Machine Learning-Based Prediction of Distributed Solar Adoption

The increasing adoption of distributed solar energy requires precise predictive models to evaluate deployment potential across diverse regions. This study leverages Random Forest, Support Vector Regression (SVR), and XGBoost to examine the key techno-economic drivers influencing solar adoption. Using a dataset derived from the NREL dGen model, data preprocessing techniques, including missing value imputation and feature engineering, were applied to extract critical indicators such as roof utilization, solar potential, and cost efficiency. To improve model accuracy, GridSearchCV was used for hyper-parameter tuning, and performance was evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> Score. The feature importance analysis identified solar potential as the key factor influencing adoption trends. The findings indicate that Random Forest and XGBoost perform better than SVR, achieving greater accuracy with fewer errors. The visualizations of error metrics and actual vs. predicted values offer clearer insights into model effectiveness and consistency. These findings highlight the effectiveness of machine learning methods in predicting solar adoption, offering useful insights for policymakers and industry professionals to enhance solar deployment strategies. Future studies could enhance predictive accuracy by incorporating spatiotemporal variables and broader economic factors, allowing for a more comprehensive analysis of solar adoption trends.