Agro-residue forecasting for biogas-based green rural electrification in India: A Gaussian process regression-based decision framework

Reliable feedstock availability remains a critical challenge for biogas-based microgrids, particularly in agrarian economies where agro-residue supply is highly dynamic. Existing planning approaches often rely on static estimates, ignoring temporal variability. This work proposes a probabilistic forecasting-based decision framework to assess agro-residue availability and corresponding biogas-based electricity potential. The methodology integrates Lasso-based feature selection with Gaussian Process Regression using a composite kernel. The proposed model, named CK-GPR, is trained on four decades of historical agro-climatic, agronomic, and policy data. The composite kernel captures multiple underlying patterns, including long-term trends, nonlinear dependencies, and stochastic variations. Model validation using K -fold cross-validation shows that Gaussian Process Regression outperforms Support Vector Regression, Random Forest, and Long Short-Term Memory models, even with a dot product kernel. Further improvement is achieved using the composite kernel ( M A P E = 2.412%, R 2 = 0.984). Rice being a major crop in India, forecasts for 2025–2027 indicate a biogas-based electricity potential of ( 152.9 − 155.36 ) ∗ 1 0 2 GWh and greenhouse gas reduction of ( 119.8 − 121.73 ) ∗ 1 0 5 tonnes C O 2 -equivalent and ( 5.24 − 4.85 ) % , ( 11.81 − 10.59 ) % , ( 5.49 − 5.22 ) % of national irrigation, rural household, and energy deficits, respectively. A sensitivity analysis and cross-crop validation using wheat further demonstrate the robustness and transferability of the proposed methodology. Bridging a critical gap between agricultural forecasting and biogas-based green energy system planning, a decision-support framework is proposed for microgrid operators, potential investors, and policymakers for planning capacity, deployment strategies, and feedstock supply chain logistics under biomass uncertainty.