Model predictive controller

In a process control system having input means for receiving signals indicative of measured process parameters, processor means for making process control decisions, and output means for transmitting command signals from said processor means which include at least one manipulated parameter value that is used to govern physical process equipment, a model predictive controller, comprising:

• interactive modeling means for estimating the current state of at least one unmeasured process variable and for determining the value of at least one manipulated parameter value, said interactive modeling means having both a process model which represents the dynamic behavior of the physical process equipment being controlled and a disturbance model which represents current deviations from said process model and future deviations over a prediction horizon, said disturbance model having variable predictor means for estimating said future deviations as a function of both past and current deviations from said process model;
• and
• real-time executive means for coordinating the operations of said interactive modeling means with said processor means and for projecting a set of future controlled parameter values over said prediction horizon in cooperation with said interactive modeling means and said processor means;
• wherein said real-time executive means includes a time-recorded database of measured data received from said processor means, and data interface means for periodically transmitting said measured data to said interactive modeling means with data identifiers employed by said interactive modeling means, said data identifiers of said interactive modeling means being different than the data identifiers used by said processor means.

A process control system, comprising:

• input means for receiving signals indicative of measured process parameters;
• output means for transmitting command signals which include at least one manipulated parameter value that is used to govern physical process equipment;
• processor means for making independent process control decisions which determine manipulated parameter values in response to said input signals;
• interactive modeling means for periodically estimating the current state of at least one unmeasured process variable and for determining the proposed value of at least one manipulated parameter, said interactive modeling means having both a process model which represents the dynamic behavior of the physical process being controlled and a disturbance model which represents current deviations from said process model and future deviations over a prediction horizon;
• and
• real-time executive means for coordinating the operation of said interactive modeling means with said processor means and for testing the validity of said signals indicative of measured process parameters, the current state of each said unmeasured process variable, and each said manipulated parameter value proposed by said interactive modeling means;
• said processor means having means for determining whether to implement at least one manipulated parameter value that has been validated by said executive means.

A method of controlling a process, comprising the steps of:

• receiving input signals which represent measured parameter values from a plurality of sensors;
• periodically transmitting a time-recorded set of said measured parameter values to an interactive modeler which includes a nonlinear process model, which is derived from physical relationships for the process, and an independent disturbance model;
• estimating the current state of at least one unmeasured process parameter in said interactive modeler from said process model;
• projecting a set of desired controlled parameter values over a predetermined control horizon;
• periodically estimating how said process will react to proposed changes to the value of at least one manipulated parameter over said control horizon, and determining a set of current and future manipulated parameter values which will minimize deviations from said set of desired controlled parameter values;
• testing the validity of said input signals, said current state of each estimated unmeasured process parameter, and each current said manipulated parameter value;
• and
• causing a process control device to implement only manipulated parameter values which are determined to be valid.

In a process control system having input means for receiving signals indicative of measured process parameters, and output means for transmitting command signals which include at least one manipulated parameter value that is used to govern physical process equipment, a model predictive controller, comprising:

• interactive modeling means for estimating the current state of at least one unmeasured process variable and for determining the value of at least one said manipulated parameter value for said process equipment at selected intervals, said interactive modeling means having both a process model which represents the dynamic behavior of the physical process equipment being controlled and an independent, evolving and step-response variable disturbance model which represents current deviations from said process model and future deviations over a predetermined prediction horizon, said process model including at least one nonlinear function which is specifically related to the physical process equipment being controlled, and said disturbance model using a window of past disturbance to predict future disturbances;
• real-time executive means for directing the operation of said interactive modeling means, said real-time executive means having means for periodically transferring some of said signals indicative of measured process parameters to said interactive modeling means and means for transferring said manipulated parameter value from said interactive modeling means to said output means;
• and
• means for testing the validity of said signals which are indicative of measured process parameters, the current estimated state of said unmeasured process variable, and the value of said manipulated parameter value.