Abstract: In order to timely detect and deal with abnormal oil temperature problems in substation transformers, a research on abnormal detection of transformer oil temperature in substations is proposed using multi-sensor data analysis. Select and deploy sensors to collect oil temperature data of substation transformers, and perform denoising processing to provide data support for subsequent detection. Using multi-sensor data analysis technology to obtain comprehensive dynamic information on oil temperature and estimate the trend of oil temperature changes. On this basis, an anomaly detection model is constructed to divide the oil temperature change rate into four operating intervals. The model receives real-time oil temperature estimation data as input, compares the real-time processed oil temperature data with the set threshold, and detects whether the oil temperature is abnormal. The experimental results show that after the application of this method, the transformer oil temperature can be accurately detected with an error within ±0.5 ℃, which is suitable for abnormal detection in substations.