Masahide Nomura

Adaptive optimal control of steam temperatures for thermal power plants

This paper describes a new adaptive optimal control method for boiler steam temperature control of thermal power plants. Fig. 1 shows a schematic diagram of the adaptive optimal control system. A coal-fired power plant and the automatic plant control (APC) system are assumed to be a single controlled object in order to allow the APC to continue controlling the power plant, even if the adaptive optimal control system fails. A process dynamics model, composed of the above mentioned controlled object and a disturbance source, is proposed. The process dynamics model is represented by a multi-input and output autoregressive moving average (ARMA) model to predict future trends in process behavior. The disturbance source is the central load dispatching office. Parameters of the process dynamics model are identified during high and low load level power plant operation on a real-time basis when load fluctuations are relatively small. Optimal controller gains are calculated using the identified model by the dynamic programming (DP) method and adapted on-line when process characteristics change as a result of changing fuel heating value or during the course of long-term operation. These control gains suppress fluctuations in both main and reheater steam temperatures.

Steam Temperature Prediction Control for Thermal Power Plant

An advanced control method using a microprocessor is developed for boiler steam temperature control of thermal power plants. A process dynamics model with the Kalman filter in the controller is presented which describes the dynamic plant behavior. Using the model, future changes in steam temperature are predicted and the fuel flow rate compensated accordingly. The model is also used to adjust the control parameters, i.e. proportional and integral gains, to provide values which ensure stable system control.

Comparison between various control methods for thermal power plant main control system

In thermal power plants, it is important to improve the control accuracy of main steam pressure and temperature and so forth during load up/down. This paper focuses on temperature control, the most difficult aspect of control due to the nonlinearity and long dead time of power plants. We applied control methods such as MRAC, neural network, and long-range predictive control to the power plant main control system. Each method was evaluated by a simulator using detailed physical models that represent accurately the power plant dynamics. We confirmed that each method can provide proper control, but long-range predictive control is better than the two other methods. In addition, thermal power plants are so complex that further analysis (e.g., persistently exciting condition, learning method of neural networks) is necessary for the application of theoretical algorithms.