Due to the low utilization of equipment information, the existing abnormal detection methods for power equipment are difficult to find potential equipment faults. Combined with big data analysis technology and equipment evaluation technology, a state data anomaly detection method based on time series and neural network is proposed. The auto-regressive time series model and self-organized mapping neural network are used to discretize the continuous power equipment data into a single sequence, and the transition probability of the state variable on the time axis is calculated, which can quickly detect data anomalies through state transition probability and clustering algorithms. The effectiveness of the proposed method is verified by experiments. The results show that this method can detect the abnormal state of power equipment quickly and effectively.