Wet flue gas desulfurization (WFGD) is very important in reduction of SO2 emission in power plant for its lower investment cost, higher desulfurization efficiency and useful by-products. Healthy slurry works as the premise of performance analysis and optimization in WFGD system. However, little research has been conducted on monitoring gypsum slurry quality deterioration. In this paper, an on-line clustering framework has been proposed to monitor gypsum slurry quality in desulfurization system based on data mining. Compound parameter gamma, is put forward to remove the influences to slurry quality from gas volume and inlet SO2 concentration to CaCO3 slurry flow. Thus, after the simplification, desulfurization efficiency, pH value and gamma are taken as parameters for gypsum slurry quality monitoring. A new clustering method, EKFCM, based on improved fuzzy clustering algorithm, Kmeans and fuzzy C-means combined with entropy theory is proposed. EKFCM is superior to basic FCM in finding the clustering number without prior knowledge when dealing with off-line data, verified by a self-defined function with validity indexes comparison. A new rule, Sub-TDFO, with time decay inserted into "first in first out" in the subset, is proposed to the framework for on-line learning. Another self-defined function is added to the former one for on-line process simulation to verify the effectiveness of the proposed framework. Then, WFGD system in a 600 MW unit is worked as the clustering case study for gypsum slurry quality on-line monitoring and quantization. The clustering results from the proposed on-line framework can be applied to illustrate the process of gypsum slurry quality variation. Moreover, this method be used in other industry processes data for its on-line features. (C) 2018 Elsevier Ltd. All rights reserved.