Research On Drilling Well Leakage Diagnosis Model Based On Neural Network Fusion Technology

It is the important means for drilling engineering in the process of oil and gas exploration and development. During the drilling engineering operationing, it is always a threat to the whole drilling process because of the drilling accidents and downhole complicated conditions, and seriously affected the drilling speed, quality of well construction and exploration efficiency. In recent years, with the rapid development of drilling technology, it has become the main goal of improving the drilling speed and quality, reducing drilling accidents and the drilling cost to the pursuit and struggle for today’s oil industry.The leak is a common complications in well drilling engineering, serious leakage can cause the well borehole pressure drop, which affecting the normal drilling, resulting in wellbore instability, and thus inducing the formation fluids into the wellbore and finally blowout occurred. Recently, with the continuous development of the exploration and development in the western regions, well leakage occurs more frequently, and it is more hader to treat. Although there is a long-term deal with advances in drilling accident and technology complex situations, the corresponding difficulty also increased with the deepening of the complexity of geological exploration. A large number of statistical data showed that:almost every drilling a well will appear the lost circulation in varying degrees and different properties, so the drilling process will face many new problems, such as the storage is carbonates, it often encounters long segment leakage and pressure sensitive tape formation induced leakage in the process of drilling wells,etc.Therefore, it is of great importance for the exploration and development of oil and gas to increase the forecast accuracy of the risk of drilling wells leaking, to study the diagnostic methods and to take the right measures of plugging leakage of drilling wells. For problems encountered in the drilling process, this paper focus on the following research aspects:(1) a systematic study of the various drilling wells leak accidents, including the porosity dropout, dropout fractured porosity-fractured dropout and caves of leakage, as well as influencing factors, accident characteristics, classification and treatment methods for the entire diagnostic system laid the foundation;(2) integratly and comprehensivly analysis the factors affecting the leakage, and set up the well-drain type diagnostic parameter space and well leakage severity diagnostic parameter space.(3) established a diagnostic model of drilling wells and drilling wells drain leakage severity evaluation model. Firstly, the author established the several neural network diagnostic models of the leakage. Secondly,used the synthesis by means of the theory of evidence ruels in dealing with vagueness and uncertainty problem of the neural network output data fusion.And to get the final drilling wells leak diagnostic model. Finally,on the basis of the final diagnosis, established the drilling wells leak severity evaluation model by means of neural networks, and to evaluate the occurrence of lost circulation intervals on the well leakage. (4) According to the established drilling well leakage diagnosis model and drilling well leakage severity evaluation model. Making use of large amounts of data collected in the field of drilling to validate two models, and the results show that it is high to use of drilling well leakage diagnosis model and drilling well leakage severity evaluation model for diagnosis of lost circulation type and the accuracy of the appraisal well leakage severity, which has a strong practical. It can provide the strong technical support and guarantee for setting economic feasible program in advance.