PERAMALAN BEBAN LISTRIK BULANAN SEKTOR INDUSTRI MENGGUNAKAN SUPPORT VECTOR MACHINE DENGAN VARIASI FUNGSI KERNEL

Abstract

Electrical energy consumption in the industrial sector is one of the factors that accounted for the production cost, and efficiency of electrical energy use can reduce the production cost. Energy efficiency in industry can be applied by regulating the use of the electric load, it can be watched from the electrical load Characteristics. The monthly characteristics of the electrical load every can be predicted, so do the efforts to regulate the amount of load and provide electrical energy required. This research present the monthly electricity load forecasting in the industrial sector by SVM with a variety of Kernel functions. There are 3 (three) types of training data given on SVM, they are types months of the year, production data and time series data past the electrical load, with a variable input data between 1 year to 6 years. Variations Kernel functions used in SVM methods are Linear, quadratic, Gaussian RBF, polynomial and Multilayer Perceptron. This research produced the smallest forecasting MAPE of 5.33% with a 6 year input data training scheme with Gaussian RBF kernel function.

Keyword: Electric Load, Monthly Forecasting, SVM, Kernel Function