During the operation of distributed photovoltaic power generation, abnormal data collected by the operating system would be caused by interference factors, such as equipment failure, meter measurement error, user personal behavior and so on.Therefore troubleshooting abnormal data helps to ensure the accuracy and reliability of the PV generation user database and identify distributed PV users with operational problems.For the above reasons, an abnormal data inspection technology for photovoltaic power generation operation system is proposed.Environmental factors such as temperature, radiation, latitude, and seasonality are taken into consideration.The mathematical model of power generation is established by the central composite design method.The accumulated power generation is corrected by the extended Kalman filter(EKF) algorithm.Thus abnormal data can be checked and eliminated.The proposed method can realize the prediction of power generation data.It can check abnormal data quickly and reliably.The principle of investigation and the specific implementation process are discussed.Finally the effectiveness of the proposed technique is proved by experimental results.