Predictors for application to real-time adaptive control of a diesel prime-mover

Adaptive control of a power prime-mover requires effective modeling and identification techniques that have good disturbance rejection properties. Existing methods tend to avoid explicit modeling of input dead-time due to resulting computational complexity. Two different approaches of modeling the diesel power prime-mover are presented. The predictors are derived from the results of two different least-squares estimates of the plant. The derived predictors are compared with each other, and with their respective least-squares models, on the basis of disturbance rejection capability and computational complexity. It is shown by extensive simulation studies that the predictors obtained converge quickly and operate with very small prediction error under severe load disturbances and large speed reference changes. A comparative discussion of the two methods is presented.<>