A K-NN Clustering Based Method to Generate PV Power Series for Power System Analysis under Typhoon

With the increasing proportion of renewable energy, today's power system become highly sensitive to the weather condition, especially the extreme meteorological events. In order to evaluate the reliability of the power system under extreme meteorological conditions, it is necessary to accurately simulate the power curves of wind farms and PV stations. In this paper, a K-Nearest Neighbors (KNN) clustering based scheme is proposed to generate the multiday power curve of PV stations during typhoon. A two-layer modeling scheme is designed to set up the weather-mode related PV daily curve libraries and the typhoon-related multiple-day curve-mode code libraries according to historical data analysis. In application of the model, an inverse procedure can be carried out to generate the multi-day PV curves of different PV stations under any specified typhoon with retaining of the randomness and diversity. Historical data from the Guangdong power system is applied to verify the model. Results show that the multiday PV power sequences generated by the proposed method well reflect the statistical and time-domain characteristics of the PV stations during the typhoon event.