有杆抽油系统故障诊断的人工神经网络方法

Application of BP neural network and self-organizing competitive neural network to fault diagnosis of suck rod pumping system

将人工神经网络用于有杆抽油系统故障的自动识别。对江苏油田的实测示功图数据进行了预处理,利用Matlab6.5进行编程,应用相同的数据对BP神经网络模型和自组织竞争神经网络模型的识别效率进行了对比。结果表明,由自组织竞争神经网络建立的模型对测试数据的正确识别率更高,识别效果稳定。因此,将自组织竞争神经网络应用于示功图的自动识别问题对实现有杆抽油系统故障诊断的自动化以及实现真正意义上的数字油田提供了一种有效途径。

On the basis of the artificial neural networks （ANN）, a self-organizing competitive neural network model was developed and used for automation recognition of dynamometer cards and fault diagnosis for suck rod pumping system. Compared with BP neural network model, the self-organizing competitive neural network model has good classification and generalization capability for recognition of dynamometer cards.