Robust configuration optimisation of low-carbon integrated energy systems under variable equipment efficiency and source-load uncertainty conditions

Integrated energy systems (IES) are critical for decarbonising urban energy, yet their design faces challenges arising from renewable intermittency, stochastic occupant demand, and equipment performance under off-design conditions, an area often overlooked in existing optimisation models. This study proposes a two stage robust optimisation framework for IES that jointly addresses renewable supply variability, stochastic demand and off design equipment performance to address this gap. This approach links micro-level occupant behaviour and equipment physics with system-level capacity planning, yielding configurations that hedge uncertainty while maintaining operational adaptability. Introducing photovoltaic (PV) uncertainty increases optimal PV capacity by 16.4%. Accounting for off-design operation shifts the configuration towards dispatchable units, gas turbine, waste heat boiler, and absorption chiller, whose capacities increase by 21.0%, 17.5%, and 17.2%, respectively, with a 3.46% increase in total annualised cost. Sensitivity to the uncertainty budget Γ shows that reducing Γ from 14 to 0 lowers annualised cost by 2.24% and investment by 5.94%, with diminishing returns beyond Γ > 9. Seasonal dispatch confirms adaptive roles: electric chillers dominate summer cooling, gas boilers dominate winter heating, and waste heat boilers dominate transitional heating. The framework supports uncertainty-aware planning and policy design; however, it has some limitations, including the absence of explicit carbon cost/constraints. Future work will extend to multi-objective (cost–carbon–resilience), dynamic pricing/demand response, and broader behavioural modelling across building types and climates. • Developed a two-stage robust optimisation framework for Integrated Energy Systems (IES) that jointly addresses PV supply uncertainty, stochastic demand, and off-design equipment performance. • Generated community-scale stochastic load profiles using k-means clustering and Monte Carlo simulation, validated with MAPE < 5%. • Incorporating off-design characteristics increased annualised cost by 3.46% , while reducing PV uncertainty (Γ) from 14 to 0 lowered cost by 2.24% and investment by 5.94%. • Seasonal dispatch analysis confirmed adaptability: electric chillers dominate summer cooling, gas boilers dominate winter heating, and waste heat boilers dominate heating during transitional periods, with other units contributing as complementary sources. • Findings inform balanced robustness targets and policy frameworks for flexible, multi-energy systems in sustainable urban planning.