Very short term load prediction in an energy management system

• US 7,305,282 B2
• Filed: 05/12/2004
• Issued: 12/04/2007
• Est. Priority Date: 05/13/2003
• Status: Expired due to Fees

• Current Assignee: Siemens AG
• Original Assignee: Siemens Power Transmission & Distribution Incorporated (Siemens AG)
• Inventors: Chen, Dingguo
• Primary Examiner(s): Picard; Leo
• Assistant Examiner(s): Rao; Sheela S.
• Application Number:
  US10/844,137
• Publication Number:
  US 20040249775A1
• Time in Patent Office: 1,301 Days
• Field of Search: 700/286, 700/291, 700/295, 702/60, 706/21, 706/45, 361/62, 361/115, 365/154
• US Class Current: 700/291
• CPC Class Codes: G06Q 50/06 Energy or water supply
• CPC Class Codes: H02J 2203/20 Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
• CPC Class Codes: H02J 3/00 Circuit arrangements for ac mains or ac distribution networks
• CPC Class Codes: H02J 3/003 Load forecast, e.g. methods or systems for forecasting future load demand
• CPC Class Codes: H02J 3/008 involving trading of energy or energy transmission rights
• CPC Class Codes: Y02B 70/3225 Demand response systems, e.g. load shedding, peak shaving
• CPC Class Codes: Y02E 60/00 Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
• CPC Class Codes: Y04S 10/50 Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
• CPC Class Codes: Y04S 20/222 Demand response systems, e.g. load shedding, peak shaving
• CPC Class Codes: Y04S 40/20 Information technology specific aspects, e.g. CAD, simulation, modelling, system security
• CPC Class Codes: Y04S 50/10 Energy trading, including energy flowing from end-user application to grid

Abstract

A method and system is provided for optimizing the performance of a power generation and distribution system by forecasting very short term load forecasts through the use of historical load data, demand patterns and short term load forecasts.

Claims

1. A method of energy management and short term load prediction in a power system, comprising the steps of:

dividing a time period into a plurality of sub-periods;

using a plurality of neural networks, each of the plurality of neural networks comprising a plurality of neurons to predict a short term load demand on the power system, the sub-periods being further divided into predicting intervals and assigning one of the plurality of neural networks to each of the plurality of sub-periods;

receiving historical load data and a short term load forecast (STLF) for a STLF interval which is greater than or equal to the largest prediction interval; and

using the output of each of the plurality of neural networks to predict the short term load demand in each predicting interval.

2. The method of claim 1, further comprising the step of using a decision algorithm module to process the plurality of outputs from each of the plurality of neural networks.

3. The method of claim 1, further comprising the step of weighting the inputs between a first neuron and a second neuron within at least one of the plurality of neural networks.

4. The method of claim 1, further comprising the step of overlapping a first sub-period and a second and adjacent sub-period.

5. The method of claim 4, wherein in the step of overlapping comprises overlapping the first sub-period with the second and adjacent sub-period by at least 30 minutes.

6. The method of claim 1 further comprising the step of overlapping a sub- period at a first end with a first sub-period and a second end with a second and adjacent sub-period.

7. The method of claim 6, wherein the overlapping of the sub-period with the first sub-period and the second and adjacent sub-period by at least 30 minutes.

8. The method of claim 4, wherein in the step of overlapping comprises overlapping the first sub-period with the second and adjacent sub-period by at least 15 minutes.

9. The method of claim 1 further comprising the step of overlapping a sub- period at a first end with a first sub-period and a second end with a second and adjacent sub-period.

10. The method of claim 9, wherein the overlapping of the sub-period with the first sub-period and the second and adjacent sub-period by at least 15 minutes.

11. The method of claim 5, wherein the prediction interval is at least one minute.

12. The method of claim 1, further comprising the step of conforming the prediction interval predictions with the short term load forecast, the short load forecast covering a period of less than or equal to 1 hour.

