Research On Intelligent Recognition Of Indicator Diagram Based On Artificial Intelligence Algorithm

Sucker rod plunger pumping wells play an important role in China's oil production.When the pumping wells fail,it will not only cause the oil exploitation not to proceed in an orderly manner,but also affect the progress goal,and even cause safety accidents when it is serious.Therefore,it is necessary to diagnose the fault of pumping wells accurately.It is the mainstream fault diagnosis method to judge the working condition of pumping wells by identifying the indicator diagram pattern of oil wells.In the traditional way,the patrol workers recognize the indicator diagram pattern according to their usual experience,but this method is inefficient and accurate,and can not meet the needs of modern production in oilfields.The traditional intelligent algorithm recognizes indicator diagrams mainly by manual pre-selection of geometric features of indicator diagrams,and then classifies and recognizes indicator diagrams according to their features.Artificially selected geometric features of indicator diagrams are often disturbed by human factors,resulting in inaccurate feature extraction and reduced classification accuracy.In order to solve the above problems,this paper proposes an intelligent identification method of indicator diagrams based on artificial intelligence algorithm.The main research contents and achievements are as follows:In this paper,the working principle of sucker rod plunger pumping unit and the formation principle of oil well indicator diagrams are studied in detail,and the corresponding relationship between different oil well indicator diagrams and pumping well operating conditions is analyzed.The theory of machine learning and deep learning in artificial intelligence algorithm is described in detail.The structures and algorithm principle of convolutional neural network(CNN)and stack sparse autoencoder neural network(SSAENN)in deep learning theory are described in detail.The collected indicator diagram data is preprocessed,and the identification models of CNN and SSANN networks are built respectively.The completed identification models are tested.The processed samples set of indicator diagrams are divided into training set and test set.The recognition model is trained by training set sample,and the performance of the recognition model is tested by the test set sample.CNN,SSAENN and Support Vector Machine are combined to form an improvedmodel.The experimental results are compared and analyzed with CNN and SSAENN.Finally,the deep learning model and its improved model are compared and analyzed with other shallow network recognition indicator diagrams.The comparison results show that the deep learning model and the improved model not only have the ability to automatically learn the characteristics of indicator diagrams,but also avoid the tedious need of manual pre-extracting the geometric features of indicator diagrams and improper selection of features in traditional shallow network,and have better recognition performance than traditional shallow network.Finally,the recognition method in this paper is applied to the intelligent fault diagnosis system of pumping wells developed by.Net development environment,and has been applied in Zhongyuan Oilfield and Yanchang Oilfield.The system solves the problem that the oil well staff can not effectively diagnose the faults of pumping wells for a long time,improves the efficiency of oilfield operation,further improves the oil production,and has a high application prospect.