论文部分内容阅读
针对油田注水站机组现有机械设备维护方式经济性差、不能完全避免事故发生的问题 ,提出用具有遗传功能的神经网络对油田注水站机组运行状态进行预测 ,给出了具有反馈功能的预测神经网络结构图 ,采用一种新型的全局随机优化搜索算法的遗传算法来训练前向神经网络。根据从大庆油田采回的现场数据建立的神经网络模型所做预测结果表明 ,遗传算法是一种启发式搜索 ,易收敛于全局最优 ;收敛速度方面遗传算法明显优于BP算法 ,预测精度明显优于常规预测方法 ,遗传算法能较好地反映机组运行状态的变化趋势。
In view of the poor economy of existing mechanical equipment maintenance methods for oilfield waterflooding units, the accidental failure can not be completely avoided. The neural network with genetic function is proposed to predict the operating status of oilfield waterflooding stations. The prediction neural network with feedback function is given. Structure diagram, using a new type of global random optimization search algorithm genetic algorithm to train the forward neural network. The prediction results based on the neural network model built from the field data retrieved from Daqing Oilfield show that the genetic algorithm is a heuristic search and is easy to converge to the global optimum. The genetic algorithm is obviously superior to the BP algorithm in the convergence rate, and the prediction accuracy is obvious Outperforming the conventional prediction methods, the genetic algorithm can better reflect the changing trend of the unit operating status.