Maximal electrical load modeling and forecasting for the tajikistan power system based on principal component analysis

Abstract

The article presents the results of the long-term forecasting for maximal power load daily graphs in the North Tajikistan power system with the use of principal component analysis orthogonal decomposition. The primary data for forecasting are maximal power load daily graphs in winter (January) and summer (July) periods from 2011 to 2015 which provide the data for the data matrix. The orthogonal decomposition of the principal components analysis is performed for uncentred daily graphs. The eigenvectors of the covariance matrix K obtained by the data matrix P make for a single orthogonal basis which performs the mapping and forecasting for both winter and summer maximal daily graphs. Here we have performed the analysis of interrelation between the orthogonal decomposition principal component analysis and the form of the studied power load daily graphs at the specified typical daily intervals (morning and evening maximum, daytime and night hours) by the example of the maximal load registered in the Northern Tajikistan power system. The identified dependencies were used to improve accuracy of the power system maximal load long-term forecasting. The forecast for 2016 was carried out in the framework of the first three principal components by the least-squares method taking into account the expected load growth and interrelations of the principal components with the daily graphs' form. The average relative error of forecasting for January 2016 amounted to no more than 6.5%.