Research And Application Of Intelligent Early Warning Method For Natural Gas Extraction Based On Isolation Mechanism

With the increase of the number of gas wells in oil and gas fields and the growth of the extraction years of gas wells,the frequency of gas well failures is also increasing,especially in the water producing gas wells of water producing gas fields,the frequency of abnormal gas well extraction has greatly increased,which has become a difficult problem to limit the production of natural gas.In order to solve the problem of frequent anomalies in the process of gas well extraction,gas production plants mainly rely on empirical rules for early warning,or experienced engineers to judge whether there are anomalies.These methods have the problems of low abnormal coverage,high false positive rate and poor portability.Therefore,how to timely and accurately detect anomalies is of great significance to ensure the production of natural gas.At present,the production of some oil and gas fields has been highly integrated with information construction,and a large number of production data have been collected,which makes it possible to realize data-driven artificial intelligence anomaly early warning.This paper deeply studies the intelligent early warning method in natural gas extraction and production.The main innovations and contributions include the following three aspects:1.In order to further improve the running speed of the anomaly detection algorithm,this paper first determines the key characteristics that determine the gas well pumping anomaly,and reduces the dimension of the data.Because the data distribution of static data in this paper is consistent with that of stream data,the feature importance calculated by XGBoost,the best supervised learning algorithm for anomaly early warning,can also be applied to stream data anomaly monitoring algorithm,which provides conditions for further improving the running speed and detection accuracy of i NNEASD algorithm.2.The existing anomaly early warning methods in the system not only need to input a large number of annotation data that are difficult to obtain in actual production,but also do not have the ability of self-adaptive update,resulting in the decrease of anomaly detection rate with the increase of running time in actual use.In this paper,an anomaly detection method of streaming data based on isolation mechanism,i NNEASD algorithm,is proposed.This algorithm is an unsupervised anomaly detection algorithm and does not need to use labeled data.Moreover,the i NNEASD algorithm uses the sliding window method,which can update the model according to the current window data to adapt to the new data.Experiments on actual production data show that compared with existing anomaly detection algorithms,i NNEASD algorithm has a higher anomaly detection rate.3.Design and implement the intelligent early warning module of natural gas production.Many gas production plants have deployed management systems based on the Internet of things.Adding an intelligent early warning subsystem to the existing management system has greater advantages than completely building an intelligent management system.Moreover,the experiment on the existing actual gas well production data shows that the proposed i NNEASD algorithm can effectively detect production anomalies in real time and meet the production requirements of gas production process because of its fast running speed and insensitive to parameter selection.To sum up,this paper uses the static algorithm with the best anomaly detection effect to determine the key characteristics of gas well pumping anomalies,and discusses the challenges and possible research directions in the field of intelligent early warning of oil and gas fields.Aiming at the shortcomings of current algorithms,this paper provides a new unsupervised stream data anomaly detection algorithm based on isolation mechanism,i NNEASD.The algorithm has the advantages of no need to label data,low parameter sensitivity,fast running speed and self updating ability.Moreover,the anomaly detection effect has obvious advantages in the current unsupervised anomaly detection algorithm based on stream data.Finally,i NNEASD algorithm is applied to the management system of gas production plant and has achieved certain results.