Prediksi Beban Listrik PT. PLN (Persero) Area Semarang Menggunakan Metode Support Vector Regression

Abstract

The trend of the increased electricity consumption and the unstable of monthly peak load make PT. PLN (Persero) as the electricity service provider needs to conduct a production planning in order to perform the electricity power system maintenance scheduling as well as fuel reserve provision to keep the sustainability of electricity production. The electricity production plan for the purpose of system maintenance scheduling and fuel reserve provision was conducted by mid-term load forecasting. This research presented the result of load forecasting using Support Vector Regression method using predictor feature consisted of load, connected power, the number of electricity subscribers, and PDRB-ADHB. The data used were 75 data (Juni 2011-Desember 2017) from PT. PLN (Persero) Semarang Area and 7 data (2010-2016) from BPS Semarang. The result of the research showed the obtained of error value using MAPE was 4,03 % for the best parameter value C = 10 8 , ɛ = 10 6 , and Kernel Linear function, with the best predictor feature was connected power, and the number of electricity subscribers. The prediction data of October-December 2017, the result of MAPE error value was 3,0384 %.