Energy Informatics最新文献

The need for predictive maintenance methods has arisen as a key element in improving operational efficiency, reliability, and life expectancy of photovoltaic (PV) systems and the future complex renewable energy infrastructure sets. The Machine learning (ML) technique is sub part of Artificial Intelligence (AI) technology which has widened their adoption in energy analytics, resulting in numerous studies proposing different algorithms for monitoring, prediction, and prevention of system failures. The overview of these approaches is yet to be exhaustive in the existing literature regarding a metric-based evaluation. In addressing this gap, the article undertakes a structured review of the state-of-the-art recent peer-reviewed literature on predictive maintenance in solar PV systems. Each work will, therefore, be appraised against standardized performance metrics models, which include aspects such as accuracy, precision, recall, F1-score, area under the curve (AUC), and model-specific indicators- Root Mean Square Error (RMSE), latency, and execution delays. A numerical analysis table summarizes and compares the predictive capabilities of techniques such as Random Forest, CatBoost, Convolutional Neural Network (CNN) ensembles, Long Short-Term Memory (LSTM) autoencoders, Supervisory Control and Data Acquisition (SCADA) IoT frameworks, and Digital Twins. High-performing models, such as CatBoost and custom CNN architectures, indicate the effectiveness of hybrid deep learning strategies in fault diagnostics. The review establishes a new benchmark for evaluating PdM systems, readying the bar between academic innovation and real-world deployment. It outlines future research directions including model generalization, real-time edge AI deployment, and integration with climate-aware forecasting systems. This work complements an important entry point for other works by researchers and industry stakeholders’ intent on deploying scalable and resilient predictive maintenance solutions in renewable energy networks.

The increasing volatility of electricity prices, driven by the growing share of renewable energy, calls for new approaches. This paper proposes a dynamic Bayesian network (DBN) method for electricity price interval forecasting. The model uses predicted values of wind power generation, total power generation, and total electricity consumption, along with historical electricity prices, as inputs. The network structure is determined using a greedy search algorithm, and the model parameters are estimated through maximum likelihood estimation (MLE). By treating the predictions of wind power, total generation, and total consumption as reasoning evidence, the method employs joint tree inference to generate discrete states and posterior probabilities for electricity prices, thereby enabling interval forecasting. The DBN-based interval predictions achieve a prediction interval coverage probability (PICP) of 95.24%, a normalized average width (PINAW) of 9.25%, and an accumulated width deviation (AWD) of 0.56%. The effectiveness of the proposed method was evaluated by comparing its predictions with actual electricity prices and with results from both particle swarm optimization-kernel extreme learning machine (PSO-KELM) and long short-term memory (LSTM)-based methods. This innovative approach not only provides prediction intervals but also associates them with corresponding probabilities, offering significant potential to enhance market participants’ decision-making and mitigate price risks.

As a supporting device for electric vehicles, DC charging piles are widely distributed and in large quantities, involving a huge emerging electricity trading market. Ensuring metering accuracy of charging pile is critical to maintaining fair electricity trading. The traditional on-site verification method for charging pile involves high personnel input and low verification efficiency, making it difficult to meet the massive metering verification demand. In this paper, based on knowledge-assisted modal decomposition, the metering error prediction method for charging pile is proposed to remotely locate the charging pile whose metering error is about to exceed the threshold in advance. First, the trend and multi-period characteristics of metering error data—driven by factors such as temperature, humidity, electrical stresses, and user behavior—are analyzed. With an adaptive data imputation method, high-ratio continuous missing values in metering data time series are completed. Then, the error data time series is decomposed into trend, multi-level periodic, and residual terms with the improved seasonal-trend decomposition method. Finally, the trend and multiple periodic terms are predicted based on the support vector regression model, and they are combined to form the error prediction. The effectiveness and superiority of the proposed method are validated through practical application.

