Research On Real-time Early Warning Method Of Drilling Overflow

Kick is one of the most common risks in the process of petroleum drilling. Without timely monitoring and warning kick, which may lead to serious well control accident, even blow out, causing heavy loss of lives and property. Therefore, the early detection and warning of kick has been a hot research topic. At present, the early detection and warning of kick mainly depends on artificial judgement analysis by field engineers at home, which is subject and uncertain, because the judgement accuracy depends on the experience of field engineers to a large extent. For all this, based on the analysis of the insufficient of current kick warning methods, this paper presented one early kick warning and kick forecasting method through analyzing real-time data by artificial intelligent analysis method.This paper built one kick risk fault tree analysis model and analyzed the kick reasons and kick risk factors. And the change rules of symptom parameters before kick and characterization parameters after kick were summarized based on real-time comprehensive logging data. Then divided the kick warning into two stages, the first warning stage was based on symptom parameters, and the second stage was based on characterization parameters to forecast the severity of kick.As the artificial neural network has a strong ability to build nonlinear approximation of nonlinear relations and its strong self-learning ability, the real-time kick warning based on symptom parameters was built by BP neural network. This paper proposed the selection principal of kick sample set, analyzed the form of input layer parameters, and optimized the number of hidden layer node.According to the variation of characterization parameters after kick, a time series forecasting model based on characterization parameters was established. The kick volume before well shut-in and pressure after well shut-in were calculated by time-domain analysis model, and the kick severity degree analysis chart was built to analysis the kick severity by the variation of kick volume or shut-in pressure.Based on the calculation model built in this paper, the drilling kick real-time warning software was developed by using VS 2010 and SQL Server 2008. And the kick risk factors were real-time calculated by this software, as well as the kick warning model and the kick severity model.The simulation calculation results show that the BP neural network warning model could warning the kick before it happened, and the time-domain forecast model could forecast the kick development tendency and then judged the kick severity degree.