Statistical Model Fitting of Remote Induction Sounding Data from Underground Coal Gasification Site---Hanna II, Phases 2 And 3

Abstract

An Underground Coal Gasification Site near Hanna, WY, was sounded from the surface with a wide-band induction system to determine remotely the location, shape, and size of the coal burned. An estimated 6700 tons (148 888 ft3) of coal 275 ft deep had been converted to low-Btu gas. A wide-band loop-loop system employing psuedonoise/cross-correlation techniques was employed to effectively combat noise and obtain time signatures. Wire grid approximations to induction models were employed to efficiently compute model responses. The reaction zone was simulated by buried metal boxes, cylinders and spheres in a conducting overburden. Box and cylinder responses were shown to be insignificantly different at coal seam depths and spheres proved to be unlikely candidates for fitting field data. A dual parameter Bayes minimum mean-square-error estimator was employed to estimate model dimensions from magnitude responses extracted from field data at 1 kHz. Estimates of the volumes of coal gasified compared favorable with that obtained formerly by chemical estimation. However, the burn may have deviated to the side of the gasification wells more than was predicted by chemical instrumentation. Measurements taken repeatedly over a one-year span on one traverse show aging characteristics that demonstrate a migration of the conducting anomaly toward the surface.