In electricity markets, evaluating collusion and market power is a critical challenge for network operators, as such behaviors can disrupt fair competition, induce price volatility, and reduce market efficiency. Effective methods are therefore required to identify and quantify the influence of each market participant on the profits of others. This study aims to assess collusion and market power in large-scale power systems through structural analysis, addressing gaps left by previous research. The proposed methodology relies on two lemmas to model market behavior. Lemma 1 quantifies the effects of various factors on local price changes and generation capacities, while Lemma 2 evaluates their impact on the profit variations of generation units. Using the matrix derived from Lemma 2, which captures profit responses to marginal unit price changes, collusion and market power across the network are assessed. Additionally, three new indicators are introduced to measure market power and collusion in large networks. The approach is applied to a 300-bus system, and detailed analysis demonstrates that changes in generation pricing strategies can substantially influence market power and collusive behavior, providing regulators with a tool for proactive market monitoring and intervention.

The proliferation of energy demand with population growth and associated costs necessitated the development of advanced demand response (DR) strategies in smart grid (SG) environments. This study proposes a novel IoT-enabled Energy Management Controller (IEMC) for smart buildings that addresses the critical challenge of optimal appliance scheduling. The proposed system integrates renewable energy sources (photovoltaic systems), energy storage systems (ESS), and advanced metering infrastructure (AMI) to enable autonomous energy management under Time-of-Use (ToU) pricing schemes. The study categorizes household appliances into schedulable and non-schedulable classes, implementing a hybrid metaheuristic optimization algorithm (HGPO) that combines Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Wind Driven Optimization (WDO) techniques. The multi-objective optimization framework simultaneously addresses four critical performance metrics: electricity cost minimization, peak-to-average ratio (PAR) reduction, carbon emission mitigation, and user comfort (UC) maximization. Extensive simulations demonstrate the superior performance of the proposed IEMC system. The hybrid HGPO algorithm achieves a 57.8% improvement in fitness cost (19.34) compared to traditional GA approaches (39.66), while maintaining the lowest emissions (3.41 tonnes/h) and optimal PAR (10). The system successfully shifts schedulable appliances from peak to off-peak hours, resulting in a 79% reduction in grid import dependency and enhanced battery state-of-charge management with peak utilization reaching 8%. Furthermore, comparative analysis with five other metaheuristic algorithms (GA, Binary PSO, WDO, Ant Colony Optimization, and Bacterial Foraging Algorithm) validates the superiority of the hybrid approach across all performance metrics.

Sustainable Development Goals (SDGS) of the United Nations must be linked with technological advancement and sustainable economic practices. The study provides empirical evidence on how Industry 4.0 (I4.0) technologies and Circular Economy (CE) practices will facilitate faster achievement of the SDGs in G20 economies from 2000 to 2024. The research supports the synergistic effect of I4.0 and CE on sustainability by using panel data analysis by industry and region. The results indicate that integrating I4.0 technologies, including internet usage by individuals, importing ICT goods, exporting high technology products, and manufacturing advanced products, alongside CE principles, leads to high resource utilisation, reduced environmental impacts, and increased innovation. The empirical data demonstrate that Industry 4.0 technology-driven transformations and circular economy practices can enhance sustainability performance, as measured by Adjusted Net Savings, in G20 economies. Digital variables such as internet utilisation, ICT imports, and resource-related sustainable decisions like renewable energy adoption are positively associated with sustainability outcomes- highlighting the synergy between digitalisation and environmental sustainability towards SDGS 7, 9, 12, and 13. To ensure effective integration of technology and CE, policy recommendations include developing more robust digital infrastructure, offering rebates with sustainability-related incentives, and establishing standards for assessment. These efforts by G20 countries should be coordinated with other initiatives, such as the G20 Osaka Blue Ocean Vision and G20 Sustainable Finance Roadmap. Such strategic arrangements can foster green, inclusive development and accelerate the realisation of the SDGS.

