A new approach of the adaptive control, MRAC, is proposed here for the level control of the steam generators. A model reference adaptive system is composed of four elements; a reference model, an adjustable system, a subtractor, and an adaptation mechanism. A reference model specifies the desired performance; the performance of an adjustable system should be as close as possible to that of the reference model. A subtractor forms the error between the outputs of the reference model and of the adjustable system and an adaptation mechanism processes the auxiliary error in order to modify the parameters of the adjustable system.
The parallel configuration is used to estimate the parameters since the structure is robust to noises. The parameters of the controller are estimated directly by the modified recursive least square method. The perturbation signal is added to the steam flow rate to estimate well the parameters of the steam flow rate.
The reference model is chosen which shows the swell phenomenon but does not show the shrink phenomenon. The variations from steady-state conditions are used to generate the control input, which gives the more robust algorithm and reduces the input variations. Cost function is derived to contain the squared error between the outputs of the reference model and of plant and the squared error between the steam flow rate and the feedwater flow rate. Using the cost function makes the integral action unnecessary and gives good performance in spite of the major disturbance of the load change. The present algorithm has the one tuning parameter of the input weighting factor: the tuning of the present controller is easier than that of the P-I controller which has four tuning parameters.