中文摘要：

通过智能采油系统自主分析与决策获取油田机采过程最佳决策参数,对解决机采系统效率低、能耗大等问题具有重要意义。受机械、地层、人为等不确定因素影响,智能采油系统难以构建生产参数、环境变量与系统性能、设定生产方式之间的机理关系并优化决策。为此,提出基于动态演化建模的偏好多目标优化方法,以实现采油系统的自主决策。利用无迹卡尔曼滤波神经网络（Unscented Kalman filter neural network,UKFNN）挖掘机采系统潜在规律,建立其动态模型;构建产液量偏好多目标优化目标函数,并利用非改进支配排序遗传算法（Non-dominated sorting genetic algorithm 2,NSGA2）获取相应的最佳决策参数。某油田试验结果表明：该方法使得系统日耗电量降低15.87%,系统效率提高4.9%。可见,所提方法可行且有效。

英文摘要：

Obtaining the optimal decision parameters by intelligent production systems＇ autonomous analysis and decision has significant meanings to deal with the low efficiency and high energy consumption in the oil extraction process. However, it is quite difficult to conduct and optimize the mechanism relationships among the operation parameters, the environment variables and the production mode settings, due to the mechanical, geological and artificial factors. Therefore, a novel autonomous decision method of oil extraction system by preference driven multi-objective optimization based on dynamic evolution models is proposed. The potential law of the pumping systems and then establish the dynamic model by unscented Kalman filter neural network（UKFNN） is found. The preference multi-objectives are constructed according to the actual production mode. The optimal decision parameters are obtained by improved non-dominated sorting genetic algorithm（NSGA2）. The experimental results show that after the proposed optimization the energy consumptions of the system decrease 15.87%, as well as the system efficiency improves over 4.9%, which illustrate the feasibility and the effectiveness of the proposed method.