Power Quality Prediction of Active Distribution Network Based on Clustering and Neural Network

Abstract

For the power quality problems that exist in the active distribution network with distributed generation, this paper predicts the power quality data in active distribution combining with various external factors affecting the power quality. Considering that the traditional BP neural network prediction is easy to fall into the local optimum solution, a new prediction approach based on K-means clustering and BP neural network for the steady-state indexes of power quality in an active distribution network is proposed. By adding the clustering step before prediction, the power quality influencing factors such as input time, light intensity, temperature and loads are clustered. After clustering, each type of error-minimum BP neural network is trained separately, and then the influencing factors of the prediction are identified by class markers. At last, the best BP neural network model of the corresponding category is called to predict the power quality of the active distribution network. The analysis of the example shows that compared with the traditional BP neural network prediction method to predict the steady-state indexes of the power quality, the use of clustering and neural network for active distribution network quality prediction can improve the prediction accuracy and has good feasibility.