Unsupervised photovoltaic installation date detection method based on interpretable characteristics

Abstract

Abstract With the vigorous development of new energy, more and more distributed household photovoltaics are connected to the distribution network. However, the high proportion of household photovoltaics can have many impacts on the operation of the distribution network. The power department needs to register users who have installed photovoltaics. However, due to the large user base and the high resource consumption of existing investigative methods, this article proposes an unsupervised photovoltaic installation date detection method based on interpretable characteristics. Firstly, a set of interpretable characteristics is proposed and characteristic extraction is performed on user electricity data. Then, based on interpretable characteristics, user electricity consumption patterns are clustered. Finally, the date when the user’s electricity consumption patterns change is captured to determine the user’s photovoltaic installation date. The simulation results show that the proposed method can correctly identify all photovoltaic installation dates within 10 days, which is in line with practical applications and policy requirements. It has certain practical significance and theoretical value for the development of photovoltaic installation date detection applications.