Robust Dynamic Economic Dispatch in Smart Grids Using an Intelligent Learning Technology

Dynamic economic dispatch is crucial for implementing real-time and efficient power management, aiming to minimize the total cost of management over a specified period. Nevertheless, the demands for emergency response and privacy protection in actual power systems pose significant challenges for dynamic economic dispatch. To this end, we develop an intelligent learning technique, involving an iterative forecasting method and a distributed reinforcement learning algorithm, to improve the robustness for effectively meeting these challenges. Specifically, an iterative forecasting method is proposed based on the long short-term memory model, chaotic time series, and fractal interpolation algorithm, achieving an accurate and robust rolling load forecasting for emergency response. Furthermore, we design a new reinforcement learning algorithm based on general function approximation and distributed optimization. This algorithm successfully addresses the dynamic economic dispatch problem over a directed communication network, guaranteeing the privacy of interactive information. Ultimately, we provide several case studies to showcase the effectiveness of the proposed technique in dynamic economic dispatch.