Comparison of Four Time Series Forecasting Methods for Coal Material Supplies: Case Study of a Power Plant in Indonesia

Abstract

Coal is the main fuel in the production process at PT PJB UBJ O&M Tenayan. As a raw material, coal needs to be considered in terms of supply to prevent losses (depreciation in caloric content) in case of oversupply. This study aimed to compare four forecasting methods for coal material supply. The four methods of time series forecasting are the moving average method, the weighted moving average, the single exponential smoothing, and the linear regression. Forecasting error calculations used the smallest MAD, MSE, and MAPE error parameters, whereas the tracking signal was used to monitor the forecasting results. The data required were coal supply and demand. Based on the data processing obtained, results of this study show that the best method is linear regression with the results of the MAD value of 13,285.63, MSE of 228,778,800, and MAPE of 15.04%. Based on the results of the tracking signal, the forecasting results were within the control limits, which shows that the linear regression method is the best forecasting method that can be applied to control coal supply in the next period.