Data processing: structural design – modeling – simulation – and em – Modeling by mathematical expression
Reexamination Certificate
2000-08-16
2004-11-09
Jones, Hugh (Department: 2128)
Data processing: structural design, modeling, simulation, and em
Modeling by mathematical expression
C703S006000, C700S032000
Reexamination Certificate
active
06816822
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to processes having operating targets and more particularly to using dynamic modeling, parameter estimation and optimization for such operations.
DESCRIPTION OF THE PRIOR ART
One example of a process that has operating targets is a blending operation. Steady-state modeling systems and steady-state optimization have been used for quite some time with blending operations such as the blending of crude oil and gasoline.
A well-developed feature of many steady-state modeling systems is their ability to take on-line, real-world data and match the response values against a model of the system to make certain data inferences. The advantages to this procedure is that it:
a. corrects errors such as thermocouple drift and orifice plate degradation that may be present in plant sensor measurements;
b. identifies values for unmeasured disturbances such as unmetered flowrates and feed quality characteristics that occur to the system that will impact how the process should most efficiently operate; and
c. establishes equipment operating parameters such as exchanger heat transfer coefficients and distillation efficiencies that change during or through the operation of the process.
There are, however, disadvantages to solely using steady state data reconciliation and parameter estimation. The disadvantages include:
a. their inability to process data from a dynamically changing process; and
b. a limited set of data points from which to process data.
In reality, most steady-state optimization systems that rely upon steady-state data reconciliation and parameter estimation are ill-designed particularly if an operator is depending upon the optimization system to make efficient changes to the system as there will always be some dynamic movement in the plant. Without data reconciliation using dynamic data, the optimization engine would be forced to operate on a model with inaccurate data.
The benefits of steady-state optimization are well known to those skilled in the art. One example of such optimization is described in U.S. Pat. No. 3,940,600 entitled “Method and Apparatus For On-Line Non-Interactive Blending Using An Uncoupling Matrix.” The method and apparatus described therein is identified for blending and in particular for blending liquid products, e.g. petroleum. The model representation of the blended product is a linear, steady-state model representation for determining the output control signals, e.g. component flowrates.
U.S. Pat. No. 5,933,345 entitled “Method and Apparatus For Dynamic and Steady-State Modeling Over A Desired Path Between Two End Points” describes the combined use of a steady-state, that is, static, model and then a dynamic model to meet the use objectives. The steady-state model is described as a neural-network representation of the plant or system to identify gains for the dynamic model. The dynamic model is described as of a linear nature. The dynamic gains are adjusted based upon the prior built static model in an effort to make the dynamic model better match the true operations of the plant.
U.S. Pat. No. 5,499,188 entitled “Flexible Method For Building A Recipe In A Process Control System” addresses the general nature of recipe formulation and how that recipe should be invoked through a control system configuration. The method described in this patent does not use actual plant response(s) to either estimate parameters or adjust the recipe. Additionally, the method described in this patent does not address planned, multiple recipe formulations.
U.S. Pat. No. 4,786,182 entitled “Method And Means For Controlling A Fodder Mixing Plant” discloses a method to optimize, in relation to the total production costs, the recipe that affects the physical properties of produced pellets based upon the current sampling of the pellets and raw materials.
U.S. Pat. No. 4,642,766 entitled “Method And Means Of Control For Multi-Source Feedstock Distribution System Including Optimization Of Supplies” discloses the optimization of such distribution where current flow and even property values if available can be adjusted to optimally distribute the feedstock in an effort to minimize the total cost of the feed. U.S. Pat. No. 3,826,904 entitled “Method And Apparatus For The Optimum Blending Of Lubricating Base Oils And An Additive” describes a method and apparatus for enabling the computation of a minimum cost for lubricating oil base stocks to its product specifications. Non-linear formulae are used in arriving at the optimum blend specification.
In contrast to the foregoing the method and apparatus of the present invention uses:
a. a non-linear, rigorous dynamic model that represents both the dynamic and steady-state behavior of the plant and the blending system;
b. parameter estimation to among other things:
i. estimate component or feed properties;
ii. update the measured disturbances and other model inputs to ensure the results actually obtained from the plant or system in real-time match the output of the model, that is, such estimation forces the model to match the plant; and
iii. avoid requiring continuous sampling of raw materials.
c. constrained dynamic optimization to among other things:
i. determine when and how multiple recipes should be invoked over the course of a blend;
ii. predict recipes or operating points over the entire run-life of a blend cycle given process constraints without the need to ever achieve a steady-state operating state;
iii. identify the optimum times the new recipes or targets should be implemented in the blend cycle to achieve the lowest potential blend cost; and
iv. state how multiple recipes can be formulated and optimized.
While the benefits of steady-state optimization have been well documented in the art, dynamic optimization provides an even more powerful tool that has the potential of providing even greater economic rewards. The need to project out and make economic determinations on both the blend recipe and the extent of time a particular recipe should be invoked can arise in many instances. Some examples include:
a. a change in product specification based upon higher level planning and scheduling decisions;
b. recognizing the loss of availability of a particular blend component;
c. lining up the blender for an on-line start to the next scheduled blend;
d. a failure to recognize or invoke the most optimum blend specification; and
e. optimally correcting for tank heal or existing tender properties.
Increasing the number of recipes over the course of the blend allows the optimizer to take the best possible advantage of the equipment utilization against the constraints for the blended components that are available.
While the technique of the present invention will be described herein in connection with a process plant that is a blending operation those of ordinary skill in the art will appreciate that the technique of the present invention may also be used in connection with any process which uses recipes as operating targets or any other operating targets such as setpoints.
SUMMARY OF THE INVENTION
A method for dynamically optimizing a process. The method has the steps of performing an optimization cycle at a first rate and simultaneously performing with the optimization cycle a simulation cycle at a second rate which is equal to or faster than said first rate. The simulation cycle has the steps of collecting data about the process; using the collected process data to execute in a predefined time interval during the simulation cycle a dynamic simulation for a dynamic model of said process; and storing data for the dynamic simulation.
The optimization cycle has the steps of:
collecting data about the process;
determining if a dynamic parameter estimation and data reconciliation procedure is needed to minimize differences between the collected process data and the data stored for the dynamic simulation where that data was calculated during a simulation cycle corresponding to a period of time which is the same as the time period during which the process data was collected during the simulation cycle;
pe
Hess Todd M.
Kroll Andreas
Richards Martyn Glenn
ABB Automation Inc.
Jones Hugh
Rickin, Esq. Michael M.
LandOfFree
System and method for dynamic modeling, parameter estimation... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for dynamic modeling, parameter estimation..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for dynamic modeling, parameter estimation... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3325260