Comparative Application of Artificial Neural Networks and ANFIS Techniques for Short-Term Load Forecasting in the Western Libyan Power Grid

Abstract

The stability and economic efficiency of modern power systems rely profoundly on accurate short-term load forecasting (STLF). This investigation presents a comparative assessment of two artificial intelligence methodologies,Artificial Neural Networks (ANN) and the Adaptive Neuro-Fuzzy Inference System (ANFIS) for STLF within the Western Libyan power grid. This network operates under considerable strain from extreme climatic conditions and infrastructural limitations, which introduce pronounced volatility and non-linearity into load patterns. Leveraging a comprehensive 2023 dataset from the General Electricity Company of Libya (GECOL), which integrates historical load data with critical meteorological variables, two models in MATLAB were developed and simulated. The findings reveal a decisive superiority of the ANFIS model, which achieved a remarkable average forecasting error of just 0.50%, starkly contrasting with the ANN model's error of 8.37%. This performance is attributed to the ANFIS architecture, which effectively marries the adaptive learning capabilities of neural networks with the transparent, rule-based reasoning of fuzzy logic. This synergy renders ANFIS an exceptionally accurate tool for short-term load forecasting in complex and uncertain environments like Libya.