Condition Monitoring of Power Plant Milling Process using Intelligent Optimisation and Model Based Techniques

Coal preparation is the first step in the whole process of coal-fired power generation. A typical milling process is illustrated in Figure 1. Also, coal-fired power stations nowadays are required to operate more flexibly with more varied coal specifications and regularly use coal with higher volatile contents and Biomass materials (Livingston 2004); this greatly increases the risks of explosions or fires in milling plants. The power stations are also obliged to vary their output in response to the changes of electricity demands, which results in more frequent mill start-ups and shut-downs. In many cases, coal mills are shutdown and then restarted before they have cooled adequately, which creates a potential fire hazard within the mill. Frequent start-ups and shut-downs of mills will also have an impact on power plant operation safety. Mill fires could occur if the coal stops flowing in the mill and the static deposit is heated for a period of time. Fires in out-of-service mills can cause explosions on mill starts. Fires in running mills can cause explosions on shut-downs. The result of a study indicated that as many as 300 “explosions” were occurring annually in the US pulverized coal industry (Scott, 1995). Especially, adding higher volatile Biomass materials greatly increases the chance of mill fires and explosions. The UK PF Safety Forum had recently reported an increase in the frequency of mill explosions in the UK. Operational safety and efficient combustion require better understanding to the milling process. However, coal mills have been paid much less attention in research compared with boilers, generators, and other power generation system components. It is difficult to identify if there will be a fire in the mill. Outlet temperature and CO are established methods of detecting fires in mills, but at present they are not very effective for detecting small fires. The CO detection system becomes ineffective when the mill is in service due to dilution effects caused by primary air flow and associated oxygen content in the mill. A wide range of literature survey shows that there are only a few reports on mathematical models of milling processes. A detailed milling process description can be found in Scott et al. 1995. An approximated linear transfer function model was obtained by Bollinger et al, in 1983. Mill modelling using system identification method was reported in 1984 (Corti et al. 1984). With specially designed input signals, a linear discrete time model was obtained by Cheetham, et al in 1990, in which system time-delay was considered. An approximated 19