Adaptive Real-Time Prediction for Oil Production Rate Considering Model Parameter Uncertainties

In the oil and gas production process, the online prediction of the oil-well production rate is an important task, that cannot only directly reflect the liquid supply capability of oil wells, but also guide the optimal control of the oil and gas production processes. However, traditional prediction methods have certain limitations in terms of accuracy and real-time properties. Therefore, to achieve an accurate prediction of the oil production rate, an adaptive integrated modeling method with a higher prediction accuracy and self-adaptability is proposed in this paper. With this method, a nonlinear mechanism model of the oil production rate is first established by analyzing the oil and gas production process and considering the nonlinear characteristics of the reservoir and multiphase flow in the wells. To reduce the influence of model parameter uncertainty and improve the prediction accuracy of the mechanism model, the least squares support vector machine (LS-SVM) method is then used to establish the error model for compensating the deviation in the mechanism model output. Moreover, to improve the adaptability of the model, an online correction strategy including a short-term correction of the LS-SVM and long-term correction of the mechanism model is proposed. Finally, through a simulation of the actual oil and gas production process in the oil production area, the results demonstrate that the proposed modeling method can not only improve the model prediction accuracy but also the model generalization, laying a solid foundation for the implementation of optimal control in the oil and gas production process.