Abstract

Canister desorption tests of cored shale samples help identify the presence of a gas resource. Gas is invariably lost during core retrieval and processing. This gas, called “lost gas,” is routinely estimated as part of the desorption test. Lost gas is added to the measured and residual gas to estimate the total gas in the shale. This provides an estimate of the relative gas resource potential.

Lost gas is typically estimated by the USBM method, a method developed for coalbed methane (CBM) sorption reservoirs. Lost gas comprises a minor fraction of the total gas in CBM and shallow gas shale core samples. However, the estimated gas lost from some deep gas-shale samples is 70% of the total gas or more, and total gas content far exceeds that reported in CBM reservoirs. The question naturally arises whether the lost gas estimate is correct, and whether the total gas measurement is accurate.

A dimensionless “model-fitting” approach was used as an alternate approach for estimating lost gas. The reference model to be fit is dimensionless fractional gas evolved from a spherical sample plotted against dimensionless normalized time. This method uses all the degassing data to estimate the lost gas, not just the slope of the first few points as in the USBM method.

The following steps are used to fit the data. First, a geometric factor converts normalized time from that of the actual core sample geometry to that of an equivalent sphere. An initial diffusivity is estimated from the slope of early degassing data. Data are fit by adjusting the equivalent lost time. Lost time is the model time during which gas is lost. As estimated lost time changes, the model automatically revises estimated diffusivity and lost gas, and recalculates normalized time and normalized fractional evolved gas. The operator iteratively modifies lost time estimate until the data best fits the model curve on the dimensionless plot. The operator then reads the estimate of lost gas and diffusivity from the spreadsheet.

Lost gas estimated here is typically less than that estimated by the USBM approach, but lost gas can still be quite high. In many tests, lost gas estimates by the two methods are similar.

To test the model, gas content estimated from the USBM method and that estimated from the model-fitting approach were compared to total gas estimated from TOC and porosity. Gas in core samples is stored as sorbed gas, free gas in porosity, and dissolved gas. Only sorbed and free gas are significant. Gas in pore space can be estimated from total porosity, water saturation, Bg and reservoir pressure. Sorption can be approximated from TOC and thermal maturity levels. Total shale gas contents estimated from the model fitting approach proposed here approximately matches gas content estimated from the gas-saturated porosity and TOC content of the core samples. In contrast, the USBM lost gas method irregularly overestimates the gas stored in porosity and by sorption.

Results also indicate that canister desorption tests should be supported by porosity and saturation measurements. Porosity, water saturation, and TOC should be routinely measured and gas content estimated from these measurements. The technology to measure porosity and saturation under restored reservoir stress is better than that of the desorption tests if a significant fraction of the gas is lost. Actual desorption patterns do not always follow desorption theory. The early part of many degassing curves is concave-upwards. This is inconsistent with any desorption theory that assumes a constant flow or diffusion parameter. This behavior indicates that flow parameter changes due to changing saturation or cleanup near the edge of the sample, perhaps similar to the capillary end effect. The value of desorption tests is that they validate the porosity and saturation estimates from core measurements.