System and method for determining harmonic contributions from non-linear loads

• US 20050057265A1
• Filed: 09/14/2004
• Published: 03/17/2005
• Est. Priority Date: 09/15/2003
• Status: Active Grant

• Current Assignee: Georgia Tech Research Corporation (University System of Georgia)
• Original Assignee: Georgia Tech Research Corporation (University System of Georgia)
• Inventors: Mazumdar, Joy, Lambert, Franklin Cook, Harley, Ronald Gordon, Habelter, Thomas G.
• Granted Patent: US 7,013,227 B2
• Time in Patent Office: Days
• US Class Current: 324/658
• CPC Class Codes: H02M 1/12 Arrangements for reducing harmonics from ac input or output
• CPC Class Codes: Y10S 706/915 Power plant

Abstract

A system and method for determining harmonics caused by non-linear loads are disclosed. Briefly described, one embodiment is a method comprising metering voltage on an electric power system; metering current on the electric power system; determining a predicted current based upon the metered voltage; comparing the predicted current with the metered current; and determining a harmonic current component using a plurality of weights determined when the predicted current converges with the metered current.

Claims

1. A method for determining harmonics caused by non-linear loads, comprising:

metering voltage on an electric power system;

metering current on the electric power system;

determining a predicted current based upon the metered voltage;

comparing the predicted current with the metered current; and

determining a harmonic current component using a plurality of weights determined when the predicted current converges with the metered current.

2. The method of claim 1, further comprising determining the weights when the predicted current converges within a predefined threshold of the metered current.

3. The method of claim 1, further comprising determining when the predicted current converges with the metered current by comparing the predicted current with the metered current.

4. The method of claim 1, further comprising:

metering single phase voltage; and

metering single phase current.

5. The method of claim 1, further comprising:

metering three phase voltage; and

metering three phase current.

6. The method of claim 1, further comprising computationally combining a sinusoidal waveform with the weights to determine the harmonic current component.

7. The method of claim 1, further comprising:

communicating the determined weights to an estimating artificial neural network when the predicted current converges with the metered current;

again determining predicted current based upon the metered voltage;

again comparing the predicted current with metered current;

determining another plurality of weights; and

determining a second harmonic current component using the other plurality of weights.

8. A system that determines harmonics caused by non-linear loads, comprising:

an artificial neural network configured to receive information corresponding to metered voltage, to train itself such that a predicted current is determined which converges with a metered current, and to determine a harmonic current component using a plurality of weights determined when the predicted current converges with the metered current; and

a processor configured to execute the artificial neural network.

9. The system of claim 8, wherein the determined predicted current converges with the metered current within a predefined threshold.

10. The system of claim 8, further comprising a memory wherein the artificial neural network resides.

11. The system of claim 8, further comprising summing circuit that compares the determined predicted current and the metered current, and that outputs a current error to the artificial neural network such that when the current error is less than a predefined threshold, the artificial neural network has determined the predicted current.

12. The system of claim 11, wherein the current error corresponds to a difference between the metered current and the predicted current.

13. The system of claim 11, further comprising a current transducer that receives a signal corresponding to current from a current transformer, and that outputs a second signal corresponding to the metered current, the second signal communicated to the summing circuit as the metered current.

14. The system of claim 8, further comprising a voltage transducer that receives a signal corresponding to voltage from a potential transformer, and that outputs a second signal corresponding to the metered voltage, the second signal communicated to the artificial neural network as the metered voltage.

15. The system of claim 8, wherein the artificial neural network further comprises:

a training artificial neural network that determines the predicted current from the metered voltage and that determines the weights determined when the predicted current converges with the metered current; and

an estimating artificial neural network that determines the harmonic current component using the determined weights received from the training artificial neural network.

16. The system of claim 15, further comprising a sinusoidal wave that is computationally combined with the weights to determine the harmonic current component.

17. A system for determining harmonics caused by non-linear loads, comprising:

means for receiving metered voltage from an electric power system;

means for receiving metered current from the electric power system;

means for determining a predicted current based upon the metered voltage;

means for comparing the predicted current with the metered current; and

means for determining a harmonic current component a plurality of weights determined when the predicted current converges with the metered current.

18. The system of claim 17, further comprising means for determining the weights when the predicted current converges with the metered current

19. The system of claim 17, further comprising means for determining when the predicted current converges with the metered current by comparing the predicted current with the metered current.

20. The system of claim 17, further comprising:

means for metering single phase voltage; and

means for metering single phase current.

21. The system of claim 17, further comprising:

means for metering three phase voltage; and

means for metering three phase current.

22. A program for determining harmonics caused by non-linear loads stored on computer-readable medium, the program comprising:

logic configured to receive information corresponding to a metered voltage from an electric power system;

logic configured to receive information corresponding to a metered current on the electric power system;

logic configured to determine a predicted current based upon the metered voltage; and

logic configured to determining a harmonic current component using a plurality of weights determined when the predicted current converges with the metered current

23. The program of claim 22, further comprising logic configured to compare the predicted current with the metered current.

24. The program of claim 22, further comprising logic configured to determining the weights when the predicted current converges with the metered current

25. The program of claim 22, further comprising logic configured to determine when the predicted current converges with the metered current by comparing the predicted current with the metered current.

26. An artificial neural network that determines harmonics caused by non-linear loads, comprising:

a training artificial neural network that determines a predicted current from a metered voltage, and that determines a plurality of weights when the predicted current converges with a metered current; and

an estimating artificial neural network that determines a harmonic current component using the determined weights received from the training artificial neural network.

27. The artificial neural network of claim 26, wherein the training artificial neural network trains itself such that the predicted current is determined which converges with the metered current.

28. The artificial neural network of claim 27, wherein the training artificial neural network compares the predicted current with the metered current to determine when the predicted current converges with the metered current.