Photovoltaic Power-Stealing Identification Method Based on Similar-Day Clustering and QRLSTM Interval Prediction

In order to defraud state subsidies, some unscrupulous users use improper means steal photovoltaic (PV) power. This behavior brings potential safety hazards grid-connected operations. this paper, a power-stealing identification method based on similar-day clustering and interval prediction of the quantile regression model for long short-term memory neural network (QRLSTM) is proposed. First, data are clustered into three similar days by according weather conditions. Second, compared with (QRNN) method, good performance QRLSTM illustrated. Third, using intervals different confidence levels days, time scale (short-term, medium-term long-term) combined electricity-stealing judgment indicators, three-layer screening framework constructed, degree user power stealing qualitatively analyzed. Last, generation eight in certain region northwest China four groups artificially constructed used as an example simulation. The simulation results prove feasibility proposed paper.