On-line monitoring for coal Low Heating Value in coal-fired power plant boiler based on CM-LSSVM-PLS

摘要

A novel comprehensive real-time on-line calculation model of coal's Low Heating Value (LHV) in coal-fired boiler is presented in this paper. Firstly, the training operational data is partitioned into several subsets by means of C-Means cluster method. Then, the Least Squares Support Vector Machine (LS-SVM) method is used to training the sub-models in each subset. Finally, the sub-models are combined into one model based on Partial Least Squares algorithm (PLS). The models' effectiveness is illustrated by the validation of online simulation in which the real operation data is used. The result shows that it is potential to considerably fulfill not only the accuracy of online monitoring of LHV of the coal into the boiler, but also the real-time computations that are supplied by using indirect heat balance approach.