Off-grid PV systems modelling and optimisation for rural communities: leveraging understandability and interpretability of modelling tools

Rural off-grid solar photovoltaic systems require careful planning to address key uncertainties, including variations in user behaviour, possible climate change impacts, and differences between software simulation and optimisation methods. Recent literature underscores the importance of understandability, examining the workflow from scenario creation to system design, and interpretability, selecting data-driven techniques that are intelligible to humans and grounded in physics-informed assumptions. Recent advances in open-source platforms offer highly adaptable and transparent alternatives for rural electrification studies and represent an alternative to proprietary software tools. This study presents a novel framework that integrates a Particle Swarm Optimisation algorithm with open-source energy demand modelling tools to size off-grid PV plus battery systems in a traditional rural building. This integrated approach emphasizes understandability by explicitly describing objective functions, constraints, and computational techniques, organised as logical blocks. Within these blocks, the incorporation of physics-informed, data-driven approaches enhances interpretability. Validation through comparisons with well-established and validated open-source software confirms the credibility of the approach. Findings indicate that incorporating multi-timescale scenarios, reflecting climate change trajectories, evolving user needs and loads (e.g., electrified heating and cooling) can substantially improve confidence in off-grid solar system solutions. The results also underscore that transparent, open-source-based models can reduce costs, increase flexibility, and simplify adaptation to evolving rural energy needs. Overall, this work highlights the broader potential modelling workflow leveraging understandability and interpretability principles in diverse contexts, including rural buildings and communities. Future research may extend the modelling framework proposed by improving the formulation of the underlying surrogate modelling methods to improve scalability and computational performance for scenarios involving multiple buildings or communities.