The accurate prediction of photovoltaic (PV) power is crucial for planning, constructing, and scheduling high‐penetration distributed PV power systems. Traditional point prediction methods suffer from instability and lack reliability, which can be effectively addressed through interval prediction. This study proposes a short‐term PV power interval prediction method based on the framework of sparrow search algorithm (SSA)‐variational mode decomposition (VMD)‐convolutional neural network (CNN)‐gate recurrent unit (GRU). First, PV data undergo similar day clustering based on permutation entropy and VMD is applied to solar radiation signals with high correlation. Then, the hyperparameters of GRU are optimized by SSA according to the comprehensive evaluation indicator of interval prediction proposed in this study. Subsequently, quantile prediction results are obtained based on CNN‐GRU using the optimal parameters from SSA optimization. Finally, the prediction interval is composed of multiple quantile prediction results. A MATLAB R2022b program is developed to compare different prediction methods. The results demonstrate that compared to single neural network methods, the proposed method effectively improves the coverage width‐based criterion. In the interval prediction of sunny and rainy similar days, the comprehensive evaluation indicators of the proposed method are only 54.3% and 37.4% of the single GRU, respectively, indicating significantly improved interval prediction accuracy.

Against the backdrop of China’s “carbon peaking and carbon neutrality” goals, the abundant renewable resources in rural areas offer natural advantages for developing distributed integrated energy systems. However, the economic feasibility of advanced technologies such as carbon capture and hydrogen energy storage in rural applications requires further evaluation. This paper proposes a rural integrated energy system (RIES) architecture that integrates hydrogen energy storage and carbon cycling, encompassing six types of energy and material carriers—electricity, heat, gas, hydrogen, carbon, and biomass—and establishes a mixed-integer linear programming (MILP) model aimed at minimizing the total annual cost. Taking a typical rural village in northern China as a case study, capacity configuration optimization is conducted for five technical scenarios, and the impacts of the local generation ratio target, photovoltaic (PV) curtailment penalty coefficient, hydrogen storage efficiency, hydrogen energy subsidies, and equipment investment costs are analyzed.The results indicate that the scenario equipped with both carbon cycling and hydrogen storage achieves the lowest total annual cost, approximately 6.8% lower than that of the conventional RIES, with the local generation ratio increasing by about 4.0 percentage points and the PV curtailment rate dropping to 2.2%. Among these, carbon cycling is the primary source of cost reduction, while hydrogen storage further improves local power supply contribution and PV utilization. Sensitivity analysis indicates that increasing the local generation ratio can reduce electricity purchases from the grid, but at the cost of higher equipment investment; the PV curtailment penalty mainly affects PV capacity configuration and the level of PV curtailment, while hydrogen energy efficiency, subsidies, and equipment cost reductions influence the economic attractiveness of hydrogen energy equipment. The findings can provide a reference for the synergistic deployment of carbon cycling and hydrogen storage in RIES.

Fault diagnosis in grid-connected grid-forming hybrid energy storage station (GFM-HESS) systems is challenging because fault transients are jointly affected by converter control dynamics, multi-source electrical couplings, operating-condition variations, and measurement noise. To address these characteristics, this paper proposes a virtual synchronous generator (VSG)-constrained physics-informed heterogeneous temporal graph neural network (VSG-PI-HTGNN) for multi-class fault diagnosis. Based on VSG control characteristics and electrical relationships, 18-dimensional node-level features and eight-dimensional global physical features are constructed from ten monitored signals. These signals are further represented as a heterogeneous graph with five node types and eight predefined relation types, while node-type-specific transformations, heterogeneous graph convolution, learnable relation-scaling factors, and a primary–auxiliary dual-output framework are integrated for feature learning. A MATLAB/Simulink electromagnetic transient model is established to generate 3200 samples covering normal operation and nine fault conditions. At a signal-to-noise ratio (SNR) of 18 dB, the proposed model achieves 97.25% test accuracy and a macro-F1 score of 0.9727. Ablation results show that removing all physical information reduces the accuracy to 89.83%, while, under the unified experimental setting, the proposed model obtains higher values of the reported diagnostic metrics than the seven considered benchmark methods. Further evaluations show that the accuracy remains between 95.50% and 98.75% across SNR levels of 10–30 dB and reaches 95.38% with only 20% of the training data. Validation using an independently acquired hardware-in-the-loop (HIL) dataset further yields 92.75% accuracy and a macro-F1 score of 0.9282. Overall, these results indicate that the proposed method provides favorable diagnostic accuracy, noise robustness, data efficiency under limited-sample conditions, and simulation-to-HIL transferability under the evaluated conditions.

Energy serves as a fundamental cornerstone for the sustainable development of human society (Duić and Urbaniec 2012; Liu et al. 2026). Driven by the strategic goal of carbon peaking and carbon neut...