Data-driven vulnerability-aware framework for hybrid renewable energy system with integrated satellite-terrestrial network

Operational security of Hybrid Renewable Energy System (HRES), amid variable operating conditions and multisource information coupling, is ensured by prioritizing the affecting components most susceptible to failure. Despite its prudence and reliability, manual overhaul suffers from manpower limitations and inefficiency, resulting in delayed responses and incomplete coverage in complex HRES. To facilitate effective allocation of overhaul personnel and equipment, a reliable forecasting process for future vulnerability spatio-temporal distributions is essential. Toward this end, an integrated sensing-assessment-localization fault-risk diagnosis framework is developed, namely the Differential Significance-based Physics-Informed Prediction Framework (DS-PIPF). Firstly, a Space-Air-Ground (SAG) cooperative observation network is formed to acquire multimodal operation and environment information across generation, grid, load, storage scenarios, which enables a unified multiscale data representation. Next, component risk is evaluated via the established Differential Significance Criterion (DSC), which can quantify both individual-component self-risks/multi-component joint-risks, characterize risk fluctuations under evolving operating conditions, thus distinguish critical factors and elucidate coupling and propagation pathways accordingly. Finally, physical constraints based on DSC are embedded in a Physics-Informed Neural Network (PINN) to determine initial values and interlayer weights for accelerated convergence, and importance sampling guided by joint-risk magnitudes is also employed to improve accuracy. Validation through a HRES empirical case study demonstrates the feasibility and effectiveness of the proposed framework during practical deployment.