Grid-Aware Sizing Optimisation Model for Low-Voltage Residential Microgrids

Residential photovoltaic (PV) and battery based microgrids are regarded as potential enablers for decarbonisation. Such systems are considered to be effective means for enhancing self-consumption and reducing reliance on the grid. However, low-voltage distribution networks can be congested especially during peak solar generation hours. During such hours not only generation quantities, but also the amount of power transferred between the microgrid and the utility network can be uncertain. This paper proposes a Monte Carlo process-integrated self-adaptive memetic algorithm and a comprehensive model for optimising the microgrid design with the objective of minimising costs for households while considering the uncertainty of solar generation and grid constraints. Energy demand is partly flexible, and can be shifted to different hours. The proposed model determines the optimal number of solar panels and battery size. A non-linear battery degradation based on depth-of-discharge is considered. Proposed algorithm combines a conventional genetic algorithm (GA) with a hill-climbing technique to improve individual selection by updating crossover and mutation probabilities via a self-adaptive process. Constraint violations are handled through an adaptive penalty parameter update. The effectiveness of the algorithm is demonstrated in a case study, where the model accuracy and computational capability are compared against a standard GA.