Short-term load forecasting based on multivariate linear regression

With the rapid development of micro grid, the power load forecast is important in system. Short-term load forecasting (STLF) plays an important role in the overall operation efficiency of micro grid. In order to improve the accuracy of STLF, this paper proposes a combined model, which is multivariate linear regression(Multi-LR) with multi-label based on K-nearest neighbor (K-NN) and K-means. We use multi-label and K-NN algorithm to give different weight of each cluster for the forecasting points and build models by Multi-LR. In this paper, the test data which include daily temperature (which include highest temperature and lowest temperature) and power load of a quarter of an hour from a community compared with the results only using Multi-LR to forecast power load, it is concluded that the combined model can achieve high accuracy and reduce the running time.