Applying predictive control to a fossil-red power station

Traditionally PID control has been used for the control of boiler plant on large fossil-red power stations. With the requirement for such plant to operate exibly rather than as base-load stations the duties imposed on the control loops have become more onerous. This paper presents the results of investigations into the potential of generalized predictive control as an alternative to PID. Due to limitations on computational load inherent in many power plant control systems, the constraint handling facility had to be modied. A suitable algorithm is developed and illustrated on a high order nonlinear test case. The algorithm is then implemented on coal-red plant and is seen to give a marginal improvement in the accuracy of the control, but with much reduced control activity. Key words: computational cost; constraints; electrical power generation; predictive control Notation z ytz-transform variable System output A(z) z tModel denominator Unknown system disturbances B(z) ûModel numerator Filtered inputs D (z) ŷDifference operator Filtered outputs T(z) D utT-lter Input increment ut RSystem input Vector future setpoints Y WuVector predicted outputs Input weighting matri