Research On Anomaly Detection Of Petroleum Drilling Process Based On Hidden Markov Model

Oil drilling is a very large project and the existence of many uncertainties in the process and these factors have strong concealment.The process of it is very prone to accidents which is a serious threat to the drilling platform safe operation.Therefore,during drilling construction in complex,if able to capture and use a variety of useful technical information to have a relatively accurate prewarning or give a certain degree of warning about the possibility of underground drilling accidents,are great practical significance to prevent and control the expanding of drilling accidents and minimize the economic loss.In this paper,the changes of various drilling parameters in the drilling engineering are analyzed comprehensively from the perspective of signal processing,and summarizes all the variation trend of drilling parameters related to the drilling accidents when the drilling accident occurs.The trend of the relevant drilling parameters is critical to the prewarning of the drilling accidents,moreover,the benchmark of the drilling parameters are changeable between different wells or even the same well with different depth.Then,the traditional non-stationary signal processing method is not suitable for the trend analysis of the drilling parameters.Accordingly,this paper constructs a kind of trend characteristic vector which can reflect the change trend of drilling parameters but not affected by the benchmark.Due to the HMM can be used to model and classify the signal on a time span,a kind of dynamic pattern recognition tool.This paper will construct a prewarning model of drilling parameters anomaly based on CHMM.Besides,the threshold of the output matching probability of the CHMM model is calculated by empirical analysis,the experimental results show that the CHMM anomaly prewarning model can predict the drilling parameters abnormally in time and effectively.It is easy to realize the fixed threshold method of selecting by experience or statistics,but it is not reasonable in essence.When selecting the threshold value is too big,it will result in a high false negative rate of prewarning,on the contrary,if the threshold value is too small,the false positive rate will increase.So the best way is to make the threshold value adaptively change the size to reduce the false negative rate and false positive rate of prewarning.By analyzing two commonly used adaptive threshold value methods and using them in the threshold value determination of CHMM model output matching probability anomaly,the experiment shows that the adaptive threshold value method can achieve better prewarning accuracy than the fixed threshold value.Its characteristics can be demonstrated by several abnormal changes in drilling parameters when a fault occurs during drilling.Therefore,it is necessary to make a comprehensive decision on several drilling parameters for drilling fault prewarning.