Multi-level energy–information model and physics-informed FUZZY–GA–LSTM adaptive algorithms for hybrid energy monitoring

Abstract

This paper proposes a multi-level energy–information model and a physics-informed PI-Fuzzy–GA–LSTM–DT algorithm for real-time monitoring and adaptive control of Smart-Grid-enabled base stations. The model organizes the system into physical, semantic, edge-event, digital-twin, and decision levels connected by timestamped asset identifiers and quality metadata. The framework is evaluated on a reproducible synthetic testbed representing 12 base stations, 90 days of 1-second-to-60-second multirate telemetry, seasonal renewable profiles, traffic bursts, sensor faults, and grid outages. In this benchmark, the proposed model achieves 4.9% renewable-power MAPE, 1.9% SOC RMSE, 0.8 kWh/day unserved energy, and 2.8 h/day diesel runtime, while 94.6% empirical coverage is obtained for nominal 95% forecast intervals. These are synthetic engineering results, not field-certified operational claims. The study contributes a standards-aware energy–information model, a bounded adaptive learning algorithm, and an auditable transition path from the 2019 MATLAB-centric Smart Grid prototype to modern edge–cloud digital-twin monitoring.