Techno-economic and environmental optimization of hybrid renewable energy systems using HOMER Pro and machine learning surrogate models

Decentralized Hybrid Renewable Energy Systems (HRESs) represent a viable pathway to achieving net-zero targets and universal energy access, yet high capital costs and the intermittency of solar and wind resources continue to impede deployment. This study addresses these challenges by developing a comprehensive techno-economic, environmental, and machine-learning-assisted optimization framework for a diesel-free HRES serving eight remote communities in Obubra Local Government Area, Cross River State, Nigeria. Five system architectures integrating solar photovoltaics (PV), small hydropower (SHP), biogas generators (BG), and battery energy storage systems (BESS) were evaluated using HOMER Pro across a design space of 1,215 feasible configurations. The study explicitly incorporates residential, commercial, public, and productive-use electricity loads, to improve realism in rural load estimation. A five-parameter sensitivity analysis was conducted, individually varying resource and economic variables around their design-case values, to capture macroeconomic and resource uncertainty. Results identify HRES-1 (PV, SHP, BG, BESS, and converter) as the optimal configuration, achieving a Levelized Cost of Electricity (LCOE) of $0.112/kWh and a Net Present Cost (NPC) of $463,371, with a simple payback period of 2.75 years. HRES-4 and HRES-5, which exclude biogas generators, achieve zero carbon emissions, while HRES-1 produces only 9.1 kg of annual CO 2 emissions. To accelerate HRES performance evaluation, 27 machine learning regression algorithms were benchmarked as surrogate models using MATLAB R2025b, trained on HOMER Pro simulation outputs. The Squared Exponential Gaussian Process Regression (GPR) model achieved the highest LCOE prediction accuracy (R² = 0.99905, RMSE = 0.00393), while the Bagged Trees Ensemble model delivered the most accurate CO 2 emission forecasts (R² = 0.99871, RMSE = 0.33794) on the test set. The proposed framework provides a scalable, replicable template for sustainable off-grid electrification in tropical Sub-Saharan Africa. It offers clear policy guidance for renewable energy subsidies and rural mini-grid financing in developing economies.