Deep neural network application in real-time economic dispatch and frequency control of microgrids

Abstract In recent years, the development of microgrids has driven the reform of the electricity market, breaking the monopoly of traditional power grids and promoting the healthy development of the electricity market. However, the stability of microgrids is significantly impacted by the integration of various energy sources and numerous users. This study explores the application of an intelligent dynamic programming algorithm based on the deep neural network algorithm, combined with adaptive dynamic programming. Subsequently, an intelligent real-time power generation control algorithm (IRPGC) is obtained by introducing rejection operation improvement. Finally, a real-time integrated scheduling and control framework for microgrids is constructed. The research results showed that the IRPGC algorithm had an average error of less than 10 −5 after 5,000 iterations. Compared with mainstream algorithms, this algorithm achieved favorable results in frequency deviation evaluation indicators, with a frequency deviation fluctuation range of −0.073 to 0.013 Hz, an average error integral of 51.45, an absolute error integral of 0.54, and a time-weighted absolute error integral of 1.58 × 10 5 . In the practical application of real-time microgrid power generation scheduling and control framework, the optimal rejection threshold range was found to be [0.94, 0.97]. The aforementioned results indicate that the proposed method exhibits good control performance and application effectiveness, providing a reference for real-time power generation scheduling and control in microgrids.