Forecasting fluctuations in coal quality delivered to a power plant via stochastic simulation

Abstract

The new generation of coal-fired power plants in Brazil are expected to use run-of-mine (ROM) coal as fuel; however, variations in the quality of the coal are likely to be transferred from the mine to the final user. Contracts frequently establish maximum/minimum limits for geological and technological parameters, allowing for rejection or imposition of penalties when the quality of the lots is outside the agreed limits. The high costs associated with such penalties necessitate quality control of the final product (ROM). Geostatistical methods can be used to predict in situ ore quality and variability. Geostatistical methods employ a block model with grades assigned to each block. The block model obtained through the use of ordinary kriging is inappropriate in accessing the uncertainty associated with the geological or technological parameters being modelled. Consequently, this procedure fails to predict grade fluctuations correctly. Contrary to kriging, simulation methods aim at reproducing in situ grade variability and spatial continuity. Once one produces a block model, with each grade uncertainty evaluated, quality fluctuation can be predicted for any given mining route. Also the fluctuations associated with the size of a lot to be delivered to the customer can be predicted. Different mining routes and sizes of ore lots are tested until the desired level of grade oscillation is achieved. Results from a case study at a Brazilian coal mine prove the adequacy and functionality of the method and that geostatistical simulation adequately predicts quality fluctuations.