Application of multivariate adaptive regression in soft-sensing and control of UCG

Abstract

The technology of underground coal gasification (UCG) is still under development and provides an alternative to conventional coal mining. Process monitoring is the necessary part of a complex control system because it provides essential information for control level. Monitoring and control improve the behaviour and effectiveness of the technological process. This paper introduces a novel approach to soft-sensing in UCG based on multivariate adaptive regression splines (MARS). This technique can support monitoring the process variable that is inaccessible for standard measuring hardware. The MARS method was applied for modelling of underground temperature from the syngas composition. The paper also presents advanced approaches to control based on an adaptive regression model. The proposed control can increase or maintain the syngas calorific value during UCG operation. The proposed methods have shown interesting results and can be applied to industrial automation devices or implemented as support algorithms for the monitoring system. Methods were verified in experimental coal gasification on an ex-situ reactor.