The analysis of transformer oil temperature data is essential for monitoring the condition of their internal insulation and for evaluating the operational reliability of transformers, which is vital for effective power transformation. Nevertheless, the positioning of the oil temperature sensor within a complex environment renders it vulnerable to malfunctions, resulting in the production of inaccurate monitoring data. Such anomalies can profoundly affect the evaluation of the transformer's condition. Consequently, after carefully investigating different categories of abnormal sensor data, a detection methodology for transformer oil temperature sensor data utilizing PEDformer is introduced. The proposed model has shown proficiency in recognizing a diverse array of abnormalities within extensive oil temperature monitoring datasets. The high precision and efficacy have also been corroborated through extensive testing with various transformer oil temperature monitoring data.