Scholarlink

<p>The global transition to solar energy is vital for addressing climate change, reducing energy inequality, and achieving sustainable development. However, the long-term economic impacts of solar adoption at the regional level-particularly across space and time-remain underexplored. This study proposes a comprehensive machine learning framework to predict future economic benefits of solar energy adoption across 10,000 regions over a 10-year period, aligned with Sustainable Development Goals (SDG) 7 and 11. Using a high-resolution panel dataset incorporating geographic, socioeconomic, and infrastructural features, we apply both linear (Ridge, Lasso) and ensemble (Random Forest, XGBoost) regression models to forecast average regional economic benefit from cumulative solar adoption. Model benchmarking was conducted using a random 60/20/20 split (train/validation/test), followed by temporal validation where models trained on Years 1–7 were used to forecast benefits in Years 8–10, simulating forward-looking policy planning. To enhance model transparency and policy relevance, we employ SHAP (Shapley Additive Explanations) and Partial Dependence Plots for interpretability. Geospatial heatmapping using Kernel Density Estimation (KDE) reveals smooth and continuous spatial patterns of economic responsiveness. Our findings show that regions with higher income levels and sustained solar adoption tend to experience significantly greater economic returns. Ensemble models outperform linear models, achieving high predictive accuracy (RMSE &lt; 20, MAPE &lt; 1%).</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predictive modeling of regional economic benefits from solar energy adoption: a multi-model regression and spatial analysis approach

  • 期刊论文

  • 发表日期：

    2026年4月10日

查看全文
  • Kezia Saini,
  • Priyam Nath Bhowmik,
  • Himangshu Deka

摘要

The global transition to solar energy is vital for addressing climate change, reducing energy inequality, and achieving sustainable development. However, the long-term economic impacts of solar adoption at the regional level-particularly across space and time-remain underexplored. This study proposes a comprehensive machine learning framework to predict future economic benefits of solar energy adoption across 10,000 regions over a 10-year period, aligned with Sustainable Development Goals (SDG) 7 and 11. Using a high-resolution panel dataset incorporating geographic, socioeconomic, and infrastructural features, we apply both linear (Ridge, Lasso) and ensemble (Random Forest, XGBoost) regression models to forecast average regional economic benefit from cumulative solar adoption. Model benchmarking was conducted using a random 60/20/20 split (train/validation/test), followed by temporal validation where models trained on Years 1–7 were used to forecast benefits in Years 8–10, simulating forward-looking policy planning. To enhance model transparency and policy relevance, we employ SHAP (Shapley Additive Explanations) and Partial Dependence Plots for interpretability. Geospatial heatmapping using Kernel Density Estimation (KDE) reveals smooth and continuous spatial patterns of economic responsiveness. Our findings show that regions with higher income levels and sustained solar adoption tend to experience significantly greater economic returns. Ensemble models outperform linear models, achieving high predictive accuracy (RMSE < 20, MAPE < 1%).