【作者】 杨玉平；

【导师】 赵永军；

【作者基本信息】 中国石油大学 ， 矿产普查与勘探， 2008， 硕士

【摘要】 煤层气资源量的大小、分布是煤层气地质评价的重要内容,同时也是煤层气开发前经济预算的主要依据,煤层气资源量计算的准确与否直接影响到煤层气开发的经济效益。研究区目前仅有少量的这类数据,而且计算方法也仅有常用的一些方法,利用人工神经网络这一新方法进行煤层气资源量预测不仅可以为煤层气的经济开采提供参考,而且解决了研究区地质状况复杂,线形预测模型无法解决的非线性问题。本文主要是以(Artificial Neural Network,以下简称ANN)误差反向传播(Error Back-Propagation,以下简称BP)神经网络为主,以自组织特征映射(Self-Organizing Feature Map,以下简称SOFM)神经网络为辅,研究这两种神经网络的算法和改进措施,以及模型的合理建立,进而利用Visual C++(以下简称VC)开发出了一套可视化的人工神经网络软件。本文以沁水盆地中南部煤层气资源量预测为实例,对研究区煤层气资源量影响因素和控制机理进行研究,提取出主要的可定量化的影响因子,本文中选择了如下的主要影响因素:煤炭的分布面积、厚度、容重和含气量。以这四个影响因素作为输入参数,以煤层气资源量作为输出参数,建立研究区人工神经网络煤层气资源量预测模型。利用所开发好的人工神经网络软件对研究区煤层气资源量进行预测,经过反复调试,得出了比较理想的预测结果。预测结果表明:研究区人工神经网络煤层气资源量预测模型建立的比较合理,人工神经网络软件操作简单、实用。根据上述预测成果,结合研究区煤层地质状况,对煤层气开发的有利区块进行了评价。评价结果显示:沁水盆地中南部煤层气资源量开发的有利区块主要位于阳城-长子、沁水、屯留-襄垣和安泽-沁源这四个区块。更多还原

【Abstract】 Coal-bed gas resources amount and distribution are two important parts of geological evaluation, and are also the bases for the budget before exploiting coal-bed gas resources. The precision of calculation of coal-bed gas resources amount directly impacts on the economic benefits of coal-bed gas development. There is only little data available in the studied area, and the methods of calculation are some used ways currently. Used of the new method of artificial neural network to predict coal-bed gas resources amount, it not only provides the reference for economic exploitation of coal-bed gas resources, but also solves the non-linear problem that the linear prediction model can not solve, due to the geological conditions being complex.The text,taking error-back propagation neural network as the main pole and self-organizing feature map neural network as the auxiliary pole, studied their fundamental principals, improvement of the algorithm and reasonable establishment of the model , then developed a visual artificial neural network software with visual c++.The text, taking prediction of coal-bed gas resources of central and southern Qinshui Basin for an example, studied the factors and their mechanism on coal-bed gas resources amount. The text selected the main counted factors as follows: coverage area of coal, thickness, density and gas content. Took these factors as input data and coal-bed gas resources amount as output data to build the model, and used the developed software to predict coal-bed gas resources amount of study areas. The predicted results indicated that the model was rational and the software had the characteristics of convenient operation and great practicality.According to the predicted results, the text assessed the favorable blocks taking into account the geological conditions of coal bed. The assessed result was that the favorable blocks were in Yangcheng-Changzi, Qinshui, Anze-Qinyuan and Tunliu-Xiangyuan.更多还原