Evolutionary Computing Methods for Optimising Virtual Reality Process Models

Modelling the technological processes used in operator training simulators requires specific approaches. Unlike simulation of control loops where a model representing a part of the controlled process needs to be as exact as possible, training simulators do not require very exact modelling. What is required is to give operators the feeling that they are working with a real process. This can even be achieved by simple approximate models that reflect changes of some measured variables in trends or steady state values that are very close to reality. Results from evolutionary computing methods can be used to establish such a virtual reality. In these methods, real process data provided by a process data acquisition system is used to obtain optimised model parameters, values of which cannot be obtained by other conventional identification methods. The proposed procedure has been developed, while creating a virtual model of a thermal power plant for operator training.