According to the distinct temperature difference characteristic between solar cells under different operation states, a new analysis and recognizing approach for the photovoltaic array operation states was proposed based on the infrared image analysis in the paper. At first, the infrared images of photovoltaic arrays are pre-processed and analyzed; the abnormal areas and their features are abstracted. To deal with influences of some factors on the temperature of photovoltaic, such as environment temperature, wind speed and solar illumination, a fuzzy reasoning technique based on the data fusion is employed and the fault areas are recognized automatically. The research results show that the normal, shadowed and aging destroyed operation states of photovoltaic array in power system can be recognized accurately with the proposed approach.