A Data-Driven Approach for Detection and Estimation of Residential PV Installations

The number of photovoltaic (PV) systems in the electric grid is growing at an unprecedented speed. This is rapidly transforming the ways in which the traditional distribution grid is being planned and operated. A problem faced by utilities is that, in many cases, the PV system installed does not correspond to the size or type filed with the installation permit, or simply the installation took place without a permit. In order to maintain grid reliability and safety, utilities must be able to detect and monitor all PV installations in their network. This paper proposes a data-driven approach for the detection, verification, and estimation of residential PV system installations. We use a change-point detection algorithm to screen out abnormal energy consumption behaviors including unauthorized PV installations. Then the existence of the unauthorized PV installation is further verified through a statistical inference known as permutation test with Spearman's rank coefficient. The proposed hypothesis test takes the customer's load profiles before and after the detected change-point as inputs, which are estimated through Gaussian kernel density method. Finally, the local cloud cover index is integrated with smart meter measurements to estimate the size of the PV system. The proposed method has been tested and validated with actual smart meter measurements under several scenarios.