Automatic calibration of a model for a complex flow system
View Patent ↗Aspects of the present invention provide a solution for calibrating a model of a complex flow system. In an embodiment, a comparison is made between the output from the model and a set of observed values for each of a plurality of nodes in the complex flow system. An adjoint sensitivity is computed for each of the nodes based on the comparison. These computed adjoint sensitivities are used to adjust a set of coefficients of the models. This calibration process can be performed multiple times, periodically and/or continuously to maximize the accuracy of the model.
1. A method for calibrating a model of a complex flow system, comprising:
making a comparison between an output from the model and an observed value for a node of a plurality of nodes in the complex flow system;
computing an adjoint sensitivity for the node based on the comparison; and
adjusting a set of coefficients of the model associated with the node based on the adjoint sensitivity, wherein the complex flow system includes a flow of water in a river system and wherein the set of coefficients includes a capacity coefficient, a slope coefficient and a friction coefficient.
2. The method of claim 1 , wherein the computing computes the adjoint sensitivity for a plurality of nodes simultaneously.
3. The method of claim 1 , wherein the observed value is collected from an embedded sensor within the complex flow system.
4. The method of claim 1 , wherein the observed value is one of a set of observed values in a stream of observed values collected by the embedded sensor in real time.
5. The method of claim 1 , wherein the adjusting and the readjusting of the set of coefficients is performed using a quasi-Newton method.
6. The method of claim 1 , wherein a friction component is adjusted during the adjusting and the readjusting of the set of coefficients.
7. The method of claim 1 , further comprising:
making a comparison between the output from the model and an observed value of a second node of a plurality of nodes in the complex flow system;
computing an second adjoint sensitivity for the second node based on the comparison; and
adjusting a set of coefficients of the model associated with the second node based on the second adjoint sensitivity.
8. A system for calibrating a complex flow system model, comprising at least one computer device that performs a method, comprising:
making a comparison between an output from the model and an observed value for a node of a plurality of nodes in the complex flow system;
computing an adjoint sensitivity for the node based on the comparison; and
adjusting a set of coefficients of the model associated with the node based on the adjoint sensitivity, wherein the complex flow system includes a flow of water in a river system and wherein the set of coefficients includes a capacity coefficient, a slope coefficient and a friction coefficient.
9. The system of claim 8 , wherein the computing computes the adjoint sensitivity for a plurality of nodes simultaneously.
10. The system of claim 8 , wherein the observed value is collected from an embedded sensor within the complex flow system.
11. The system of claim 8 , wherein the observed value is one of a set of observed values in a stream of observed values collected by the embedded sensor in real time.
12. The system of claim 8 , wherein the adjusting and the readjusting of the set of coefficients is performed using a quasi-Newton method.
13. The system of claim 8 , wherein a friction component is adjusted during the adjusting and the readjusting of the set of coefficients.
14. The system of claim 8 , the method further comprising:
making a comparison between the output from the model and an observed value of a second node of a plurality of nodes in the complex flow system;
computing an second adjoint sensitivity for the second node based on the comparison; and
adjusting a set of coefficients of the model associated with the second node based on the second adjoint sensitivity.
15. A computer program product stored on a non-transitory computer readable storage medium, which, when executed performs a method for calibrating a complex flow system model, comprising:
making a comparison between an output from the model and an observed value for a node of a plurality of nodes in the complex flow system;
computing an adjoint sensitivity for the node based on the comparison; and
adjusting a set of coefficients of the model associated with the node based on the adjoint sensitivity, wherein the complex flow system includes a flow of water in a river system and wherein the set of coefficients includes a capacity coefficient, a slope coefficient and a friction coefficient.
16. The program product of claim 15 , wherein the computing computes the adjoint sensitivity for a plurality of nodes simultaneously.
17. The program product of claim 15 , wherein the observed value is collected from an embedded sensor within the complex flow system.
18. The program product of claim 15 , wherein the observed value is one of a set of observed values in a stream of observed values collected by the embedded sensor in real time.
19. The program product of claim 15 , wherein the adjusting and the readjusting of the set of coefficients is performed using a quasi-Newton method.
20. The program product of claim 15 wherein a friction component is adjusted during the adjusting and the readjusting of the set of coefficients.
21. The program product of claim 15 , the method further comprising:
making a comparison between the output from the model and an observed value of a second node of a plurality of nodes in the complex flow system;
computing an second adjoint sensitivity for the second node based on the comparison; and
adjusting a set of coefficients of the model associated with the second node based on the second adjoint sensitivity.