Urban-Rural Energy Disparities in Focus: An XAI Approach to Biomass Allocation Strategies

This research explored how explainable AI (XAI) can inform more just, efficient, and regenerative biomass policies — including better distribution of alternative biomass sources like crop residues, identifying sustainable options, reducing emissions, and improving rural livelihoods.

Transcript

1. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation Leveraging XAI for Policy Action Ronald Atuhaire, CoCIS, Department of Computer Science, Makerere University, Uganda GBEP AWARD FINALIST
2. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation 95% biomass dependence → 2.6% annual deforestation rate (World Bank, 2020) Urban-Rural Divide: Rural: 72% lack clean cooking → 15hrs/week fuel collection Urban: 34% reliance but grid instability (UBOS, National Household Survey 2021) Research Context Problem Statement
3. Bioenergy potential of agricultural and forest residues, Collins Okello & et.al Science Direct
4. Methodology Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation GAT Random Forest TabTransformer Evaluation Explainability Training Raw Data Pre-processing EDA Data Engineering a Best Model Step 1 Problem Step 2 Data + EDA Step 3 ML Models
5. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation Exploratory Data Analysis Top 10 Districts by BE Potential Biomass Energy Potential - Districts
6. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation Exploratory Data Analysis Uganda District-Level BE Potential showing region wise didtribution Correlation heatmap for the different dataset features
7. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation Key Findings Top Predictor: Hardwood stocks (42% impact, per SHAP). Spatial Insight: 63% of high-loss areas border national parks (GAT maps). Policy Impact: 5% conservation → +210,000 GJ/year in Rakai district.
8. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation Explainable AI Actual Vs Predictions of RF, GAT & TabTransformer RF Feature Importances (Gini)
9. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation Explainable AI Intergrated Gradients - GAT Feature Attribution SHAP Plot - Model Output (RF)
10. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation Energy Conversion 1 kg dry firewood = 15 MJ (LHV, 20% moisture) 210,000 GJ/year savings = 14,000 tons firewood spared ≡ 840,000 mature trees* (assuming 50 kg/tree)
11. Uganda Energy Policy Aligns with Uganda’s 2023 Energy Policy Deadlines Government targets 60% clean cooking access by 2027 Timing Why This Research Matters NOW African Forests Climate Tipping Point for African Forests. UNEP warns Uganda may lose all forests by 2040 at current rate. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation
12. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation Dashboards Real-time dashboards for policymakers. Future Work Conclusions + Next Steps Satellite Data Sentinel-2 satellite integration. Hybrid models Combine Multiple models or even include knowledge distiilation
13. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation Ronnie A Lead Christine T DevOps George S ML Engineer Meet the Team
14. Urban-Rural Energy Disparities: An XAI Approach to Biomass Allocation ronlinx6@gmail.com Email Address +256 703 151 746 Phone Number Let's Connect