Short-term load forecasting based on multivariate time series prediction and weighted neural network with random weights and kernels

Forecasting short-term load is a basic but indispensable problem for power system operations. This paper treats the forecasting problem as a multivariate time series forecasting problem. The electricity load and the corresponding temperature data are analyzed as correlative time series, and are reconstructed to the multivariate phase space. A neural network with random weights and kernels, which combines the advantages of the neural network and support vector machine including simple training and good generalization performance, is used as the forecasting model. Then, in order to further improve the forecasting performance, different weights are applied to the input data in the phase space according to the predictive value, and the resulting model is called weighted neural network with random weights and kernels. Simulation results based on the real world data set from the EUNITE competition show the effectiveness of the proposed method.