DUAL-FORECASTING AND ADAPTIVE-WEIGHTED HYBRID GA-PSO ENERGY MANAGEMENT FOR IOT-ENABLED RURAL SOLAR MICROGRIDS

Reliable rural electrification requires solar microgrid controllers that can predict renewable-power variation, protect battery health, and maintain essential loads under communication constraints. This paper proposes an IoT-enabled rural solar microgrid using dual CNN-BiLSTM-Attention forecasting and an adaptive-weighted Hybrid Genetic Algorithm–Particle Swarm Optimization energy-management system. Separate forecasting models predict one-hour-ahead photovoltaic generation and rural load demand using historical power, weather, time, and appliance-state features. The optimization weights are dynamically updated based on forecasted power deficit and battery state-of-charge risk. The proposed controller is evaluated for a 250-W PV, 12-V battery rural microgrid scenario. Compared with rule-based control, the proposed method improves energy-utilization efficiency from 72.1% to 91.2%, reduces unserved load from 58.4 to 14.6 Wh/day, decreases the battery-stress index from 0.71 to 0.31, and increases reliability from 82.6% to 96.5%. The PV and load forecasting models achieve RMSE values of 7.46 W and 5.03 W, respectively. At a 60-s IoT update interval, the communication model gives 91.23 ms latency, 1.39% packet loss, and 98.61% packet-delivery ratio.