Equipment failures

weather conditions and other factors can lead to a large amount of abnormal data in distributed photovoltaic(PV) power generation systems

causing serious effects on their safe and stable operation.In order to accurately identify and remove these abnormal data

a distributed PV power generation abnormal data identification method is proposed based on dynamic time warping(DTW) and two-stage quartile. Firstly

continuous abnormal data identification and elimination are achieved by comparing the mean photovoltaic power under similar irradiance. Abnormal data are eliminated based on the comparison of the mean photovoltaic power at the same period

taking into account the fluctuation of the photovoltaic power generation curve. A comprehensive curve similarity judgment method based on DTW and Euclidean distance is used to consider the fluctuation characteristics of the data more comprehensively

thereby improving the recognition and elimination effect of continuous abnormal data. Secondly

the DTW-Two-Stage Quartile abnormal data identification algorithm is proposed

and the first-order change rate and the second-order change rate are used to eliminate discrete abnormal data from the fused data

effectively identifying and eliminating discrete abnormal data. Finally

it is determined whether a fault has occurred based on the results of abnormal data identification and elimination. Experimental results show that

after the proposed algorithm eliminates abnormal data

it can better fit the distribution of normal photovoltaic power data.Compared with the quartile method and the 3-Sigma algorithm

the linear correlation degree of the proposed algorithm before and after the elimination of abnormal data has increased by 58.15% and 68.41% respectively

with better identification results.