Short Term Forecasting of Electricity Load: A Comparison of Methods to Paiton Subsystem East Java & Bali

Abstract

Electricity consumption in Indonesia is expected to continue to grow by average of 6,5% per year until 2020. Therefore, PT. PLN had to make an effective subsystem that can provide electrical energy based on customer needs. The electrical energy is converted from the mechanical energy and can’t be stored. Because of that reason, if the electrical energy isn’t channeled properly then PT. PLN will suffer losses. It is necessary to plan a proper distribution system of electrical energy. The aim of this research is to predict short-term electricity consumption for Paiton’s subsystem in East Java Indonesia by using ARIMA and Multilayer Perceptron. The best model is measured based on MAPE, SMAPE, and RMSE value in data sample. The result of the analysis shows that Multilayer Perceptron method provides better accuracy rate for electricity consumption forecasting in Paiton subsystem based on peak load compared to ARIMA

Short-Term Electricity Load Forecasting Model Based DSARIMA

“…Based on the final parameter estimation results in Table 4, the ARIMA coefficient parameters are obtained as follows: AR ( Based on the prediction model parameters, the DSARIMA model ( [1,2,5,6,7,11,16,18,35,46], 1, [1,3,13,21,27,46]) , (1,1,1) 23 (0,0,1) 445 with the following model equation:…”

“…Based on Table 1, the data has met the white noise criteria with a p-value greater than the error tolerance value 𝛼 = 5%, with an alpha significance level of less than 0.0001. In addition, the model at the time of initial estimation has an improvement pattern with 3 MA parameters, namely MA(1,1), MA(2,1) and MA (3,1), so these three parameters must be included in the model estimation. While the residual assumption test which includes the assumption of white noise must meet the independent criteria and have a normal distribution (0, 𝜎 $ ).…”

“…Time series analysis has become a research trend, where the results of this study are very effective with a very high level of accuracy in predicting future electrical loads. Forecasting short-term electrical load based on time series has been carried out by several Indonesian researchers [3]- [5]. This time series forecasting method is then also compared with artificial intelligence methods, as has been done previously [6], [7].…”