In order to improve the positioning ability and fault tolerance of clustering algorithm in fault area identification, this paper proposes a grid fault area identification algorithm based on relay protection information and spectral clustering. The method firstly uses the relay protection state value provided by each Intelligent Electronic Device (IED) to form a feature vector, and then analyzes the spectral clustering result according to the previously delineated association between the component and the IED, and finally obtains the faulty component. In terms of fault tolerance, this paper analyzes the influence of eigenvector distortion on clustering analysis and finds that using density-adjusted spectral clustering can achieve better results. The simulation experiments show that the density difference of the kernel is improved, and the fault location algorithm based on the clustering principle has higher positioning ability and fault tolerance, and the positioning capability is improved. This work is supported by Shanghai Sailing Program (No. 16YF1404800) and the Three Gorges University Cascade Hydropower Station Operation and Control Hubei Provincial Key Laboratory Fund Project (No. 2015KJ11).