An Application-Oriented Analysis on Forecasting Models of Installed RES Capacity: Solar and Wind Scenario in Italy by 2030

The global shift towards renewable energy highlights the urgency of a sustainable transition. As solar and wind capacity expand rapidly, their integration into existing infrastructures demands accurate forecasting to support strategic planning. This paper offers an application-focused analysis of forecasting models for installed renewable energy capacity, focusing on solar and wind energy in Italy by 2030. The proposed approach combines traditional statistical methods with machine learning algorithms in a unified evaluation framework, offering a novel comparative assessment that bridges methodological gaps identified in the literature. To improve model performance, preprocessing strategies including smoothing and dataset reduction were implemented. The comparative analysis highlights each method's strengths and limitations in capturing the growth patterns and delivering accurate long-term predictions. When compared to the baseline Linear Regression model, several approaches achieved substantial error reductions. In particular, the LSTM architecture achieved improvements of up to 98.8% (wind) and 98.8% (solar), while the best feedforward neural network configurations reached 92.6% and 98.2% error reduction, respectively. These findings highlight the potential of advanced machine learning models to significantly enhance longterm RES capacity forecasts, although trade-offs exist between statistical performance and alignment with national targets. This work contributes to identifying the most effective forecasting strategies to support informed policy decisions and optimize renewable energy integration. Future research will explore hybrid approaches, inclusion of additional predictive variables, and scenario-based modeling to further improve the robustness and policy relevance of long-term forecasts.