ESTIMATED USE OF ELECTRICAL LOAD USING REGRESSION ANALYSIS AND ADAPTIVE NEURO FUZZY INFERENCE SYSTEM

The rapid growth of Indonesia's population increases electricity consumption. Unfortunately, this growth is not followed by the development of new electrical energy sources. Therefore, there needs to be a comprehensive study to ensure that the electrical power supply meets the people's electricity needs. The estimated electrical load is one of the important factors in determining the number of the electrical power system. This study deals with the electrical load estimating using an Adaptive Neuro-Fuzzy Inference System (ANFIS) and Linear Regression methods. The data used for this estimation were data on the use of electricity in the provinces of Central Java and the Special Region of Yogyakarta, Indonesia. The electricity usage data were measured every half hour for six months, so there were 48 data loads in one day. This study uses the methods of analysis with regression analysis and ANFIS. The ANFIS consists of two inputs and is measured using two rules. The Fuzzy inference system applies the first order of the Takagi-Sugeno-Kang model. The result of estimation shows the Public Electricity Company Ltd. or Perseroan Terbatas Perusahaan Listrik Negara (PT PLN) has a good accuracy rate, with a Mean Absolute Percent Error (MAPE) error of 4.47%. On the other hand, based on the simulation results, the Regression Analysis method is fairly accurate for estimating electrical loads with a MAPE error of 1.92%, while the ANFIS method shows the most accurate estimation result with MAPE error of less than 0.96% of the actual value.