A data-driven analysis of levelized cost dynamics, operational performance, and emissions in food waste-based hybrid renewable energy systems

This study evaluates the techno-economic and environmental performance of a grid-connected photovoltaic (PV)-biogas hybrid energy system using food waste (FW) in a campus setting. Unlike the conventional studies focussed on a single optimal configuration, the present study analyses the solution space to examine system behaviour across a wide range of feasible designs. The system is modelled in HOMER for a load demand of approximately 951 kWh/day (∼347 MWh/yr) considering varying levels of FW availability. The optimisation-derived solution space that represents all feasible system configurations generated through simulation is analysed using regression, machine learning (ML), and clustering to examine the behaviour of levelized cost of energy (LCOE) across the designs. The findings show that LCOE variation remains stable, regression models achieving strong performance with R ² > 0.98 and low prediction errors. Additionally, the ML models capture non-linear behaviour and threshold effects under varying levels of information. Four operating patterns are identified: grid dominant (GD), grid integrated mix (GIM), hybrid biogas optimised (HBO), and biogas intensive (BI). GD relies on 75% grid energy while BI derives 77% from biogas. Moreover, the environmental analysis shows that CO ₂ emissions decrease by about 76% in HBO and 75% in BI compared to GD. The carbon intensity decreases from 0.602 to 0.141 kg CO ₂ /kWh. However, higher renewable penetration also increases LCOE, indicating a cost-carbon trade-off. Overall, the study shows how optimisation-derived data can be translated into economic insights and system-level understanding beyond a single optimal solution. This further enables the systematic interpretation of cost and performance under varying conditions.