基于SVM的注水机组状态预示技术研究

Techniques for predicting conditions of water injection machine train based on support vector machine (SVM)

油田大型注水机组在连续运转过程中,由于其自身的因素以及受外界条件的干扰,其运行常处于非线性非平稳状态。在充分研究和比较多种设备状态预示方法的基础上,提出一种基于支持向量机(Support Vector Machine,SVM)的状态预测新方法。该方法应用最终预报误差(FinalPrediction Error,FPE)准则确定样本的嵌入维数。通过比较SVM预测模型与自回归预测模型的单步和多步预测结果,证明基于SVM的预测方法在较长区间内具有良好的预测效果。用SVM预测大庆油田旋转注水机组时域的振动烈度,取得了较好预测效果,证明该算法能有效提高预测精度。

In the continuous running process of large water injection machine train in oilfields, operation of the machine train is often in nonlinear and non-steady conditions, because of the interference caused by the factors of itself and external conditions. On the basis of studying and contrasting the methods for predicting equipment status, a new method for status prediction is proposed based on SVM. In the method, the criterion of final prediction error (FPE) is employed to determine the embedding dimension of samples. Through the comparison of single-step prediction result with multi-step result of SVM predicting model and auto-regressive model, it is demonstrated that the method based on SVM prediction method has better predicting result in longer sections. SVM model is used to predict the time domain of rotating train, where shake is intensive, and better result is achieved. It indicates that the method can be used to effectively improve the prediction precision.