Research On Fault Diagnosis Method Of Pumping Wells Based On Hidden Markov Model

At present,the way of sucker rod pumping is widely used in oil fields,but most of the equipments such as sucker rod pumps are in the field,the location is dispersed,and the working conditions in the down-hole are very complex,which will cause oil production failures to happen frequently.The traditional method of checking the working conditions of the oil well by manual inspection can't find the faults in time and take measures,which will cause inestimable loss.The indicator diagrams can reflect the working states of sucker rod pumping well,and they have different graphical representations under various working conditions,which can effectively reflect the failure of pumping wells.In this article,selecting indicator diagrams as the research object,through the analysis of the indicator diagrams to classify the faults of the pumping well.Firstly,analyzed the production principle of rod pumping wells,through the introduction of the working process of the oil pump,the forming process of the indicator diagrams is discussed.The characteristics of several typical working conditions of indicator diagrams are introduced,which lays a theoretical foundation for the subsequent establishment of sample database for typical working conditions.Secondly,the feature extraction methods of indicator diagrams are studied.In this article,the Freeman chain code is selected to represent the outline of the indicator diagrams,and the chain code histogram is used to extract the features of the indicator diagrams.For the case that the indicator diagrams are different but the chain code histogram is the same,the spatial distribution entropy of chain code histogram is introduced.The method of chain code histogram and spatial distribution entropy of chain code histogram are combined to extract features of the indicator diagrams,this method not only considers the statistical characteristics of chain code but also considers the distribution characteristics of chain code space,improves the accuracy of expressing indicator diagrams.Finally,the purpose of fault diagnosis for pumping wells is achieved through classification and identification of indicator diagrams.In this article,the Hidden Markov Model is used to classify the indicator diagrams,and the feasibility of the method is verified by the simulation experiment.Aiming at the problem that the parameter training process in Hidden Markov Model tends to fall into the local optimal solution,the improved harmony search algorithm is used to optimize the initial model of Hidden Markov Model,and then enter the Baum-Welch algorithm into the optimal model.The simulation results show that compared with the traditional Hidden Markov Model,the model trained by the optimized algorithm accelerates the convergence speed,and it is easy to jump out of the local optimal solution,and the recognition accuracy is relatively high.