### Abstract

Language | English |
---|---|

Pages | 104-131 |

Journal | Control and Intelligent Systems |

Volume | 27 |

Issue number | 3 |

Publication status | Published - Dec 1999 |

### Fingerprint

### Cite this

*Control and Intelligent Systems*,

*27*(3), 104-131.

}

*Control and Intelligent Systems*, vol. 27, no. 3, pp. 104-131.

**Neural Network Model-based Multivariable Predictive Control Algorithms with Application in Thermal Power Plant Control.** / Prasad, G; Swidenbank, E; Hogg, B W.

Research output: Contribution to journal › Article

TY - JOUR

T1 - Neural Network Model-based Multivariable Predictive Control Algorithms with Application in Thermal Power Plant Control

AU - Prasad, G

AU - Swidenbank, E

AU - Hogg, B W

PY - 1999/12

Y1 - 1999/12

N2 - Two neural network (NN) model-based multivariable nonlinear model predictive control (NMPC) algorithms along with their application to more efficient thermal power plant control are presented in this paper. The first algorithm uses a global nonlinear NN model and a nonlinear programming (NLP) routine for optimization to compute optimal control moves. With the help of suitable observer polynomials, modelling errors and other disturbances are given a proper and consistent treatment to achieve offset-free closed-loop behaviour. The second algorithm carries out continuous linearization of the nonlinear global NN model to estimate the local linear model and then applies the generalized predictive control (GPC) method with quadratic programming to compute optimal control moves. Using these algorithms, SISO and MIMO controllers have been designed and implemented in a simulation of a 200 MW oil-fired drum boiler-based thermal power plant. By running the plant simulation over a load profile along with suitable pseudo random binary sequence (PRBS) signals superimposed on controls, the operating data is generated. The NN modelling technique has been used for identifying global dynamic models (NNARX models) of the plant variables off-line from the operating data. The effect of variation in controller parameters and higher plant model mismatch is shown through several test results Excellent performance in controlling main steam pressure and temperature, anti reheat steam temperature in the plant simulation during load cycling as well as other severe plant operating conditions are demonstrated.

AB - Two neural network (NN) model-based multivariable nonlinear model predictive control (NMPC) algorithms along with their application to more efficient thermal power plant control are presented in this paper. The first algorithm uses a global nonlinear NN model and a nonlinear programming (NLP) routine for optimization to compute optimal control moves. With the help of suitable observer polynomials, modelling errors and other disturbances are given a proper and consistent treatment to achieve offset-free closed-loop behaviour. The second algorithm carries out continuous linearization of the nonlinear global NN model to estimate the local linear model and then applies the generalized predictive control (GPC) method with quadratic programming to compute optimal control moves. Using these algorithms, SISO and MIMO controllers have been designed and implemented in a simulation of a 200 MW oil-fired drum boiler-based thermal power plant. By running the plant simulation over a load profile along with suitable pseudo random binary sequence (PRBS) signals superimposed on controls, the operating data is generated. The NN modelling technique has been used for identifying global dynamic models (NNARX models) of the plant variables off-line from the operating data. The effect of variation in controller parameters and higher plant model mismatch is shown through several test results Excellent performance in controlling main steam pressure and temperature, anti reheat steam temperature in the plant simulation during load cycling as well as other severe plant operating conditions are demonstrated.

M3 - Article

VL - 27

SP - 104

EP - 131

JO - Control and Intelligent Systems

T2 - Control and Intelligent Systems

JF - Control and Intelligent Systems

SN - 1480-1752

IS - 3

ER -