Load profiles reflect electric energy consumption of consumers, including the information of day-to-day operations and system reliability. However, some random factors such as channel errors, unexpected interruption or shutdown of power stations can result in load profiles contain outliers and missing values. In this paper, a data preprocessing model based on fuzzy clustering and grey relational analysis is proposed. Firstly, the similar sample set with larger correlation degree is determined by grey correlation analysis. Then the typical load profiles are obtained by adopting fuzzy clustering algorithm and clustering validity index. Finally, the correction is performed on the abnormal data of identification. The proposed model is applied to a city grid SCADA system, which proves the model has high accuracy and practicability.