The Application Of RBF Neural Network In Under-balanced Drilling

The under-balanced drilling technology is emerged as required under the following conditions: the work in exploratory development is harder and harder, the international petroleum market has become more and more difficult since the competition grew, it is a kind of new technology which has been risen again since 1990s in the international markets. Utilizing this technology could resolve many troubles and complex issues in exploratory development, increasing output and decreasing cost. In 21st century, exploratory development of hydrocarbon resources in our country is facing the following: development of hydrocarbon resources under complex reservoir property and complex geological conditions, developing low pressure, low permeable and low output hydrocarbon resources, development of rebuilding and top potential against middle and later stages of oil-gas field, development of the unconventional hydrocarbon resources such as thickened oil, tight gas reservoir and coal bed methane, and challenge coming from the world oil markets. therefore, the under-balanced drilling technology is a necessary and is one of the main technologies In 21st century's development of hydrocarbon resources in our country. The under-balanced drilling technology in development and application in our country is a system engineering with involving wide range, the high investment and risk. The under-balanced drilling technology has become the another development direction and hotspot after technology of horizontal wells. Doing the work of bottom pressure forecasting well could provide important pressure technical data for drilling parameter design and structural design in well bore, and it can play a active role for research of effectively utilizing the under-balanced drilling in the wells developed and increasing oil and gas output.This thesis is focused a characteristics of the stratum and the geologic reservoir in Daqing Field, and studying application of the under-balanced drilling technology in Daqing Field. Analysing the factors may effect bottom pressure, introducing the recent achievements of the domestic and overseas controlling technology in bottom pressure, summarizing the status of knowledge on forecasting technique of the under-balanced drilling in bottom pressure. Passing through analysing the current lack in this research field, including the application limitations, based on the above a research direction is raised for resolving bottom pressure forecasting. This thesis is divided into the following parts:1. Analyzing the computation module for the under-balanced drilling in bottom pressure to find a feasible method for decreasing and eliminating error. In this thesis the semi-empirical method is adopted, based on the early adopters'research, and considered the influence of bottom liquid column hydrostatic pressure and circle friction drag by rock-fragment solid phase, and analysing acceleration of gas-liquid flow in pressure drop, therefore this module comes nearer to the actual situation of under-balanced drilling, and can improve the bottom pressure module in calculation accuracy.2. Seeking the applicable technology which can resolve the troubles, i.e. artificial neural network technology with development of under-balanced drilling technology, the requirement to bottom under-balanced value in accuracy control is more and more slashing. Results indicate that, when the pressure gauge with drilling measures the actual pressure at the bottom, for the original gas-liquid biphasic well, its computation module in under-balanced control in error is larger and can reach 13%. Therefore, when air cutting is happened on a drilling process set up by H. V. Nickens, forecasting model of RBF neural network in bottom pressure can be set up based on the module of two phase flow. In this module the influence of rock-fragment solid phase and multi-phase acceleration in pressure drop has been considered well, and having higher accuracy, utilizing the actual data measured from deep-level under-balanced drilling. Results indicate that, the error is no more than 3%, therefore, it provides a theoretical base for under-balanced design and calculation, and accurately controlling bottom under-balanced value.3. Through the method of combining theoretical analysis and simulation experiment, designs reasonable artificial neural network model, and gives optimization to the algorithm, with which to establish the predicting system. Factors which have great influence on bottom pressure of under-balanced drilling as well as obtain data information on the spot are taken as input layer Neuron. Under the condition of air cutting, selects typical sample as neural network study training input sample. Theoretically makes the comparison between BP network and RBF network, and separately carries out simulation comparison on the performance of BP network and the RBF network with data. Also gives test on network model with three wells of Dashen 2, Xushen 5, Xushen 7 and so on, with this foundation, analyses and evaluates the test result with high precision so as to satisfy the design requirements of under-balanced drilling. Besides, the software system has been developed, and has been successively carry out under-balanced drilling field test for six wells such as Xushen 5 and so on in the east Changyuan in-depth in north part of Songliao basin, which has made ideal effect.In the entire process of the under-balanced drilling, this system is able to analyze the bottom under-balanced value, judge the tendency of under-balanced drilling, provide scientific foundation to give on-site adjust to under-balanced drilling technology plan promptly and accurately according to the instrument real-time gathering data and the ground monitor parameter. In addition, analysis can be implemented on under-balanced the drilling process and afterward, including the relationship between well depth and bottom under-balanced value, and the calculation of annular pressure distribution rule. To see from the on-site usage situation, this system has realized the real-time prediction of bottom pressure in the under-balanced drilling process, which can satisfy the on-site request of analysis and decision-making of under-balanced drilling, effectively guides the under-balanced drilling construction, as well as accumulates vital information for deep research into under-balanced drilling.Through the application of RBF neural network in the under-balanced drilling, research and develop the exploratory well under-balanced drilling prediction system for bottom pressure, which not only can give real-time analysis, predicting the under-balanced state under the well, promptly adjust under voltage value of the well, but also realize the preliminary interpretation and evaluation to a reservoir during the drilling, and it can inquiry about the bottom under-balanced status at a certain and early time or a certain depth in any time, improving the site deciding ability and level in under-balanced drilling. This system has higher precision of prediction, in the deep level of the long wall eastern part, the north part of Songliao basin the six wells including Xushen 5 have been perform the site under-balanced drilling test one after another, and results show they are effective, and provide more efficient technology and safety control in the next under-balanced drilling. It would have a important significance for oil-gas increasing output in Daqing Field, and effectively developing exterior fields, and it will provide strongly technical support to Daqing Field in sustainable development.