Study Of Methods Of CBM Productivity Numerical Simulation And Its Prediction

Coalbed methane(CBM) productivity is the key indicator for measuring CBM well’s performance, and the productivity directly affects the economic benefits. How to predict the productivity of CBM well accurately is key problem in efficient development of CBM reservoirs. The output mechanism of CBM is different from the traditional petroleum reservoir. The exploring how to establish a numerical simulation technique for the characteristics of CBM reservoir has important significance to the development of CBM. Therefore, developing effective CBM productivity numerical simulation model has an important practical value for the exploration and development of CBM.The conventional finite differential method(FDM) is hardly adapted to the influence of the complex structure of coal-rock mass, especially the structure bodies such as fault, etc. Finite volume method(FVM) has the advantage of adapting to complex boundary and unstructured grid. In addition, integral conservation of the method can be satisfied total calculation region or every control region. The result of computation is very accurate. It shows the accurate integral conservation in case of coarse grid. Second, FVM can avoid numerical shake and numerical disperse frequently happening in FDM. Third, integral grid’s partition is very flexible. So complicated boundary conditions could be treated conveniently. FVM has not only the accuracy of finite element method( FEM) but also the simplicity of FDM.Following a great number of experiments and theoretical studies, this thesis is finished main concerning CBM productivity numerical simulation and prediction method. The detail is following:(1) In this paper, the mechanisms of the reservoir, migration and production for CBM are studied and the geological model for CBM reservoir simulation is presented, From the theories of dynamic of fluids through porous media and CBM geology, a three-dimensional, two-phase, dual-porosity, pseudo-steady, non-equilibrium sorption, pseudo-steady CBM productivity numerical simulation model is established by numerical-simulation method.(2) Finite Volume Method(FVM) is used to solve the governing equations of the CBM migration. The formulations of cell-centered FVM are presented. The discretized nonlinear equations are handled fully implicitly. In addition, this complex mathematical model is solved with FLUENT software.(3) Qinshui basin as example, using the ejection data of gas and water, history matching of CBM productivity is calculated and the primary parameters which affect the production of CBM well are rectified.This model is applied to the prediction of productivity trend of the CBM well. After that, the sensitivity analysis of the primary parameters which affect the production of CBM well was done.The result is believable.(4) CBM productivity forecasting model based on chaotic time series and Least Square Support Vector Machine(LS-SVM) with Bayesian evidence framework is established by using the theories of phase space reconstruction theory and Bayesian evidence framework. The local minima problem is solved in support vector machine. It is faster and easier to use. As Support Vector Machine model not only takes a long time to determine the model parameters but also has "over- fit" phenomenon, the optimal parameters of the model can be found through three-layer Bayesian evidence inference. This method is proposed to select the input vector automatically. The adaptability can be enhanced with this method. Phase space reconstruction theory is used to study the chaotic time series properties of CBM production. Then after the reconstruction phase space of time series, Bayesian-LS-SVM forecasting model is proposed with a case study. Compared with SVM forecasting method and the BP network forecasting method, the results show that this method has quick computing speed, high fitting precision, flexible structure, strong generalization ability, etc. So, it is a new way to forecast CBM productivity.