摘要
有杆抽油系统悬点示功图的特征参数是合理选择地面机电设备的主要依据。由于井下工况的复杂性和部分参数难以确定,使得基于求解高维时变非线性方程的传统方法的计算结果存在偏差。本文将传统方法和神经网络相结合,给出了一种能比较精确地确定定向井有杆抽油系统悬点示功图特征参数的方法,避免了建立和求解复杂的非线性动力学方程。首先依据传统方法,计算出简化悬点示功图的特征参数,然后考虑抽油杆柱弹性振动、抽油杆与油管之间的库仑和粘性摩擦、气体和供液能力等因素的影响,利用BP神经网络和RBF神经网络建立了不同工况下悬点示功图特征参数的计算模型。利用现场实测数据对建立的神经网络进行了训练和测试。测试结果表明了本文方法的正确性和有效性。
Calculation of hanging-point dynamometer card characteristic parameters is important to the selection of mechanical and electrical equipment for sucker-rod pumping to obtain an accurate result by way of the traditional method of solving system. However, it is difficult high-dimensional and time-variable nonlinear equations,because of the unpredictability of down-hole working condition and the difficulty of determining some parameters such as friction force well. In this paper, a method combining the traditional and viscous damping parameters in directional means and the artificial neural network （AAN） is put forward to calculate the hanging-point dynamometer card characteristic parameters of sucker-rod pumping system in directional well, which avoids establishing and solving nonlinear equations. First, the simplified dynamometer card is calculated based on the traditional method. Then, by means of the BP and RBF artificial neural network, a model is established for calculating the hanging-point dynamometer card characteristic parameters for different working conditions, which takes into account the influences of elastic vibration of rod-string, coulomb friction and viscous damping between rod and tube, as well as gas effect. The network is trained and validated with the system＇s working condition monitoring data, which shows that the present method is correct and effective.
出处
《计算力学学报》
EI
CAS
CSCD
北大核心
2008年第4期557-562,共6页
Chinese Journal of Computational Mechanics
基金
国家自然科学基金(50575180)
陕西省科技攻关(2004K06-G23)资助项目
