Method, computer program with program code means, and computer program product for determining a future behavior of a dynamic system

• US 7,464,061 B2
• Filed: 04/07/2004
• Issued: 12/09/2008
• Est. Priority Date: 05/27/2003
• Status: Active Grant

• Current Assignee: Monument Peak Ventures, LLC (Dominion Harbor Enterprises, LLC)
• Original Assignee: Siemens AG
• Inventors: Grothmann, Ralph, Tietz, Christoph, Zimmermann, Hans-Georg
• Primary Examiner(s): Holmes; Michael B
• Application Number:
  US10/558,623
• Publication Number:
  US 20070022062A1
• Time in Patent Office: 1,707 Days
• Field of Search: None
• US Class Current: 706/15
• CPC Class Codes: G05B 13/027 using neural networks only

Abstract

An approximation is determined for the future system behavior by a similarity analysis using a previously known behavior of the dynamic system, whereupon the future system behavior is determined by using the approximation for the future behavior of the dynamic system as well as a neuronal network structure, especially a causal retro-causal network (causality analysis).

Claims

1. A method for determining a future response of a dynamic system, comprising:

determining an approximation of the future response using a known response of the dynamic system, performing a comparison and using a similarity analysis;

feeding the approximation of the future system response to a neural network as an input quantity;

determining the future response using the approximation of the future response, using the neural network and performing a causality analysis;

producing an output quantity from the neural network representing the future response of the dynamic system; and

displaying the output quantity on a display device using a time axis.

2. The method according to claim 1, wherein

the neural network has first and second linked neural sub-structures,

the first neural sub-structure has a first mapping response that describes a forward response of the dynamic system, and

the second neural sub-structure has a second mapping response that describes a backward response of the dynamic system.

3. The method according to claim 2, wherein

at least one of the first and second neural sub-structures is a neural network developed over one or a plurality of points in time.

4. The method according to claim 1, wherein

the dynamic system responds based on market mechanisms and planning influences.

5. The method according to claim 4, wherein

the dynamic system is a human-controlled system.

6. The method according to claim 4, wherein

the neural network has first and second linked neural sub-structures,

the first neural sub-structure has a first mapping response that describes a forward response of the dynamic system,

the second neural sub-structure has a second mapping response that describes a backward response of the dynamic system, and

the first neural sub-structure takes into consideration an influence of market mechanisms on the dynamic system and/or the second neural sub-structure takes into consideration a planning influence on the dynamic system.

7. The method according to claim 1, wherein

the neural network is a causal-retro-causal network.

8. The method according to claim 1, wherein

the known response of the dynamic system is determined using historic system data.

9. The method according to claim 1, wherein the approximation of the future response of the dynamic system is determined by:

dividing a time required for the known response into segments of a predefined duration,

determining a segment response of the dynamic system for each segment, and

determining the approximation using selected segments with respectively associated segment responses and a calendar effect.

10. The method according to claim 9, wherein

segments are selected that have a relatively large segment response.

11. The method according to claim 9, wherein in determining the approximation of the future response of the dynamic system:

the future response is determined by interpolating between the segment responses of the selected segments,

a mean is determined of the segment responses of the selected segments, and/or

at least one historic segment response is considered together with the segment responses of the selected segments.

12. The method according to claim 1, further comprising using the future response of the dynamic system as a forecast.

13. The method according to claim 1, further comprising using the future response of the dynamic system to forecast energy consumption.

14. The method according to claim 1, further comprising using the future response of the dynamic system to forecast gas fuel consumption.

15. The method according to claim 5, wherein

the neural network has first and second linked neural sub-structures,

the first neural sub-structure has a first mapping response that describes a forward response of the dynamic system,

the second neural sub-structure has a second mapping response that describes a backward response of the dynamic system,

the first neural sub-structure takes into consideration an influence of market mechanisms on the dynamic system, and

the second neural sub-structure takes into consideration a planning influence on the dynamic system.

16. The method according to claim 15, wherein the approximation of the future response of the dynamic system is determined by:

dividing a time required for the known response into segments of a predefined duration,

determining a segment response of the dynamic system for each segment, and

determining the approximation using selected segments with respectively associated segment responses and a calendar effect.

17. The method according to claim 16, wherein in determining the approximation of the future response of the dynamic system:

the future response is determined by interpolating between the segment responses of the selected segments.

18. The method according to claim 16 wherein in determining the approximation of the future response of the dynamic system:

a mean is determined of the segment responses of the selected segments.

19. The method according to claim 16, wherein in determining the approximation of the future response of the dynamic system:

at least one historic segment response is considered together with the segment responses of the selected segments.

20. A computer readable medium storing a program for controlling a computer to perform a method for determining a future response of a dynamic system, the method comprising:

determining an approximation of the future response using a known response of the dynamic system, performing a comparison and using a similarity analysis;

feeding the approximation of the future system response to a neural network as an input quantity;

determining the future response using the approximation of the future response, using the neural network and performing a causality analysis;

producing an output quantity from the neural network representing the future response of the dynamic; and

displaying the output quantity on a display device using a time axis.

21. A method for determining a future response of a dynamic system representing a humans controlled system related to energy consumption, power consumption or gas consumption, the method comprising:

determining an approximation of the future consumption using a known consumption of the dynamic system, performing a comparison and using a similarity analysis;

inputting the approximation of the future consumption to a neural network corresponding to the dynamic system;

determining the future consumption by performing a causality analysis in the neural network, the neural network including a first neural sub-structures yielding a first mapping response, which describes a forward response of the dynamic system taking into consideration an influence of market mechanisms on the dynamic system, and a second neural sub-structure yielding a second mapping response, which describes a backward response of the dynamic system taking into consideration a planning influence on the dynamic system;

outputting the future consumption of the dynamic system from the neural network in response to input of the approximation of the future consumption; and

displaying the future consumption on a display device using a time axis.