Tuning model structures of dynamic systems

Patent metadata

Field	Value
Publication number	US-9645563-B2
Application number	US-201213599712-A
Country	US
Kind code	B2
Filing date	Aug 30, 2012
Priority date	Aug 30, 2012
Publication date	May 9, 2017
Grant date	May 9, 2017

Abstract

Tuning model structures of dynamic systems are described herein. One method for tuning model structures of a dynamic system includes predicting a variable for each of a number of models associated with a number of model structures of a dynamic system, calculating a rate of error of the predicted variable for each of the number of models compared to an observed variable, determining a best model structure among the number of model structures based on the calculated rate of error, and creating a revised model structure using the best model structure to tune the number of model structures of the dynamic system.

First claim

What is claimed:

1. A method for tuning a data model structure including:
predicting, using a computing device, a variable for each of a number of models associated with a number of model structures of a dynamic system, wherein each of the number of model structures include a dependency between variables of a model among the number of models;
calculating, using the computing device, a rate of error of the predicted variable for each of the number of models compared to an observed variable;
determining, using the computing device, a best model structure and a worst model structure among the number of model structures based on the calculated rate of error;
creating, using the computing device, a revised model structure using the best model structure by combining a plurality of model structures among the number of model structures;
tuning the number of model structures of the dynamic system by defining a number of new model structures eligible to be the best model structure and replacing the worst model structure with the revised model structure; and
controlling, using the computing device, a heating, ventilation, and air-conditioning (HVAC) system by using the best model structure to optimize control variables associated with controllers of the HVAC system.

2. The method of claim 1 , wherein creating a revised model structure further includes migrating the best model structure.

3. The method of claim 1 , wherein creating a revised model structure further includes migrating a well performing model structure toward the best model structure.

4. The method of claim 1 , wherein creating a revised model structure further includes migrating a random model structure towards the best model structure.

5. The method of claim 1 , wherein creating a revised model structure further includes creating a new model structure using surrogate modeling including exploration and exploitation.

6. The method of claim 1 , further including:
receiving prioritization of a number of variables from a user; and
wherein the prioritization is used to calculate the rate of error.

7. The method of claim 1 , wherein calculating a rate of error further includes comparing a number of predicted variables to a number of observed variables for each the number of models associated with a number of model structures.

8. The method of claim 7 , wherein a model structure further includes a variable among the number of variables that impacts a critical variable.

9. A non-transitory computer-readable medium storing a set of instructions executable by a processor to cause a computer to:
identify a number of models associated with a number of model structures of a dynamic system using data, wherein the number of model structures include dependencies between a number of variables for each of the number of models, including:
a linear dependency; and
a bandwidth for regression;

predict a value for each of the number of variables for each of the number of models;
calculate a rate of error for each of the number of models including comparing the predicted value for the number of variables to an observed value for the number of variables for each of the number of models;
compare the calculated rate of error to determine a best model structure among the number of model structures and a worst model structure among the number of model structures;
update the number of model structures with a revised model structure using the best model structure by combining a plurality of model structures among the number of model structures;
tune the number of model structures of the dynamic system by defining a number of new model structures eligible to be the best model structure and replacing the worst model structure with the revised model structure; and
control a heating, ventilation, and air-conditioning (HVAC) system by using the best model structure to optimize control variables associated with controllers of the HVAC system.

10. The medium of claim 9 , wherein the instructions are further executable to save a number of immature model structures.

11. The medium of claim 9 , wherein a number of immature model structures includes a number of model structures with less than a threshold number of evaluations.

12. The medium of claim 9 , wherein the instructions are further executable to update the calculated rates of errors of the number of model structures of the dynamic system over time.

13. A system for tuning a model structure of a dynamic system, the system comprising a processing resource in communication with a non-transitory computer readable medium, wherein the non-transitory computer readable medium includes a set of instructions and wherein the processing resource is designed to carry out the set of instructions to:
identify a number of models associated with a number of model structures using data, wherein the number of model structures include dependencies between a number of variables for each of the number of models, including:
a linear dependency; and
a bandwidth for regression;

evaluate the number of model structures, wherein evaluating each model structure includes:
calculate a predicted value for the number of variables in a model associated with a model structure over a time period;
compare the predicted value for the number of variables to an observed value for each of the number of variables over the time period; and
calculate a rate of error for the model structure based on the comparison of the predicted values and the observed values;

identify a best model structure and a worst model structure based on the calculated rate of error for the number of model structures;
create a revised model structure using the best model structure by combining a plurality of model structures among the number of model structures, wherein the revised model structure includes a model structure with a calculated rate of error less than the calculated rate of error of the best model structure;
update the number of model structures with the revised model structure to tune the number of model structures of the dynamic system by defining a number of new model structures eligible to be the best model structure and replacing the worst model structure with the revised model structure; and
control a heating, ventilation, and air-conditioning (HVAC) system by using the best model structure to optimize control variables associated with controllers of the HVAC system.

14. The system of claim 13 , wherein the system is a portion of a supervisory control system and the system runs adaptively using the Internet.

15. The system of claim 13 , wherein the worst model structure includes a model structure with a higher calculated rate of error than the remaining number of model structures.

16. The system of claim 13 , wherein the worst model structure includes a model structure with less than a threshold number of evaluations.

17. The system of claim 13 , wherein the processing unit is designed to further carry out the set of instructions to save a threshold number of well performing model structures.

18. The system of claim 17 , wherein the threshold number of well performing model structures include a number of model structures with a lower calculated rate of error than the remaining number of model structures.

Assignees

• Macek Karel
• Marik Karel
• Honeywell Int Inc

Inventors

• Macek Karel
• Marik Karel

Classifications

• F24F11/64

  using pre-stored data · CPC title

• F24F11/58

  using Internet communication · CPC title

• F24F11/46 Primary

  Improving electric energy efficiency or saving · CPC title

• F24F11/30 Primary

  for purposes related to the operation of the system, e.g. for safety or monitoring · CPC title

• G05B17/02 Primary

  electric · CPC title