Receding horizon based adaptive control having means for minimizing operating costs
AssigneeHoneywell Inc.
InventorsJ. Ward MacArthur, David A. Wahlstedt, Michael A. Woessner, Wendy K. Foslien
Life of this patent
Estimating the term…
Abstract
Disclosed is an adaptive and predictive controller having self tuning capabilities. The receding horizon window and sample interval are determined by the tunning mechanism automatically, based on the relationship between the controlled and manipulated variables. The interval and window may be set by the user as well. Constraints on the manipulated and controlled variables at the end of the horizon window are accommodated automatically. In addition to providing servo regulatory control the invention teaches simultaneous minimization of operating costs.
Claims16
A controller for controlling manipulated variable inputs to a plant to achieve a desired controlled variable output trajectory which is sensitive to the outputs from said plant wherein said controller receives user input and provides output signals to effect control moves wherein said controller establishes a model indentification based on the manipulated variable inputs and the controlled variable outputs from the plant and comprises:
- a tuner for establishing a control sample rate with reference to how long the controlled variable settling time is and on how many controlled and manipulated variable samples are used to identify the model, and for establishing the first m increments of a horizon window which has a total of m+1 increments and
- a control law calculator means having a recursive model based predictor to predict the nominal control variable output trajectory that would occur without changes to the manipulated variables, said predictor deriving its model from the indentified model, and having a model based corrector, also based on the same identified model, to specify the changes required of the manipulated variable outputs to satisfy the following criteria:
- a) the errors between the controlled variables and their respective setpoints become zero after m control moves,
- b) the errors between the controlled variables and their respective setpoints are zero for 1 additional control moves, and
- c) the change in the manipulated variables during the last n control moves of the m+1 horizon take on prescribed beta values which are determined by said control law calculator to be those values necessary to smooth the change in manipulated variables after m control moves and reduce the steady state controlled variable errors, and "means for controlling the controlled variable output based on said specified manipulated variable changes."
A process for regulating a system with a controller which provides inputs to a plant and receives outputs from the plant indicating changes in the plant comprising the steps of:
- a) establishing by the controller a horizon window comprised of sampling increments of time such that the sampling interval will be long enough to provide useful control move change periods and short enough to substantially reduce the appearance of unwanted oscillations of the controlled variable trajectory.
- a) predicting the nominal controlled variable output trajectory,
- c) predicting the controlled output trajectory based on the control moves required to reach a user-defined setpoint trajectory at a given time of m increments within the horizon window,
- d) constraining the controlled output trajectory to remain at the user defined output state for l increments beyond said m increments, and constraining the control moves to equal n user-specified betas for the last n increments of the l part of the horizon window.
Dependencies are inferred from the claim text.