A Data-Driven SVM-Based Method for Detection and Capacity Estimation of BTM PV Systems

In the past years, an increasing number of residential customers have installed rooftop solar photovoltaics (PVs). Most of these PVs are located behind-the-meter (BTM), i.e., invisible to utilities, which poses challenges to system operation and planning. Therefore, a methodology to detect and estimate the installed capacity of customer-level BTM PVs utilizing only net power curves and weather data is proposed in this paper. The methodology is based on the grouping of PV generation and pairing of native demand that feed algorithms based on support vector machine (SVM) built to detect and estimate the installed capacity of BTM PVs. The results using two datasets (real and synthetical) show that false positive and false negative rates in detection are limited to 9.09%. In the estimation method, the root mean square error (RMSE) is lower than the rated power of a single PV panel, ensuring the precision of the developed method.