The measured dynamometer card used to analyze working conditions of pumping units requires a large amount of manpower and material resources, and thus it is very necessary to develop a novel and simple method for the diagnosis of pumping conditions. In this paper, a pattern recognition algorithm is proposed for the classification of working conditions of oil wells based on electrical parameters. To this end, the data of electrical parameters are obtained by the acquisition system; according to the algorithm theories (logistic regression, support vector machine, principal components analysis) of pattern recognition in machine learning, the binary and multiple classifiers of condition diagnosis are used to realize the diagnosis of the working conditions of the oil wells based on electrical parameters. The improvement of the proposed algorithm has been verified by analytical and simulation results.