Automated stochastic modelling for design-stage electrical power demand prediction in multi-dwelling residential buildings

Traditional methods for estimating electrical power demand (EPD) rely heavily on engineers’ past experience and established guidelines that prioritise continuous power availability. While effective for ensuring reliability, these approaches typically overestimate demand, resulting in oversized systems that operate inefficiently under partial loads and contribute to unnecessary greenhouse gas emissions. In the context of climate change targets, there is a pressing need for more accurate and sustainable methods. This study enhances an existing occupancy behaviour-driven stochastic modelling tool, originally validated for predicting a single day EPD of a standalone household, by adapting it to forecast demand for multiple households of varying sizes over extended periods. To broaden applicability, results are processed using Monte Carlo analysis, covering a wide range of scenarios. When applied to a 219 flat residential building, the proposed method predicted a maximum electrical power demand of approximately 206 kW, around 44% and 35% lower than the industrial partner and 2006 design-guidance estimates, respectively. The improved tool significantly simplifies practical use, making it suitable for design-stage EPD estimation in residential building types. The methodology promises cost savings, efficiency improvements, and reduced environmental impact.