Power Grid Dispatching Optimization Assistant Decision-making Model Based on Deep Learning

Abstract

At present, the power grid dispatching assistant decision-making model relies on the model-driven solution method, which leads to too long a calculation time. Therefore, an auxiliary decision-making model for power grid dispatching optimization based on a deep learning algorithm is proposed. The objective function of power grid dispatching optimization is defined for the power grid connected with multiple renewable energy sources. Relying on cloud computing technology, the power grid dispatching and monitoring data are obtained and saved as different knowledge systems. The long and short-term memory network is selected from the deep learning field, and the Bi-LSTM network structure unit is designed. The scheduling decision model is constructed using the bi-layer Bi-LSTM neural network, and the optimal scheduling decision is obtained by deep learning of the historical scheduling data. The calculation results show that the longest time of the model is 0.04s. The shortest time is only 0.02s, which improves the efficiency of auxiliary decision-making and enhances the stability of power grid operation to a certain extent.