A thermal plant as a controlled system has many state parameters, which cannot be measured directly in many cases. The control efficiency can be improved by applying estimated parameters to the control of the plant. A fluidized-bed incinerator is not constant in the quantity and quality of the feed refuse, which is the fuel, and combustion control is difficult owing to the rapidity of completion of combustion. However, it is thought to be possible to improve the efficiency of CO control and NOx control drastically if the state parameters such as the combustion rate on the upper bed site and the bed site, and the effective air ratio are known.
This paper proposes a method that estimates these state parameters by means of sensor outputs such as temperature, air flow rate, and cooling water rate, using dynamic characteristic analysis and neural networks. This paper also shows that it is possible to estimate the state parameters of an actual incinerator. Further, it is shown that the generalization of parameters estimation equations enables the application of the method to other plants.