To adapt to the complex and volatile environment of the electricity spot market, this study proposes a flexible resource characterization and identification method for Integrated Energy Systems (IES). To address the non-stationarity of multi-energy loads, a Variational Mode Decomposition (VMD) enhanced Temporal Convolutional Network-Graph Convolutional Network-Long Short-Term Memory (TCN-GCN-LSTM) spatiotemporal fusion model is developed, achieving significant improvements in forecasting accuracy compared to benchmark models. For electricity price forecasting, a hybrid Random Forest-Improved Attribute Generalization Importance Value-Complete Ensemble Empirical Mode Decomposition with Sample Entropy-Long Short-Term Memory (RF-IAGIV-CEEMD-SE-LSTM) model is constructed, which combines feature selection, subsequence decomposition, and noise reduction to capture temporal dynamics. Experimental results demonstrate that the proposed models reduce RMSE by up to 42.7% across load types and keep market-clearing deviations within 3% under multiple scenarios. The contributions of this study lie in three aspects: (1) developing a collaborative framework for multi-energy load and price forecasting; (2) proposing advanced spatiotemporal feature extraction and hybrid data preprocessing strategies; and (3) providing case-based validation with diverse market architectures. These results highlight the method’s strong potential for supporting intelligent scheduling and decision-making in modern electricity spot markets.

The purpose of this study is to determine the most important barriers for the improvements of the renewable energy projects in the hospitals. Within this context, a novel artificial intelligence-based fuzzy decision-making model is created. In the first stage, selected barriers are weighted by using artificial intelligence-based Spherical fuzzy CRITIC methodology. In the next process, emerging seven countries are ranked via Spherical fuzzy MAIRCA. An important novelty of the study is the integration of the CRITIC and MAIRCA methodologies with artificial intelligence. Owing to this situation, the weights of experts can be identified based on their qualification. This situation contributes to a more accurate analysis. The findings demonstrate that the most important factor in clean energy projects is operating costs. Similarly, technology and operational infrastructure factor also has an important impact on this situation. On the other side, the ranking results show that the most successful countries in clean energy projects in hospitals are Russia and China. India and Mexico are the last ranks in this regard. To increase the efficiency of projects, systems and equipment need to be analyzed regularly. In this context, the use of current technologies for renewable energy applications allows efficiency to be increased.

This study intends to address the issues of low recognition accuracy, delayed response, and insufficient efficiency of multi machine collaboration in unmanned aerial vehicle (UAV) inspections of transmission lines in extreme environments. Thus, the study proposes an intelligent operation and inspection framework that integrates multimodal perception, deep reinforcement learning, and dynamic scheduling, which is divided into three stages. In the first stage, this study proposes an UAV hardware system integrating Light Detection and Ranging (LiDAR), infrared thermal imagers, and high-resolution visual sensors to enhance data collection efficiency. In the second stage, this study then presents a Transformer-based multimodal data fusion algorithm to improve defect recognition accuracy and robustness. It also uses a deep reinforcement learning algorithm for dynamic path planning to optimize UAV inspection routes, thereby enhancing inspection coverage and energy efficiency. In the third stage, a dynamic task allocation and resource scheduling model combining Mixed Integer Programming (MIP) and heuristic rules is proposed to achieve real-time task allocation and resource optimization for multi-UAV collaborative inspection. Experimental results show that this method achieves an F1-score of 89.8% for defect recognition in extreme environments (improved by 11% compared with TransPathNet), shortens emergency response time to 45 s (improved by 28.6% compared with PPO-MultiDrone (Proximal Policy Optimization-Multi-Drone)), increases inspection coverage to 98.7% (improved by 10.7% compared with PPO-MultiDrone), reduces energy consumption by 28.4%, and achieves task completion rate and resource utilization rate of 95.6% and 91.5% respectively (Improved by 8.4% and 16.0% respectively compared to the optimal baseline Genetic Algorithm-Mask Region-based Convolutional Neural Network). This study provides a reference method for the further development of power Internet of Things defect detection.