Short-term load forecasting of power system based on similar day method and PSO-DBN

摘要

Short-term load forecasting is an important part of the energy management system (EMS) and the basis for the safe operation of the power system. In view of the problems of high dimensionality, long time and general precision in load forecasting of existing artificial intelligence algorithms, this paper proposes a short-term load forecasting method based on particle swarm optimization and deep belief network (PSO-DBN). The characteristics of this method are: (1) Calculate the similarity according to the date distance, the type of the week and the meteorological characteristics, and select the similar day according to the similarity; (2) Replace the traditional historical daily load with the similar daily load as the partial input of the algorithm. Improve the prediction accuracy; (3) Construct the DBN prediction model to overcome the problem that the support vector Machines (SVM)training time is long and the BP neural network method is easy to fall into the local optimum; (4) Optimize the weight by usingPSO to optimize the DBN algorithm. Further reduce the degree to which the algorithm is affected by the initial value and reduce the number of iterations. The simulation example demonstrates the effectiveness and good engineering application value of the proposed method.