SHORT-TERM WIND POWER PREDICTION IMPLEMENTING THE GRADIENT BOOST ALGORITHM IN PARTICLE SWARM OPTIMIZATION-EXTREME LEARNING MACHINE MODEL

Abstract

As wind power output increases internationally, accurate wind power assesses are fundamental to maximizing the use of wind power and defend a safe and reliable electricity grid. We suggest a novel strategy that combines a gradient boosting algorithm with a limit learning machine optimized using particle swarm optimization (PSO-ELM), taking into account the inherent unpredictability and variability of wind power as well as the limitations of existing forecasting models. We first refine the ELM starting thresholds and input weights before building the PSO-ELM prediction model. A Gradient Boost technique is then used to generate a large number of PSO-ELM weak predictors. The input weights and thresholds of each weak predictor, which each includes a unique hidden layer node, are optimized using the PSO technique. The results from each weak predictor are combined and weighed to get the final forecast result using a robust wind power forecast model. We use measured data from Turkish wind turbines to verify the efficacy of our suggested strategy. We may evaluate the Gradient-PSO-ELM model's accuracy and dependability in predicting wind power output under real-world circumstances by contrasting the forecasts it produces with the actual measured data.. The findings demonstrate that the Gradient-PSO-ELM wind power prediction model has improved generalizability and accuracy.