Peak load estimation based on consumer heating type classification powered by deep learning

The heating system of a building could significantly impact the aggregated electricity peak load. It is in the interest of the distribution system operators (DSOs) to both know what type of heating system that is connected to the grid, and the impact of end-users changing their heating system. Focusing on the transition from non-electric heating systems to heat pumps, this paper investigates its impact on peak load consumption. A state-of-the-art heating type classification method using smart meter data and deep learning was used to to first classify the heating types of single-family dwellings. Building upon previous work, a multi- label approach was adopted with the classifier to accommodate buildings with multiple heating sources. To assess the impact of heating system changes, smart meter data were substituted with data from similar buildings equipped with heat pumps. This process was repeated for statistical confidence. A geographical analyses identify areas susceptible to a large peak load increase, demonstrating practical application.