13. The method of claim 12, wherein the step of conforming to the short term load forecast requires that the total of the summation of all actual load values for a time prior to or equal to an instant in time within a current sub-period and the summation of the product of a scaling factor and predicted load values for the time remaining within the sub-period equals the value of the forecasted short term load forecast.

14. The method of claim 13, wherein the scaling factor is dynamically time varying between prediction intervals.

15. The method of claim 14, wherein the scaling factor is substantial equal to one.

16. The method of claim 15, wherein the scaling factor changes as a function of immediately past actual load values and forecasted load values for each of the prediction intervals.

17. The method of claim 16, wherein the past actual load values exclude load values of non-conforming loads.

18. A computer-readable medium having stored thereon instructions which when executed by a processor, cause the processor to perform the steps of:

dividing a time period into a plurality of sub-periods;

using a plurality of neural networks, each of the plurality of neural networks comprising a plurality of neurons to predict a short term load demand on the power system, the sub-periods being further divided into predicting intervals and assigning one of the plurality of neural networks to each of the plurality of sub-periods;

receiving historical load data and a short term load forecast (STLF) for a STLF interval which is greater than or equal to the largest prediction interval; and

using the output of each of the plurality of neural networks to predict the short term load demand in each predicting interval.

19. A system for predicting short term loads within an energy management system of a power system, comprising:

a processor for dividing a time period into a plurality of sub-periods and prediction intervals; and

a plurality of neural networks, each of the plurality of neural networks comprising a plurality of neurons to predict a short term load demand on the power system, the sub-periods being further divided into predicting intervals and assigning one of the plurality of neural networks to each of the plurality of sub- periods;

the processor, receiving and processing historical load data and a short term load forecast (STLF) for a STLF interval which is greater than or equal to the largest prediction interval; and

using the output of each of the plurality of neural networks to predict the short term load demand in each predicting interval.

20. The system of claim 19, further comprising a decision algorithm module in communication with the processor to process the plurality of outputs from each of the plurality of neural networks.

21. The system of claim 19, wherein at least one neural network is operative for weighting the inputs between a first neuron and a second neuron within at least one of the plurality of neural networks.

22. The system of claim 19, wherein the processor is operative for overlapping a first sub-period and a second and adjacent sub-period.

23. The system of claim 22, wherein overlapping comprises overlapping the first sub-period with the second and adjacent sub-period by at least 30 minutes.

24. The system of claim 19 wherein the processor is operative for overlapping, and overlapping comprises overlapping a sub-period at a first end with a first sub-period and a second end with a second and adjacent sub-period.

25. The system of claim 24, wherein overlapping comprises overlapping of the sub-period with the first sub-period and the second and adjacent sub-period by at least 30 minutes.

26. The system of claim 22, wherein overlapping comprises overlapping the first sub-period with the second and adjacent sub-period by at least 15 minutes.

27. The system of claim 19 further wherein the processor is operative for overlapping and overlapping comprises overlapping a sub-period at a first end with a first sub-period and a second end with a second and adjacent sub-period.

28. The system of claim 27, wherein the overlapping of the sub-period with the first sub-period and the second and adjacent sub-period by at least 15 minutes.

29. The system of claim 23, wherein the prediction interval is at least one minute.

30. The system of claim 19, further comprising the step of conforming the prediction interval predictions with the short term load forecast, the short load forecast covering a period of less than or equal to 1 hour.

31. The system of claim 30, wherein the processor is operative for conforming to the short term load forecast requires that the total of the summation of all actual load values for a time prior to or equal to an instant in time within a current sub-period and the summation of the product of a scaling factor and predicted load values for the time remaining within the sub-period equals the value of the forecasted short term load forecast.

32. The system of claim 31, wherein the scaling factor is dynamically time varying between prediction intervals.

33. The system of claim 32, wherein the scaling factor is substantial equal to one.

34. The system of claim 33, wherein the scaling factor changes as a function of immediately past actual load values and forecasted load values for each of the prediction intervals.

35. The system of claim 34, wherein the past actual load values exclude load values of non-conforming loads.