Application of geostatistics to the construction of in situ coal quality models

Abstract

For a coal producer to remain competitive, coal has to be produced which satisfies the market specifications at the maximum possible coal recovery and minimum cost. To reach this goal requires an accurate knowledge of the in situ reserve. This knowledge includes a good estimation of both the in situ coal quality and the probable fluctuation of the in situ coal quality during production. Based on this knowledge, an optimum mine planning and scheduling strategy aimed at maximising the preparation plant recovery can be investigated. Geostatistical kriging provides one of the best approaches for estimation of in situ coal quality and geostatistical conditional simulation provides a promising way to characterize fluctuations in the in situ coal quality. The aim of this project is to investigate and develop a practical procedure for applying geostatistics to the in situ coal quality prediction and simulation using coal washability data. The term coal quahty used in this project refers only to yield and ash. The coal washability data can be expressed by two basic relationships: the instantaneous ash vs relative density relationship and the yield vs relative density relationship. For the coal seams or individual coal plies studied, the first relationship is relatively constant over a significant area. Consequently, it is only necessary to study the second relationship geostatistically. There are significant correlations between the incremental yields at different separating densities. These correlations of the coal washability data have been reproduced in the models by constructing them in a multivariate way. Another characteristic of the coal washability data is that the sum of the incremental yields and the associated sink is 100. A quadratic programming approach was employed to force the originally constructed multivariate models to have this characteristic. As a result, the procedures for constructing two coal quality models: a prediction model and a simulation model have been successfully formulated. A mine scheduling exercise has also been conducted based on the models developed. Geostatistical analysis was undertaken on a commercial software package. All additional computer programs needed for carrying out the procedures have been specially written. Consequently, the procedures developed in this project are readily generalized to other mines. One merit of the models developed is that they can be used as inputs to a coal preparation plant simulator. This allows mine planning and scheduling to be carried out by taking into account the imperfect separation of a coal washing plant and hence, the prediction of plant yields. The simulation model also allows the scale of the fluctuations in the plant feed for a given mine operating configuration to be predicted. Data from the Newlands Coal Pty. Ltd., a member of the M.I.M Holdings Limited Group of Companies have been used for procedure development. The resultant procedures have, in addition, been successfully applied to data from the German Creek mine of Capricorn Coal Management Pty Ltd.