Research And Imple Mentation Of Predictive Analysis Of Oilfield Anomalous Wells Based On Support Vector Machine

Abnormal wells are one of the major causes of oil field production fluctuations,and this issue has received more attention from professionals.Detecting abnormal wells in advance and taking corresponding repairs in time can reduce the loss of oil fields.The method of analyzing dynamometer cards is one of the traditional methods of abnormal well prediction.Professional technicians analyze dynamometer cards to determine whether the well is abnormal.However,this method relies on human experience,which is time-consuming and error-prone.At the same time,idle historical data cannot be fully utilized,limiting the value of its use.For the above problems,this paper is based on the data of the dynamometer cards,using the curve moment theory and machine learning algorithms on Hadoop to predict the abnormal wells.The predicting scheme is divided into the following parts: firstly,extract eigenvalues of the dynamometer card by the curve moment theory.Then,construct a model for predicting abnormal wells.The model adopts SVM multi-classification algorithm of decision tree method.And the decision tree is constructed by PSO clustering algorithm.Finally,describe the characteristics of the dynamometer cards by bag of words,which constructs a predictive model with greater generalization ability.The input data of the prediction model is eigenvalue data extracted by the curve moment theory and bag of words.The output data is an abnormal type of oil wells.In SVM algorithm,the parameters are optimized by the grid search method implemented by MapReduce,which can improve the speed of optimizat ion and the accuracy of prediction.And the accuracy of the prediction results can reach more than 90%.