众所周知,页岩层是烃类的源岩层。页岩气是吸附于页岩微颗粒表面的天然气,其主要成分为甲烷气。评价页岩气地质资源量的方法是重量体积法。由于页岩的渗透率极低,常以纳达西（nano Darcy）级别表示（1 nD=10^-9 D=10^-6 mD）,而吸附气处于饱和的吸附状态。因此,页岩吸附气的开发,需要通过钻水平井多段压裂,降压解吸后才能投产。对于页岩气田,每口水平井都是一个独立封闭的开采单元,因此,需要先评价井控的可采储量,而后再评价页岩气田的可采储量。本文基于Baihly提出的归一化产量递减曲线位移法,利用陈元千近年提出的泛指数产量递减（GEPD）模型,对美国的Barnett、Woodfort和Haynesville三个大型页岩气田的井控可采储量和评价区的可采储量进行了有效的评价,并对页岩气井产量的递减率进行了理论上的分析,其结果完全符合实际页岩气井的产量递减特征。这一特征就是,气井初期的产量递减得快,在生产20个月后的产量递减逐渐变慢。同时,笔者对现有的五种递减模型进行了简评。

It is generally known that shale layers are the source rock layers of hydrocarbons. The shale gas is a kind of methane gas which is adsorbed on the surface of shale particles. The weight-volumetric method is used to estimate the geological resources of shale gas. Due to the extremely low permeability of the shale, it is represented by the nano Darcy levels （1 nD=10^-9 D=10^-6 mD）, and the adsorbed gas is in a saturated adsorption state. Therefore, the development of shale adsorbed gas requires horizontal well drilling, multi-stage fracturing, depressurization and desorption before putting into production. Since the horizontal well is an independent closed production unit for the shale gas fields, it is necessary to estimate the well-controlled recoverable reserves firstly, and then to estimate the recoverable reserves of the shale gas field. Based on the normalized production decline curve shifting method proposed by Baihly, the well-controlled recoverable reserves and the reserves of the estimation areas from the three large shale gas fields （Barnett, Woodfort, and Haynesville） in the United States have been effectively predicted using the generic exponential production decline（GEPD） model proposed by Chen Yuanqian in recent years. Furthermore, the production decline rate of shale gas wells is theoretically analyzed, and the results completely accord with the actual production decline characteristics of shale gas wells. It shows that the production of a gas well decreases rapidly in the initial stage, and gradually slows down after another 20 months of production. At the same time, five kinds of decline models are studied and estimated in this paper.