Abstract: We propose a failure diagnosis method for an electric submersible plunger pump on the basis of the mind evolutionary algorithm (MEA) and the back propagation (BP) neural network to solve the problem of high failure rate and short pump inspection period of the electric submersible plunger pump. This method can effectively diagnose failure accidents to prolong the pump inspection cycle. To solve the problem of minimal historical fault data and fault data that do not consider the electric submersible plunger pump in the actual production process, an experimental platform that can simulate the working condition of the electric submersible plunger pump is established. First, we simulate different failure states of the electric submersible plunger pump on the experimental platform, and the operating parameters are measured by using a multi-parameter acquisition module fixed at the bottom of the pump and wellhead instruments. Then, the most representative parameters are extracted from those relative operating parameters to structure the failure feature vectors and sample set. We use the sample set to train and validate the failure diagnosis model. Finally, the effectiveness of the fault diagnosis method is verified by the fault data set of the electric submersible plunger pump obtained from the actual production process. Experimental results show that this failure diagnosis method can diagnose the failure states of electric submersible plunger pump accurately and avoid failure accidents. Thus, this method can prolong the pump inspection cycle of the electric submersible plunger pump effectively.