Research on Accurate Load Forecasting and Optimization Algorithm for Distribution Networks Based on Improved BiLSTM and DRL (High Load Density Areas)

Accurate load forecasting and optimization are fundamental to the reliable operation of modern distribution networks and intelligent electromagnetic energy transmission systems, particularly in high load density areas where complex multifactor interactions and significant load fluctuations present substantial challenges. This paper proposes an integrated algorithm based on improved Bidirectional Long Short-Term Memory (BiLSTM) and Deep Reinforcement Learning (DRL) to address these issues. First, the Maximal Information Coefficient (MIC) is employed to identify highly correlated load-influencing factors and construct a multidimensional feature set incorporating meteorological variables, temporal information, and historical load data. Second, chaotic mapping and an elite opposition-based learning strategy are introduced to enhance the Crested Porcupine Optimization Algorithm (CPOA) for hyperparameter optimization of the BiLSTM model, while a multi-head self-attention mechanism is incorporated to adaptively assign feature weights and improve forecasting performance. Finally, based on the forecasting results, a multi-time-scale optimization framework is established for the coordinated regulation of energy storage and flexible loads by formulating the problem as a Markov Decision Process (MDP) and solving it with the Deep Deterministic Policy Gradient (DDPG) algorithm. The proposed framework provides an effective solution for intelligent load prediction and adaptive energy management, offering practical support for electromagnetic energy distribution and resilient operation in next-generation smart power systems.