Application of Different Membership Function for Short-term Load Demand Estimation: A Neuro-Fuzzy Approach

Abstract

Electricity management and power sources must be properly managed to ensure efficient usage of electricity. This necessitates the requirement for accurate electrical demand forecasting to ensure that electricity generation is sufficient to meet demand. This paper investigates two methods of adaptive neuro-fuzzy inference systems (ANFIS), grid partitioning (GP) and subtractive clustering (SC), in predicting the load demand of Abuja City. The input parameters used for modelling are wind speed, solar radiation and air temperature. The performance of the two models - ANFIS-GP and ANFIS-SC - are compared using determination coefficient ($R^2$) and mean absolute percentage error (MAPE). The results showed that ANFIS-SC outperformed ANFIS-GP with a better goodness-of-fit ($R^2$) of 93.21\% and lesser error result (MAPE) of 8.99\% which proves it can be used for load demand forecasting.