Abstract: We propose a submersible oil plunger pump stroke frequency optimization method based on a particle swarm optimization-extreme learning machine (PSO-ELM) model to improve low pump efficiency, high energy consumption, and other defects. We establish the soft measurement model of a working dynamic liquid level by combining partical swarm optimization (PSO) arithmetic with extremc learning machine (ELM). According to the change of the working current of the working dynamic liquid level and submersible oil plunger pump, we solve the problem that the pumping frequency cannot be adjusted accurately during oil well production by establishing the objective function to obtain the pumping frequency that will result in the optimal economic performance of oil well production. Finally, we establish a model of fuzzy controller by employing the relationship of objective function to input parameter to adjust the pumping frequency of a submersible oil plunger pump. The experiment result shows that the established soft measurement model has high precision of moving liquid level, and the fuzzy controller can adjust the pumping frequency more reasonably to intelligently adjust the size of pumping, thereby improving oil recovery and saving energy.