Industry load composition proportion forecasting of substation based on SVM

Abstract

A new methodology based on support vector machines (SVM) for the industry load proportion forecasting of a substation is presented to solve the problem that parameters of substation composite load model are randomly time-varying. The SVM algorithm is used to forecast a substation daily load curve and extract characteristic quantities of the substation daily load. Based on this, typical characteristic quantities of each industry are obtained through fuzzy C-means clustering with the consumer daily load curve from load control system and then project weights on the substation daily load characteristic quantities respectively. Load proportion of each industry is finally worked out by further calculation of the weights. According to the characteristics of a region's electricity, this prediction method is taken to forecast industry load composition proportion of a substation in the region on its summer peak load day. The result shows that the approach is consistent with the actual operation of the grid.