Status evaluation method for arrays in large-scale photovoltaic power stations based on extreme learning machine and k-means
Ling Liang, Zhenqing Duan, Gengda Li, Honglu Zhu, Yucheng Shi, Qingru Cui, Baowei Chen, Wensen Hu
Abstract
Large-scale photovoltaic (PV) power generation has developed rapidly, and its installed capacity has reached 512 GW worldwide by the end of 2019. The status evaluation for arrays is an important guarantee of safe running of large-scale PV power stations. However, there exist the following problems in status monitoring: first, the lack of weather information hinders theoretical power calculations; and second, traditional methods focus on whole power stations other than arrays. To solve such problems, a status evaluation method for arrays is proposed. First, an extreme-learning-machine algorithm is used to calculate the output reference value of the targeted array. Then, we found that different indicators can effectively reflect the status of PV arrays. The performance assessment method was designed in conjunction with the k-means clustering algorithm. Finally, a case study was employed to evaluate the performance of different arrays in a 40-MW PV power station. The status assessment accuracy reaches approximately 90%, which confirms the effectiveness of the proposed method.