Effective solar prosumer identification using net smart meter data

Abstract

With the continued growth of distributed power generation, the number of customers who consume and produce energy via solar photovoltaic (PV) power generation (solar prosumers) has been increasing. Unfortunately, electric utilities are not aware of the location of all solar prosumers due to unauthorized or unreported installations and lack of separate PV metering in areas with Net Energy Metering policies. Knowledge of the location of solar prosumers can inform circuit protection and voltage regulation settings, and help grid operators improve situational awareness and better plan for daily variations in demand, which are further magnified by generation intermittency. This paper proposes an innovative approach to identify customers with PV power generation using net energy consumption data from smart meters. Methods of dimensionality reduction are evaluated to propose an effective means of reducing the amount of data needed to accurately identify solar prosumers to a single data point and improve the clustering time by a significant factor. The proposed advanced classification approach for solar prosumer identification, based on agglomerative clustering, demonstrates very significant classification accuracy and outperforms k-means and neural network based approaches on real consumption data from a large number of residential customers. As the number of solar prosumers is growing daily, a novel Solar Prosumer Identification-Duration Curve is also proposed to give operators and distribution planners insight into how quickly new solar prosumers can be accurately identified.