Design and implementation of a real-time knowledge-based controller

A general structure and design specification of a real-time knowledge-based controller have been proposed in this thesis. The controller consists of (1) a knowledge-base, (2) a real-time inference engine, (3) a real-time information pre-processor, and (4) control mechanisms. The control objective is to maintain the overall system stability and performance not only during normal plant operation, but also during contingencies, such as system component failures, operating condition changes, and large load variations. The controller specifications are classified into two groups; one for the normal system operation, and one for the system under contingencies. The inference engine uses the controller knowledge-base and the current information about the system operation provided by the information pre-processor to detect, classify, and correct underlying system failures.;The knowledge-base is constructed from the following knowledge sources: (1) experience of human experts; (2) the model-based control system theory; and (3) extensive off-line computer simulations.;The information pre-processor contains a set of digital signal processing schemes. These schemes can be divided into two groups depending on their complexities. The first group consists of several simple and robust algorithms which are used to monitor overall system performance, and to detect any steady-state or transient abnormalities. The second group is composed of several more complex digital signal processing algorithms, such as adaptive filters, parameter identifiers, and state observers. These algorithms are used by the inference engine for system failure classifications.;To meet the requirement of real-time decision-making, the inference engine employs both forward-chaining (data-driven) and backward-chaining (goal-driven) inference mechanisms. In the forward-chaining process, the first group of information pre-processing algorithms is used, whereas, during the backward-chaining process, the second group of algorithms is employed.;A prototype knowledge-based controller has been designed, and implemented for a hydraulic turbine generator governor system to achieve reliability and security. The hydraulic turbine generator system is simulated on an analog computer. The knowledge-based controller is implemented in real-time on a microcomputer. The inference engine and the knowledge-base are constructed using PROLOG, and the C language is used to implement the information pre-processor. The performance of this prototype knowledge-based controller is evaluated by simulating in real-time various types of system failures.