When a fault occurs in the power grid, a large number of remote signaling alarms and amount of displacement information are uploaded to the dispatching terminal. This makes it difficult for dispatchers to make accurate judgments on the faulty equipment and the type of the fault in a short time. Therefore, this paper proposes a power grid fault diagnosis method that combines classification with fault space optimal coding and intelligent state estimation and error correction of remote signal data. It proposes mapping the remote signal data to the fault diagnosis space, and compares and classifies it with the optimal coding set of the fault space to realize fault diagnosis of the power grid. Through the remote signal displacement data of different failure modes, the intelligent state estimation model is trained to correct the misplaced or missed data of the remote signal to realize the front-end data error correction of the remote signal data, and thus improve the accuracy of fault diagnosis. Finally, an intelligent fault diagnosis method for power grid faults suitable for big data platform applications is formed. The validity of the intelligent state estimation model and fault diagnosis model for fault diagnosis of power grids is verified by case simulation verification and actual big data platform network operation. This work is supported by the National Natural Science Foundation of China (No. 51777119) and the Science and Technology Project of State Grid Shanghai Electric Power Company (No. 520900180030).