Neighbor histories for short term electrical load forecasting

This paper shows the application of the Neighbor Histories (NH) algorithm to the problem of short term electrical load forecasting in a utility company. This algorithm is a simple application of embedding theorems recently used in chaotic time series prediction. The choice of the parameters of the algorithm is usually done manually by trial and error. In this paper the possibility of automatic selection of parameters is investigated in order to obtain an easily customizable prediction tool. The basics of the algorithm are presented along with some experimental results. Some modification are proposed and tested, showing the improvements in the predictions. The NH algorithm with the automatically selected parameters is finally compared against a NN predictor.