The invention discloses an elevator fault prediction method based on a BP neural network. The method comprises the following steps that 1, real-time data of the movement of an elevator are collected through a sensor set installed on an elevator car; 2, the collected historical data and standard data of the movement of the elevator are preprocessed and characteristic parameters are extracted, wherein a part of the characteristic parameters serves as a data sample, and the other part serves as a test sample; 3, a bp neural network diagnosis model is built, and the collected data sample is inputfor training; and 4, then the test sample is input to the trained bp neural network, the recognition accuracy of the trained sample and the test sample is determined, a training algorithm is optimized, the optimized neural network parameter configuration is used, and the elevator is subjected to fault detection. The elevator fault diagnosis method has the advantages of being high in real-time performance and diagnosis precision, capable of judging whether the elevator has potential safety hazards or not in real time, reducing the cost for manually maintaining the safety of the elevator, and finally achieving the balance of the safety performance and the economic benefit.