy is proposed considering the energy flow relationship and utilizing the adaptive virtual resistance algorithm. The simulation results demonstrate the effectiveness of the proposed strategy in achieving SOC balancing and decreasing bus voltage unbalance.
Article
Energy & Fuels
Raad Z. Homod, Basil Sh. Munahi, Hayder Ibrahim Mohammed, Musatafa Abbas Abbood Albadr, Aissa Abderrahmane, Jasim M. Mahdi, Mohamed Bechir Ben Hamida, Bilal Naji Alhasnawi, A. S. Albahri, Hussein Togun, Umar F. Alqsair, Zaher Mundher Yaseen
Summary: In this study, the control problem of the multiple-boiler system (MBS) is formulated as a dynamic Markov decision process and a deep clustering reinforcement learning approach is applied to obtain the optimal control policy. The proposed strategy, based on bang-bang action, shows superior response and achieves more than 32% energy saving compared to conventional fixed parameter controllers under dynamic indoor/outdoor actual conditions.