This paper puts forward a new fault trace method for power grid based on big data platform. It extends the data source of fault diagnostic to the transformer substation and deals with the variety of fault date by the technology of Spark. This method can solve the problem of mass data processing. This paper analyzes the fault information by the data mining technology. The incorrect operation of protection or circuit breaker is traced-back by decision tree while the fault components are found. The reasons of the faults are given and the function of fault diagnosis system is optimized. Compared with the traditional fault diagnosis system which is based on the alarm messages

the proposed method can effectively use the monitoring data at every level in the substation and give the reasons of the faults.