Electric energy forecasting in crude oil processing using Support Vector Machines and Particle Swarm Optimization

In this paper, support vector machines (SVMs) are applied in predicting fuel consumption in the first phase of oil refining at oil refinery ldquoNIS Rafinerija Nafte Novi Sadrdquo in Novi Sad, Serbia. During cross-validation process of the SVM training particle swarm optimization (PSO) algorithm was utilized in selection of free SVM parameters. In particular widths of radial basis functions, as well as widths of regression tube and penalty factor were optimized by means of PSO. Incorporation of PSO into SVM training process has greatly enhanced the quality of prediction.