Ιδρυματικό Αποθετήριο
Πολυτεχνείο Κρήτης
EN  |  EL

Αναζήτηση

Πλοήγηση

Ο Χώρος μου

Development of new geostatistical methods for spatial analysis and applications in reserves estimation and quality characteristics of coal deposits

Pavlidis Andreas

Πλήρης Εγγραφή


URI: http://purl.tuc.gr/dl/dias/E1D3F174-4ABE-4B73-8E6E-A844E8038928
Έτος 2016
Τύπος Διδακτορική Διατριβή
Άδεια Χρήσης
Λεπτομέρειες
Βιβλιογραφική Αναφορά Andreas Pavlidis, "Development of new geostatistical methods for spatial analysis and applications in reserves estimation and quality characteristics of coal deposits", Doctoral Dissertation, School of Mineral Resources Engineering, Technical University of Crete, Chania, Greece, 2016 https://doi.org/10.26233/heallink.tuc.66731
Εμφανίζεται στις Συλλογές

Περίληψη

Coal is an important energy resource, especially for countries that have few other energy resources. While the use of fossil fuels is linked to several environmental challenges, these problems can be mitigated. Coal reserves are distributed in space. A better understanding of their spatial distribution would improve exploitation plans and help to better assess financial risks. This research attempts to supplement and improve existing methods of reserves estimation for coal and provide easy-to-use tools to assist in mine planning. The methods proposed herein could also be used for different ore deposits. We introduce the spatial profitablity index (SPI) that evaluates the profitability of mining each individual seam. As prices and mining costs change during the lifetime of a mine, the profitabilty of different mine sectors or lignite seams may change from profitable to unprofitable or vice versa.The SPI is a flexible tool that can easily and quickly investigate different economic scenarios. The SPI can be used to re-evaluate the pit limits and mine reserves with the current prices and mining costs or near-future estimates. The SPI is applied in the multiseam lignite mine of Mavropigi, Northern Greece, to evaluate the different lignite seams based on data from 341 drill holes provided by the Public Power Corporation (PPC). The differences in the reserves estimation between the SPI-corrected data and the original data are investigated with regression kriging. The uncertainty of the prediction is investigated based on 5000 conditional simulations. Based on the SPI, different economic scenarios are investigated that cover a large range of revenue and cost changes. The estimated reserves and revenue difference based for these scenarios are approximated by an empirical function. This function or the resulting graph can give a quick and accurate approximation of the expected difference in reserves for specific mine sectors or the entire mine. The uncertainty of the prediction is assessed with conditional simulations.In mining, reserves estimation is usually performed using the family of kriging methods. While these methods are the best linear estimators for Gaussian data, they require the computationally intensive inversion of a large covariance matrix. To reduce the computational load, a user-defined neighborhood radius is usually introduced. In this thesis, we investigate the recently proposed method of the stochastic local interaction model (SLI) as a possible alternative. This method avoids the inversion of the covariance matrix. The bandwidth parameter required by the method is self-consistenly defined by the geometry of the data-set without any input from the user.SLI models are used with three different data-sets and compared with kriging methods. The first data used for SLI estimation are the SPI-corrected data from Mavropigi mine. The second data-set is from Campbell county, in the state of Wyoming, USA to assess the distribution of the coal reserves there. The coal reserves of Campbell country constitute an important economic and energy resource for the area. The third and final data set investigated by SLI and kriging, is a non-Gaussian dataset from a gray scale photograph of Pluto. While this dataset does not involve coal deposits, it is employed in this thesis to showcase the performance of SLI methods compared to kriging methods in non-gaussian datasets.In all three datasets SLI performs faster than the kriging methods and performs as accurately or in the case of non-Gaussian datasets more accurately than kriging methods.

Διαθέσιμα αρχεία

Υπηρεσίες

Στατιστικά