Closed-loop forecast–optimize–control for renewable-rich microgrids under uncertainty: LSTM forecasting, hybrid GA–PSO dispatch, and adaptive voltage–frequency regulation

A lot of renewable energy sources can cause quick, nonstationary uncertainty that makes it hard for smart microgrids to use decoupled forecast, scheduling, and control pipelines. This study suggests a single closed-loop architecture for predicting, optimizing, and controlling that strongly links short-term renewable forecasting, multi-objective dispatch, and adaptive voltage-frequency regulation for microgrids with a lot of renewable energy. To begin, we utilize Long Short-Term Memory (LSTM) models to guess how photovoltaic and wind power will change over time. Second, these predictions are given to a hybrid Genetic Algorithm–Particle Swarm Optimization (GA–PSO) scheduler that finds the best balance between dispatch cost, bus-voltage variation, and feeder losses while also meeting DER, battery, and network limits. Third, forecast residuals help gain-scheduled adaptive voltage–frequency regulation, which makes it possible to quickly fix problems when renewable energy sources or loads change. The framework is tested on a high-fidelity MATLAB/Simulink microgrid that has PV, wind, diesel power, and Li-ion battery storage and runs for 24 h with typical disturbances. Compared to traditional droop control with fixed scheduling, the suggested solution cuts the bus-voltage RMSE from 0.0324 p.u.–0.0125 p.u. (about 61%), the daily feeder losses from 52.8 kWh to 42.9 kWh (about 19%), and the time it takes to recover from a disturbance from 7.2 s to 1.8 s (about 75%). The forecasting layer has an RMSE of 0.94 kW, which means it can accurately estimate short-term renewable energy. These results show that tightly linking learning-based forecasting, hybrid metaheuristic scheduling, and residual-aware adaptive control can greatly improve the stability, efficiency, and resilience of a microgrid when there is uncertainty about renewable energy sources. This approach is also suitable for future hardware-in-the-loop and edge-control implementation.

Hali tarjima qilinmagan