Short-Term Building Load Forecasting Based on Data Mining Technology

Abstract

Short-term building load forecasting is an important part of building energy efficiency management system to assess and diagnose energy consuming subsystem, optimize control and schedule planning. In this paper, K-Means clustering is used to cluster the daily load curve and the DBI evaluation index is used to determine the clustering number. In addition, the Pearson correlation coefficient is used to calculate the correlation coefficient between the load and its influencing factors. And then the classification rules are established by probabilistic neural network (PNN) to find out the basis of the clustering result. Finally, the BP neural network model optimized by particle swarm optimization is used to predict the load value of one day in the future. The prediction and verification of a building load data of Shanghai proves the rationality and effectiveness of the model.