Enhancing wind energy prediction accuracy with a hybrid Weibull distribution and ANN model: a case study across ten locations in Java Island, Indonesia

Abstract

Accurate wind speed forecasting is essential for optimizing renewable energy (RE) systems, especially in coastal and island regions with high variability. This study proposes a hybrid predictive model that combines Weibull distribution parameters with artificial neural networks (ANN) to enhance forecasting accuracy. Using ten years of hourly NASA POWER data from 10 locations across Java Island, 24 scenarios were tested with varying combinations of Weibull and meteorological variables. Results demonstrate that incorporating both Weibull shape (k) and scale (c) parameters significantly improves performance, with the best configuration (Scenario 1) achieving a MAPE of 0.44% in Garut. Excluding one or both parameters sharply reduced accuracy, with errors rising up to 35.12%. Beyond technical accuracy, the findings emphasize the practical relevance of Weibull-informed ANN models for energy planning. Reliable forecasts support better wind resource assessment, grid integration, and investment decisions, reducing uncertainties that often hinder wind power deployment. By providing accurate and stable predictions across diverse locations, this approach offers policymakers and planners a robust tool to accelerate RE development and meet national energy targets.