Data dynamic feature analysis and control technology are considered as important data mining methods. The coal-fired boiler data have obvious characteristics such as timing and drift. In view of the current lack of dynamic, real-time problems of data tracking monitoring algorithms, we propose an adaptive clustering model based on improved growing neural gas(GNG). The improved GNG is used to realize real time monitoring for drift data. The node generation and deletion mechanism are established. based on probability, range search and node average distance. Finally, the dynamic data for coal-fired boilers are taken as the analysis object. The experimental results show that the model and algorithm have stronger real-time tracking ability for dynamic drift data, and can effectively supervise and control the dynamic data of coal-fired boilers.