Maximal Information Coefficient Based Residential Photovoltaic Power Generation Disaggregation

Abstract

Due to policy support, low cost and easy applicability, distribution photovoltaic systems (DPVSs) are increasingly popular among residential community. However, small-scale DPVSs of less than 10 kWp are always installed behind the meter (BTM), which results in the invisible of the photovoltaic (PV) power generation. Only access of composite power data can result in non-optimal distribution network control and optimization, leading to a series of energy management problems. In order to solve the aforementioned problems, this paper proposes a BTM composite power disaggregation method focusing on small-scale DPVSs, with only composite power data of residential users in a community, without relying on weather data and models assumption. Considering that community users’ DPVSs usually exhibit approximate output characteristics, neighboring composite power is used to extract PV power generation information as mutual proxies. After obtaining approximate PV proxy data by subtracting composite power of inter-users, a grid search algorithm guided by Maximal Information Coefficient (MIC) is performed to obtain final PV power generation disaggregation results. The proposed method is evaluated using data gathered from residential customers located in Ithaca, New York and Austin, Texas in real-life scenarios. Testing results show that our proposed method achieve considerable disaggregation accuracy in the absence of solar radiation and temperature data as compared to other state-of-art methods.