Electricity Consumption Forecasting Analysis using Linear Regression Method, DKL 3.2 Method and BaU Scenario

Muhammad Zainal Roisul Amin (University of PGRI Banyuwangi)
Ratna Mustika Yasi (University of PGRI Banyuwangi)
Wahyu Setyo Aji (University of Jember)
Bambang Sri Kaloko (University of Jember)

Journal of Educational Engineering and Environment
Vol. 5 No. 1 (2026): Journal of Educational Engineering and Environment

Accurate long-term electricity demand forecasting is essential for ensuring reliable power system planning and sustainable energy development. Previous studies have generally focused on the application of a single forecasting approach, such as linear regression, or a limited comparison between two methods, resulting in insufficient evaluation of forecasting performance across multiple sectors and forecasting models. This research addresses this gap by conducting a comparative analysis of three forecasting approaches Multiple Linear Regression, Electricity Demand List Method 3.2 (DKL 3.2), and Business as Usual (BaU) for projecting electricity consumption in the ULP Kencong service area during the 2025–2029 period. The novelty of this study lies in the integration and comparison of these three forecasting methods within a single framework, combined with the utilization of customer growth, connected power capacity, and Land and Building Tax (PBB) indicators as forecasting variables. Forecasting simulations were performed using LEAP software and Microsoft Excel, while forecasting accuracy was evaluated using the Mean Absolute Percentage Error (MAPE). The results indicate that all sectors are expected to experience continuous growth in electricity consumption, with the industrial sector showing the highest increase. Among the evaluated methods, Multiple Linear Regression demonstrated the best forecasting performance, achieving the lowest MAPE values in three of the four analyzed sectors, namely residential (3.49%), business (6.68%), and social (4.77%) sectors. In contrast, DKL 3.2 produced the highest forecasting errors, particularly in the industrial sector (45.98%), while BaU showed moderate and relatively stable performance. These findings support the claim that Multiple Linear Regression is the most suitable and accurate method for long-term electricity consumption forecasting in the ULP Kencong region, providing a reliable basis for future electricity supply planning and infrastructure development.