Steam temperature controller with LS-SVR-based predictor and PID gain scheduler in thermal power plant

Abstract

Nonlinearity and time delay in thermal power plant processes during the period of steady state load have been reported to be efficiently compensated for by controlling the open rate of sprays and an angle of burner tilt using a conventional cascade PID controller. However, it is not easy to compensate severe nonlinearity and time delays simultaneously due to feed water and coal flow variations during load changes. This paper introduces advanced method to mitigate these problems. A predictor based on the least square support vector machine for regression (LS-SVR) algorithm is developed to efficiently compensate the nonlinearity in the boiler system, and it enables accurate modeling of the steam temperature by applying LS-SVR algorithm by using one variable, steam temperature, for the superheater and reheater systems. Moreover, the predictor enables to compensate the time delay by generating a prior control action, based on the predicted steam temperature after a certain time interval. An LS-SVR-based predictor is combined with a PID controller that uses a gain scheduler based on an anti-reset windup algorithm to enable more sensitive and efficient steam temperature control during load changes to the boiler system in a thermal power plant. A load-changing simulation is conducted, and the proposed steam temperature controller demonstrates a more stable and efficient performance than a conventional cascade PID controller.