ARTIFICIAL INTELLIGENCE APPLICATIONS FOR PREDICTING RENEWABLE-ENERGY DEMAND UNDER CLIMATE VARIABILITY

Abstract

This quantitative, cross-sectional, case-study–based research investigates how artificial intelligence (AI) applications can improve renewable-energy demand forecasting under climate variability while being adopted in real operational settings. The study addresses the problem that utilities and renewable-intensive enterprises still underuse climate-variability indicators in AI models, limiting the accuracy and decision value of demand forecasts. Using a structured Likert five-point questionnaire and secondary operational and climate data from renewable-energy systems, 280 questionnaires were distributed and 214 valid responses were obtained (76.4% usable rate). Key constructs included organizational analytics capability, data quality and integration, climate-variability integration in AI models, AI model transparency, perceived forecast accuracy, trust in AI outputs, and intention to use AI-based forecasts. Reliability analysis showed Cronbach’s alpha values between 0.82 and 0.91, and multiple regression and correlation analyses were used alongside benchmarking of traditional regression and AI models. Compared with a traditional multiple regression model (MAPE 7.8%, RMSE 18.4 MW) and an AI model without explicit climate indicators (MAPE 6.1%, RMSE 15.2 MW), the climate-enhanced AI model achieved substantially lower error (MAPE 4.3%, RMSE 11.6 MW). Survey results indicated moderately high perceived forecast accuracy (mean 3.92) and trust (3.88), with intention to use AI forecasts averaging 4.03. Data quality and integration and climate-variability integration were the strongest predictors of perceived accuracy, while perceived accuracy and trust primarily drove intention to use. The findings imply that climate-aware AI forecasting delivers measurable accuracy gains but requires robust data pipelines and analytics capability to be trusted and embedded in renewable-energy planning and operations.