The Design of an Optimal Computer Controller for a Steam Generating Plant

Abstract

The paper describes part of a study into the application of modern control theory to the design of computer controllers for industrial processes. The study is centered on a 50kW oil-fired steam generator interfaced with a digital control computer.The form of controller being developed incorporates a simple linear model of the boiler to predict its response to particular controls. Provision is made for constraints upon the amplitude of critical plant variables, and for finite limits upon the excursion of the control variables . Full use is made of prior knowledge of future load changes to be demanded of the boiler. To allow for model imperfections and unknown plant disturbances, the state of the model is frequently updated from measurements on the plant, and also the control is periodically recomputed.In developing the controller in a form suitable for implementation by a small computer, a fairly sophisticated non-linear model having S non-linear differential equations and a corresponding set of non-linear algebraic equations has been developed from physical reasoning. Its parameters have been determined from observations of actual plant responses using least squares followed by the Gauss-Newton procedure. This model has been programmed into a large computer to simulate the plant.Using the simulated plant, experiments have been carried out with simplified models (3 - and 5-state linear models) in the controller which are also set up in the same computer for prediction purposes. These have revealed the configuration and speed of the control computer which is necessary for implementation of optimal control on the actual plant. They have also indicated some of the improvements over existing closed loop analogue control to be expected from the on-line optimal controller. The practical implementation of the on-line optimal control scheme is nearing completion.