采油控制系统神经网络建模和优化算法设计  

Study on Neural Network Model and Optimal Algorithm for Oil Pumping Control System

作　　者：王彩云[1] 李英[1] 李元春[1] 

机构地区：[1]吉林大学通信工程学院,长春130025

出　　处：《吉林大学学报（信息科学版）》2006年第2期164-171,共8页Journal of Jilin University（Information Science Edition）

基　　金：吉林省科技计划发展基金资助项目(20040532)

摘　　要：油田抽油机普遍存在抽取能力大于油井实际负荷的问题,产生泵空或空捞现象,其结果增加了无效行程,浪费了大量电能,同时也增加了抽油设备的维护费用。针对这一问题,提出结合BP(Back Propagation)神经网络和遗传算法的采油控制系统。为了克服常规BP神经网络容易陷入局部极小和收敛速度慢的缺点,采用改进的非线性同伦BP神经网络进行采油模型辨识,并用遗传算法优化停机时间。在保证采油量的前提下,采油控制系统的节电率达25%以上,并可延长抽油机寿命30%左右,实现了抽油机采油的智能控制,经济效益十分显著。Oil pumping control has the problem that the designed capacity of pumping units is beyond the actual load of oil wells. Thus, pumping emptiness and under loading often happen in the process of oil pumping, which increases inefficient work. It consumes large amount of power and wears the machine down badly. So oil pumping control system based on neural network and genetic algorithm is p speed of conventional BP （ Back Propagation） neural network and overcome its drawbacks of getting stuck at local minima, improved nonlinear homotopy BP neural network is presented to identify oil-pumping model, and also optimize downtime through GA （ Genetic Algorithm）. The results of the computer simulation and spot experiment showed that the energy-saving rate ran up to 25 percent and the life span of oil pumping prolonged 30 percent with the precondition of oil output equality, eventually realized the intelligent control of the oil pumping and gained much economy profit.

关 键 词：采油控制系统 非线性同伦 改进的同伦 BP神经网络 遗传算法 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]