Research on estimating methods of coalbed proximate analysis using geophysical log data

摘要

The coal proximate analysis results are the most important indicators of coal quality analysis as well as the basic of evaluation and analysis of the underground coalbed. Methods like linear and multiple regressions, constrained optimization algorithm based on volume model, and BP neural network were introduced into analyze the moisture, ash, volatile-matter and fixed carbon by using in-situ geophysical log data under non-laboratory conditions. The results showed that in the case of enough samples, the priority choice should be regression analysis method based on the statistical model, for little sample data taking BP network to construct intelligent algorithm between logging data and coalbed proximate analysis is feasible though should consider improving its generalization capability and avoiding local minimum. Based on constrained optimization algorithm and constructing the minimum error between theoretical and log data, and combine the coalbed proximate analysis distribution range then could get high precision ash predication and relatively high precision for fixed carbon and volatile-matter.