With the wide application of smart electricity meters and power consumption management terminals, the relevant monitoring terminals in power grid stations can collect massive line loss data every day and identify the abnormal situations. However, the interference of data noise to line loss data in power grid stations leads to the decline of identification accuracy and recall rate. In order to solve these problems, an anomaly recognition method for line loss data in power grid area based on multidimensional features is proposed. In this method, the line loss data samples of power grid area are first formed into corresponding two-dimensional data and denoised by two-dimensional wavelet threshold method. Hasusdorff distance formula is improved according to the location and time characteristics of two-dimensional data after denoising, which is used to calculate the multi-dimensional characteristic similarity of line loss data in power grid area, and the similarity matrix between line loss data is obtained. Finally, multi-dimensional Hasusdorff distance is applied to hierarchical clustering algorithm to identify anomalies in line loss data of power grid area. Simulation results show that the proposed method has high accuracy and recall rate. The abnormal identification time of line loss data in power grid area is short, which meets the requirements of practical engineering.