The present disclosure is applicable to the technical field of maintenance of submarine cables, and provides a fault diagnosis method and apparatus for a submarine cable, and a device. The method includes: obtaining deployment data and sensed data of a target submarine cable; and inputting the deployment data and the sensed data into a trained fault diagnosis model to obtain a diagnosis result of the target submarine cable, where the fault diagnosis model is a probabilistic neural network model, and the diagnosis result includes a fault type. The present disclosure uses the trained probabilistic neural network model as the fault diagnosis model to diagnose a fault of the target submarine cable, and can quickly and accurately obtain the fault type of the target submarine cable.