Aiming at the problem of identifying the characteristics of ash accumulation and shadow of photovoltaic modules, the difference of photovoltaic characteristic curves of ash accumulation and shadow was analyzed in detail, and the time-varying characteristics of the inflection point of the shadow photovoltaic curve were revealed. The number of inflection points of the characteristic curve and the current and voltage characteristic conditions were proposed to form the input feature quantity of the training model, and the dust accumulation and shadow recognition model was trained based on the CatBoost algorithm. Finally, the performance analysis and comparative test of the recognition model trained by CatBoost algorithm, ID3 and GA-BP algorithm were carried out by using the measured data of photovoltaic modules, and the results show that the recognition model trained based on CatBoost has strong discrimination and high diagnostic accuracy, which is of great engineering application value.