基于SVM煤层气井井底流压预测方法研究

发布时间：2018-04-29 17:12

本文选题：煤层气 + 井底流压　；参考：《西安科技大学》2017年硕士论文

【摘要】：随着我国工业化发展水平越来越高以及进程的越来越快,国家对新能源的需求也越来越大。终端能源需求正在逐步从传统能源向优质高效洁净能源转化,作为新型环保能源的煤层气在我国的能源结构中占有极其重要的地位。然而,虽然我国煤层气资源丰富,但有超过一半的煤层气低产量井由于排采管理不善造成了产能束缚,所以急需新型的煤层气排采技术及装备来提高产能,实现合理的煤层气排采管理及产能的提高。影响排采效果的主要因素包括:非连续性排采因素、井底流压控制因素、排采强度因素等。本文在探讨了煤层气单井采气的基本原理以及工艺流程的基础上,深入分析了基于中国煤层气特殊资源条件和气井工程状态的井底流压预测方法,保证连续、平稳、逐级降低井底流压,控制排采强度,提高开发井产能释放成功率,增加气井产气年限,从根本上改变我国排采严重依赖个体经验的局面。针对煤层气单井采气系统中井底流压特征参数具有连续平稳、逐级降压的变化规律,本文研究了基于支持向量机的煤层气井井底流压预测模型,该方法可以有效地解决预测模型中支持向量机的参数寻优问题。针对支持向量机的参数选择不同对预测性能的影响也不同的特点,本文介绍了几种常见的用于支持向量机参数寻优的优化方法,分别为交叉验证法、网格搜索算法、遗传算法以及粒子群算法。通过方法对比,找到了适合本文对应煤层气采气系统样本数据的参数寻优最佳方法,而且通过可视化的编程使结果以图表的形式清晰的显示出来。在此基础上,分析研究了模糊信息粒化的方法,将信息粒化方法与支持向量机方法相结合,建立模糊粒化模型,获取目标的近似解范围。该模型采用最佳参数寻优算法优化模糊粒化模型参数,并采用误差评价模型的精确度。仿真结果表明,该方法具有较好的实用性,可以有效的预测井底流压的变化趋势,具有良好的预测和分析效果。
[Abstract]:With the development of industrialization in China, the demand for new energy is increasing. The terminal energy demand is gradually changing from traditional energy to high-quality and efficient clean energy. As a new type of environmental protection energy, coal bed methane (CBM) occupies an extremely important position in the energy structure of our country. However, although our country is rich in coal bed methane resources, more than half of the coal bed methane low production wells are constrained by production capacity due to poor production management. Therefore, it is urgently needed to develop new coal-bed methane drainage technology and equipment to improve production capacity. To achieve reasonable coal bed methane production management and productivity improvement. The main factors affecting the drainage effect include: discontinuous production discharge factor, bottom hole flow pressure control factor, drainage intensity factor and so on. On the basis of discussing the basic principle and technological process of single well gas recovery of coalbed methane, this paper deeply analyzes the prediction method of bottom hole flow pressure based on the special resource condition of coal bed methane in China and the engineering state of gas well, so as to ensure continuity and stability. The downhole flow pressure is reduced step by step, the production intensity is controlled, the success rate of productivity release of development well is increased, and the gas production life of gas well is increased, which fundamentally changes the situation that the production of production in our country depends heavily on individual experience. In view of the fact that the characteristic parameters of bottom-hole flow pressure in a single well gas recovery system of coalbed methane have the regularity of continuous steady and stepwise pressure reduction, a prediction model of bottom-hole flow pressure of coalbed methane wells based on support vector machine is studied in this paper. This method can effectively solve the parameter optimization problem of support vector machine in prediction model. Aiming at the different influence of parameter selection of support vector machine on prediction performance, this paper introduces several common optimization methods for parameter optimization of support vector machine, which are cross-validation method and grid search algorithm, respectively. Genetic algorithm and particle swarm optimization. Through the comparison of methods, we find the best method for optimizing the parameters corresponding to the sample data of CBM production system in this paper, and through visual programming, the results can be clearly displayed in the form of charts. On this basis, the method of fuzzy information granulation is analyzed and studied. Combining the information granulation method with the support vector machine method, the fuzzy granulation model is established, and the approximate solution range of the target is obtained. The optimal parameter optimization algorithm is used to optimize the parameters of the fuzzy granulation model and the accuracy of the model is evaluated by error evaluation. The simulation results show that this method has good practicability and can effectively predict the change trend of bottom hole flow pressure and has good prediction and analysis effect.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TE37;TP18

【参考文献】