Rahmadwati Rahmadwati, Abdul Hafiz Harmizi, Rusmi Ambarwati, Miroslav Markovic, Rini Nur Hasanah, Hadi Suyono
Accurate short-term electrical load forecasting (STLF) is essential for reliable and efficient power system operation, particularly at the substation level where load variations occur rapidly. This paper presents a comparative evaluation of fuzzy logic-based forecasting methods and a conventional operational approach for short-term load prediction at a distribution substation operated by the state-owned company PT PLN Batam in Indonesia. Three fuzzy inference systems - Mamdani, Sugeno, and Tsukamoto - are implemented and benchmarked against the Coefficient of Load (COL) method commonly used in PLN operational practice. Active power and system frequency are employed as input variables to represent real-time operating conditions under parallel transformer operation. Hourly load data collected over multiple weeks are evaluated using Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE). The results show that forecasting accuracy varies across different days due to changes in daily load characteristics. Overall, the Tsukamoto fuzzy inference system achieves the best performance, yielding the lowest MAPE and RMSE on most testing days. However, under stable load conditions, the COL method and Mamdani inference with appropriate defuzzification techniques demonstrate competitive accuracy. These findings indicate that no single forecasting method is universally optimal, highlighting the potential of adaptive or hybrid forecasting strategies for substation-level STLF. © 2026 IEEE.
Universitas Brawijaya, Department of Electrical Engineering, Malang, Indonesia; College of Applied Studies, Serbia
