In the research that delves into the predictive analysis of renewable energy returns, this project has comprehensively considered multiple key indicators, aiming to enhance the accuracy of forecasts through both univariate and multivariate methods. In the multivariate analysis phase, we employed advanced dimensionality reduction techniques to select the most representative variables from a multitude of indicators, thereby constructing several forecasting models. By evaluating out-of-sample forecasts on 20% of the dataset, we found that the 3PRF model demonstrated exceptional performance in predicting monthly renewable energy returns. To further verify the model’s robustness and generalization capabilities, univariate and multivariate forecasting assessments were also conducted for quarterly and semi-annual periods. The results showed that the 3PRF model not only excelled in monthly forecasts but also maintained superior predictive performance in quarterly and semi-annual forecasts, consistently outperforming other models. These findings not only confirm the applicability and reliability of the 3PRF model across different timescales but also provide robust data support for investment decisions in the renewable energy market.