Method of model identification for a process with unknown initial conditions in an industrial plant
A method of model identification for a process with unknown initial conditions in an industrial plant, the method comprising collecting a set of manipulated variables and corresponding set of process variables from the process; obtaining a plurality of manipulated variables from the collected set of manipulated variables; for each of the plurality of manipulated variables, obtaining optimal model parameters of a model transfer function and computing a model fitting index for optimized simulated process variables generated by the model transfer function using the optimal model parameters; identifying a best model fitting index among the model fitting indices computed; selecting a manipulated variable associated with the best model fitting index as an initial steady state condition for the model transfer function; and selecting the optimal model parameters corresponding with the best model fitting index as the best model parameters of the model transfer function to tune the controller.
1. A method of model identification for a process in an industrial plant, the method comprising:
collecting a set of manipulated variables and a corresponding set of process variables from the process, the manipulated variables being sent from a controller to the process for adjusting each process variable to reach a set-point, the set point being a desired value for each process variable, each process variable being obtained by a sensor provided in the process, wherein initial conditions of the process are unknown;
selecting a plurality of manipulated variables from the collected set of manipulated variables;
for each of the plurality of manipulated variables,
obtaining optimal model parameters of a model transfer function, the model transfer function being one of: a first order plus time delay model transfer function and an integrator plus time delay model transfer function, and
computing a model fitting index for optimized simulated process variables generated by the model transfer function using the optimal model parameters wherein the computing uses the following equation:
model
fitting
index
=
100
×
[
1
-
Sum
of
Squared
Errors
Total
Sum
of
Error
Squares
]
;
identifying a best model fitting index among the model fitting indices computed;
selecting a manipulated variable associated with the best model fitting index as an initial steady state condition for the model transfer function; and
selecting the optimal model parameters corresponding with the best model fitting index as the best model parameters of the model transfer function to tune the controller.
2. The method of claim 1 , wherein obtaining optimal model parameters comprises searching for and identifying model parameters that result in a minimized fitting error between simulated process variables generated by the model transfer function and the collected set of process variables.
3. The method of claim 2 , wherein the model fitting index is a function of the minimized fitting error.
4. The method of claim 1 , wherein the best model fitting index is a highest model fitting index among the computed model indices that are above a predetermined threshold value.
5. The method of claim 4 , wherein identifying the best model fitting index includes confirming that the model parameters computed for a second highest model fitting index among the computed model fitting indices that are above the predetermined threshold value are within a specified difference from the model parameters computed for the highest model fitting index.
6. The method of claim 1 , wherein the plurality of manipulated variables comprises manipulated variables selected at intervals from a first percentage of the collected set of manipulated variables.