Research on Real-Time Reconfiguration of Distribution Network Based on Data-Driven Approaches

As a key component of power generation, the transmission network and user demand, the distribution network is of great significance to safe and stable operation of the power system. However, in recent years, owing to the increase in complex loads and access to distributed power sources, the network operation has become increasingly complicated and uncertain. In view of the complexity of the power flow of the distribution network and the uncertainty of the distributed power supply, it becomes increasingly difficult to quickly restore the normal power supply of the power system after a fault occurs. In this paper, an improved heuristic algorithm combined with convolutional neural network (CNN) model is used to solve this problem. First, the reasons for the uncertainty of the distribution network are analysed, and a mathematical model is established for the uncertainty factors; then, the minimum network loss and voltage quality factors are selected as the objective functions of reconfiguration, and, at the same time, the corresponding constraints are determined; the algorithm is finally verified on the IEEE 136-bus test system, which validates the correctness of the used method. On this basis, the deep learning method is used to improve the reconfiguration rate and efficiency, that is, the convolutional neural network is used to train the historical data set, the one-to-one mapping relationship between the historical reconfiguration optimal solution and the distribution network topology is determined, and, finally, the model is applied to the test data set to verify its validity and correctness, so as to achieve the objective of real-time reconfiguration.