Short-Term Wind Power Forecasting Using Optimally Tuned Adaptive Neuro-Fuzzy Inference System

Abstract

Wind energy is well known for its high volatility, primarily due to its strong dependence on climatic conditions. To achieve optimal management of smart grids, it is essential to develop accurate forecasting tools capable of predicting the future production of wind power plants. The existing literature presents a wide range of forecasting methods, which can generally be categorized into four main groups: statistical methods, physical methods, machine learning approaches, and hybrid models. This study aims to investigate a novel hybrid forecasting technique that integrates An Adaptive ANFIS (Neuro-Fuzzy Inference System) optimized using the Manta Ray Foraging Optimization (MRFO) algorithm is developed to improve the precision and robustness of wind power forecasting. The effectiveness of the proposed method is validated using real-world wind power data collected from the Sotavento Galicia wind farm in Spain. The forecasting performance is further benchmarked against several conventional approaches, including DT (Decision Tree) and RF (Random Forest) models, traditional Support Vector Regression (SVR) and hybrid SVR-Gray Wolf Optimization (SVR-GWO) model.