Extracting the Extreme Mode Operation of Coal Mills Using the Gaussian Mixture Model Method

Coal mills have been the heart of running coal-fired power stations. Coal mills must operate safely and normally to avoid unnecessary unit load losses and complete shutdowns. It has been desirable to have a model that can detect anomalous behavior of a system. The objective of this paper is to be able to isolate unsafe operations or clusters from an unlabeled coal mill dataset. In this paper, Gaussian Mixture Model (GMM) algorithm was implemented on a two-year dataset from an MPS-type vertical spindle coal mill of a power plant located in South Africa to label and classify them into different clusters. To implement GMM, one must assume the number of clusters in advance. Cluster performance evaluation criteria were selected to confirm and evaluate the assumed number of clusters. Furthermore, the operating behavior of coal mills was considered. GMM algorithm was found to fit the system dataset well, and it is good at isolating safe and unsafe modes of operation. This depends on finding the right balance in choosing the appropriate number of clusters and features.