# Automated assessment of technological and financial drivers of greenhouse gas reduction in sustainable renewable energy systems

## TL;DR

This study explores how renewable energy systems can reduce greenhouse gases by analyzing technological and financial factors using predictive modeling.

## Contribution

The study introduces a dual-perspective analysis combining global sensitivity analysis and explainable machine learning to reconcile structural and predictive drivers of emission reduction.

## Key findings

- Energy storage efficiency is the most significant factor in reducing greenhouse gas emissions.
- Financial incentives play a crucial role in short-term emission reduction predictions.
- The CAAO configuration showed the best predictive performance and faster convergence for energy planning.

## Abstract

This study analyzes the capacity of renewable energy facilities to reduce greenhouse gas emissions using feature-based analysis approaches. The main goal is to identify the technological, economic, and environmental elements that most substantially influence emission reduction, serving as a basis for strategic planning and policy development. The dataset includes multiple renewable energy sources and financial variables. Predictive modeling was conducted via CatBoost Regression (CAT R) and Random Forest Regression (RFR), along with hybrid optimization via Transit Search Optimization (TSP) and Arithmetic Optimization Algorithm (AOA). Among the assessed configurations, the CAAO configuration not only achieved the highest predictive performance but also converged faster, demonstrating computational efficiency advantageous for real-time and large-scale energy planning. Feature analysis utilizing SHAP values, K-fold cross-validation, and sensitivity evaluation via the FAST method revealed that energy storage efficiency is the predominant factor, followed by financial incentives, underscoring the significance of both technological and economic aspects in emission reduction strategies. These findings offer an initial investigation and pragmatic suggestions rather than conclusive determinations. The findings indicate that feature-oriented assessments, when integrated with sophisticated predictive modeling, may substantially improve renewable energy planning and facilitate the formulation of context-specific, low-carbon policies. Importantly, by jointly employing variance-based global sensitivity analysis (FAST) and explainable machine learning (SHAP), the study reconciles an apparent discrepancy between structural system drivers (e.g., energy storage capacity) and predictive policy drivers (e.g., financial incentives). This dual-perspective analysis demonstrates that while storage dominates the physical response of emission reduction, incentive mechanisms primarily govern short-term predictive variability, offering a nuanced interpretability framework rarely achieved by single-method studies.