Research On Risk Early Warning And Recognition Method

In the process of oil exploration, drilling accidents and complex problems always exist objectively. According to the analysis of drilling data in recent years, in the drilling activities, to deal with complex situation and drilling time, accounts for about6to8%of the total time of construction, and the sticking accidents accounted for40%-50%of the whole drilling accident, which caused by sticking funding costs account for more than50%of non-production cost. In the national science and technology major projects (contract number:2011zx05021-006)"Drilling real-time monitoring and decision-making system", set up how to identify the risk of all kinds of underground and early warning, research task, and sticking risk is one of the most complex drilling accidents. In this paper, based on the real-time data to study early warning and recognize for sticking accidents.Artificial neural network algorithm is a kind of nonlinear, strong adaptive ability to learn data information processing method, its strong nonlinear approximation ability can truly express the nonlinear relation between input variable and output variable. For the parameter change rule of sticking accident, using artificial neural network for feature parameters based on sample training to establish relationship between accidents and data changes. On the basis of real-time data transmission, all parameters of the sticking accidents to guide change nonlinear maps the recognition of sticking accident.In this paper, according to the sticking data recorded, analyzed and summarized the categories and characteristics of sticking accidents. Studied the process of sticking accidents, and parameterized the whole process, identified the characteristic parameters of all sticking accidents in the whole process.Categorized down-hole drilling tool activity ways for normal drilling and tripping, which could distinguish all kinds of sticking the corresponding state with real-time calculation parameters such as tool rest time. According to sticking risks omen laws in each state, using the neural network method identify anomalies in down-hole, establishing sticking risks warning model with the real-time computation parameters. According to representation law of sticking accidents had already happened, using the neural network method to identify sticking drilling accidents happened, combining classification of risk, establishing sticking accidents classification recognition model.Having designed sticking risk warning and recognition software based on sticking risks warning model and sticking accidents recognition model. For those drilling risks has certain regularity, warning and identification can be performed, result and the actual match partly. Through examples further perfect, sticking accidents can be on the scene to provide early warning and identification of reference.