Short term load forecasting using evolutionary algorithm for Tajikistan

TL;DR: In this article , the authors proposed particle swarm optimization (PSO) to improve working support vector machine (SVM), SVM regression model is derived; also derived SVM with PSO.

Abstract: Load forecasting is a significant element in the energy management system of power systems. Precise load forecasting aids electric utilities to conduct decisions of unit commitment, reduction of spinning reserve capacity, and schedule device maintenance plan. Furthermore, load forecasting contributes to reducing the generation cost, and it is fundamental to the reliability of the power systems. On the other hand, short-term load forecasting is substantial for economic running. The forecasting precision directly affects the reliability, economy running and supplying power quality of the power system. Hence, finding the required load forecasting method to enhance the accuracy is valuable for forecasting precision. This paper proposed particle swarm optimization (PSO) to improve working support vector machine (SVM), SVM regression model is derived; also derived SVM with PSO. Support vector machine (SVM) model is adopted with and without PSO based on the historical load data and meteorological data of Tajikistan country, analysis the various factors affecting the forecast. The historical data and the load forecasting factors to be considered are normalized. The two parameters of SVM significantly influenced the model, and therefore it optimized using evolutionary algorithm.