||Due to the multi-condition in oilfield production process, it is difficult to accurately describe the relationship among production parameters, environment variables and system performance. Therefore, this paper proposed that the PSO-ELM is utilized to contract the multi-condition model of beam pumping units to effectively solve the faced problems. Firstly, the indicator diagram is divided into several typical working conditions by K-means clustering method. Secondly, a principal component analysis is adopted to reduce the dimension of data in each condition. Finally, particle swarm optimization is adopted to optimize the key parameters of extreme learning machine to establish high precision model corresponding to conditions. The proposed method is applied to one Oilfield Company in China. And the results show that condition-specific model of beam pumping units can accurately reflect the operating characteristics of working condition and obtain high precision model. In addition, the comparative experiments with ELM show that the PSO-ELM model has stronger generalization ability. Thus, the proposed method could provide theoretical guidance to build high precision model of beam pumping units and lay foundation for the research of the consumption reduction.