الفهرس 
1. INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 156 |  |  |
| | | from | 156 | | |
Abstract
ABSTRACT Many Industrial processes are large in dimension and have focused problems in their technology and behavior. These systems are large scale systems. Large scale systems are stochastic in nature and highly nonlinear complex systems due to the large size of dynamic models, the varying system parameters, and the interconnected structure of the system.
Power system is one of the large scale systems that have very nonlinear complex dynamic units. The control of power plant system has a central role of plant performance where the power plant control system should have the capability to achieve an optimal tracking property of the nonlinear MIMO units. The importance of power plant control is highlighted to adjust the output power to meet the electrical load demands. Several types of problems which affect controller performance are increasingly taken into account in the controller design, especially for power plant under large disturbances. Such as, the nonlinearities and uncertainties of power plant models, undesirable interactions among multiple controllers, different control objectives within varying operating regions and their inherent conflicting objectives.
Boiler- turbine unit is a crucial part that has an important role in the steam power plants. Boiler- turbine system is a nonlinear, time varying multi-input multi-output (MIMO), highly interconnected and uncertain industrial process whose states generally vary with operating conditions. There is, therefore, a strong motivation to design a good controller for stabilization of this system. The central task of the boiler- turbine system control is to adjust the output power to meet the electrical load demand, while maintaining the steam pressure and water level in the drum within acceptable tolerance.
This work proposes a decentralized control system of the nonlinear boiler-turbine system based on adaptive wavelet neural networks. The proposed control system is formed from three separated local controllers, wavelet neural networks, for controlling three outputs of the boiler-turbine system; drum steam pressure, electrical output power and drum water level. The
stability of the control system is analyzed through this work by deriving adaptive learning rates from the discrete Lyapunov stability theorem and the convergence of the local controllers was guaranteed by training the local wavelet neural networks with the these learning rates. Simulation results showed the improvement of the wavelet neural networks in comparison between the decentralized controller based on the neural networks and wavelet neural networks. The results also illustrated the feasibility and benefits of the decentralized control system instead of the centralized control system.
This work also, presents a decentralized single node wavelet neural network controller of the nonlinear boiler-turbine system. The local wavelet neural network is composed of single layer. This simple network structure reduces the complexity of the control system and accelerates the learning process and therefore reduces the time required to the system stabilization. Finally, the simulation results show that the good performances of the proposed controllers and their robustness to satisfy stable tracking of the boiler-turbine system to the desired outputs under large disturbances, load changes and uncertainties.