Artificial intelligence and energy transition: An LSTM predictive modeling of sectoral climate trajectories in Europe, Russia, and India

Abstract

This study presents a comparative analysis of the impact of renewable and non-renewable energy dynamics on sectoral CO₂ emissions in the European Union, Russia, and India. Using daily electricity production data disaggregated by source—renewables (solar, wind, hydro) and fossil fuels (coal, oil, gas)—we examine emissions trajectories across five key sectors: ground transport, international aviation, industry, power generation, and residential. Emissions are forecasted using Long Short-Term Memory (LSTM) neural networks, which capture nonlinearities, asymmetries, and memory effects in energy–climate interactions. Our approach integrates hybrid data (energy, climate, economy), fine sectoral granularity, and advanced AI modeling. The results reveal distinct regional vulnerabilities: the EU, though advanced in renewables, remains exposed to gas-related risks; Russia's heavy reliance on fossil fuels increases its sensitivity to geopolitical shocks; and India faces growing climate risk amid surging energy demand. LSTM forecasts indicate that fossil fuel–reliant sectors (transport, aviation) show the most volatile emissions, while electrified sectors exhibit greater stability. This study contributes a robust, AI-driven, multi-sector, multi-region framework to the energy transition literature and offers actionable insights for policymakers. It underscores the need for differentiated policy tools tailored to specific sectoral and regional contexts.