Oil well production is an important basis for grasping oil field dynamic changes, and scientific and accurate prediction of oil well production is of great significance for planning oil and gas field development scale and production optimization. However, the traditional prediction methods have some problems such as long time consumption or low accuracy. In view of the above problems, firstly, the calculation mechanism model of oil well production is analyzed, and the main control factors of oil well production are determined by using grey correlation method. Secondly, the machine learning models of oil well production prediction based on KNN, SVM and BPNN are constructed. Finally, four ensemble learning models with higher prediction accuracy and stronger applicability are established based on the model fusion theory, and the nested cross verification method is used to test the whole data set. In order to provide more stable prediction results. The calculation results of an example show that the prediction errors of KNN, SVM and BPNN models are 15.73%, 14.9% and 13.96%, respectively. Compared with the single machine learning model, the performance of all ensemble learning models is superior. When the model is fused by BP neural network, the prediction accuracy of ensemble learning model is the highest, and the average relative error is 9.22%. It shows that ensemble learning can improve the performance of the model, predict oil well production more accurately, and provide technical support for on-site grasp of oil field dynamic changes and production optimization.