Hybrid Deep Learning and Reinforcement Learning Framework for Power Prediction and Optimization in Photovoltaic Storage Systems

This paper presents a novel hybrid deep learning and reinforcement learning (DNN-RL) framework for power prediction and control optimization in photovoltaic (PV) storage systems. The proposed method integrates a deep neural network (DNN) for accurate power generation forecasting and a reinforcement learning (RL) agent to optimize real-time control of the energy storage system. The algorithm was evaluated on three open-source datasets, demonstrating superior performance in terms of prediction accuracy, power loss reduction, and energy utilization efficiency compared to traditional machine learning and rule-based control methods. The results highlight the effectiveness of combining prediction and control in a unified framework for improved energy management. Future work will focus on extending the model to include more complex environmental factors and applying multi-agent reinforcement learning to enhance system-wide energy distribution coordination.