Abstract: The curtailment data in PV power generation is a special type of abnormal data. Traditional abnormal data recognition algorithms rely on the data distribution hypothesis or empirical model and cannot work well for recognizing this special type of abnormal data. Aiming to address with this problem, an abnormal data recognition algorithm based on the mathematical morphology denoising theory was proposed in this paper. The proposed abnormal data recognition algorithm took the curtailment data as the noise signal of the original data, so it did not have any requirements on the distribution characteristics of the original data. It only needed to transform the original data into a binary image, and then adaptively identify the curtailment data through the mathematical morphology denoising operations such as dilation and erosion. The simulation results show that compared with the traditional abnormal data recognition algorithms, the proposed algorithm has significantly improved the recognition rate of the curtailment data, which verifies the applicability of the proposed algorithm in the field of the curtailment data recognition.