Novel optimisation approach for community energy systems: Grid-connected capacity sizing with hydrogen storage and lifecycle

Abstract

The transition to high-renewable energy systems at the community level demands optimisation frameworks that balance economic efficiency, operational flexibility, and sustainability. While many existing studies focus on either static sizing or simplified dispatch heuristics, they often fail to co-optimise key system parameters such as inverter capacity, grid constraints, and hybrid storage integration under dynamic conditions. This thesis addresses that gap by developing a deeply integrated optimisation architecture that unites long-horizon sizing with short-horizon control, tailored for islanded and weak-grid energy communities. The central objective is to design a techno-economically robust and energy sustainable hybrid PV–battery–hydrogen system that minimises lifecycle cost and enhances renewable selfconsumption while accounting for real-world constraints. To this end, a nested optimisation approach is proposed, integrating a Genetic Algorithm (GA) for capacity sizing with a Mixed Integer Linear Programming (MILP) framework that embeds a Model Predictive Control (MPC) dispatch strategy. The GA generates candidate system layouts, each of which is validated via the MILP model that co-optimises hourly dispatch under fixed tariff structures and inverter-grid limits with AC/DC nodal representation. To capture operational uncertainty and improve flexibility, a rolling-horizon MPC layer executes every 12 hours using a 24-hour forecast window, incorporating flexible loads up to 8% of daily average demand, a level selected to reflect realistic load-shifting potential based on typical non-critical applications such as water pumps. Results show that the framework achieves Net Present Cost (NPC) and Levelised Cost of Energy (LCOE) reductions of 10% and 10.2%, respectively, compared to static or rule-based baselines. Grid-related operational charges fall by 46% under MPC with load flexibility, and self-consumption rises to 44.56%. A novel, extended Energy Return on Investment (EROI) metric is introduced to capture full energy pathways, revealing battery storage as the dominant contributor to lifecycle efficiency. To explore trade-offs between system size, energy return, and cost, generalisation heatmaps of EROI and NPC are developed around the optimised Formentera case study design from Chapter 4, which serves as the baseline (i.e. the configuration with the lowest NPC). These heatmaps identify design “sweet spots” around 1.0– 1.1× the baseline capacity, where high EROI (>5.0) and low NPC (≤€610,000) are simultaneously achieved. Beyond which oversizing leads to diminishing energy and cost returns due to increased curtailment and underutilisation of grid infrastructure. The proposed GA–MILP–MPC framework thus provides a replicable, scalable, and practical tool for optimising community-scale energy systems. By tightly linking planning, operation, and sustainability metrics, it enables planners to make data-driven decisions that are financially sound, operationally feasible, and environmentally justified. As distributed energy infrastructures continue to evolve, such integrative methods will be crucial for shaping resilient and sustainable energy futures.