Load forecasting and energy storage optimization control for distribution networks with high distributed photovoltaic penetration

With the advancement of carbon peaking and neutrality goals, the penetration of distributed photovoltaic (PV) in distribution networks is rising, but its intermittency and volatility cause voltage violations, reverse power flows, and higher losses. To address this, we propose an integrated “forecast-optimization-control” framework. A RIME (Rime optimization algorithm)-CNN (Convolutional Neural Network)-LSTM (Long Short-Term Memory)-Attention mechanism model is developed, where the RIME algorithm optimizes hyperparameters and weights, and CNN, LSTM, and Attention are combined for high-accuracy load and PV forecasting. Based on the forecasting results, a multi-objective storage optimization model is formulated to minimize voltage vulnerability, losses, and storage scale, solved by an improved Multi-Objective Particle Swarm Optimization (IMOPSO) for coordinated configuration and control. Simulations on the IEEE 33-bus system verify that the method reduces voltage fluctuations and losses, enhances PV accommodation and storage utilization, and shows scalability to larger networks.