SYSTEMS AND METHODS FOR IMPROVING THE ACCURACY OF DAY-AHEAD LOAD FORECASTS ON AN ELECTRIC UTILITY GRID

• US 20130096983A1
• Filed: 04/04/2012
• Published: 04/18/2013
• Est. Priority Date: 04/04/2011
• Status: Active Grant

• Current Assignee: Catholic University of America (Catholic University)
• Original Assignee: Catholic University of America (Catholic University)
• Inventors: St. Cyr, O. Chris S., Zampelli, Ernest M., Forbes, Kevin F.
• Granted Patent: US 10,127,568 B2
• Time in Patent Office: Days
• US Class Current: 705/7.31
• CPC Class Codes: G06Q 10/04 Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" market predictions or forecasting for commercial activities G06Q30/0202
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 50/06 Energy or water supply
• 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 50/14 Marketing, i.e. market research and analysis, surveying, promotions, advertising, buyer profiling, customer management or rewards

Abstract

Systems and methods improve the forecast of electricity consumption, and/or refining such predictions. Predictions may be refined by accounting for factors such as preliminary predictions, pricing and cost information associated with future supply of energy, the extent of anticipated changes in the predictions, the time of day and/or anticipated daylight for the period of time. Coefficient values are calculated for a forecast error model that takes into account factors related to electricity consumption using existing historical electrical grid data. Using the calculated values, the forecast error model may be applied to current electricity demand forecasts.

Claims

1. A method of adjusting forecast load predictions on an electric grid using one or more operatively connected computers, the method comprising:

obtaining, at the one or more computers, electrical grid information comprising;

a day-ahead profile of a forecasted load for the electric grid,

a day-ahead sparks ratio,

a day-ahead price profile, and

an hourly forecasted load relative to the forecasted profile;

accessing, one or more databases operatively connected to the one or computers, electrical grid historical data;

calculating, by the one or more computers, coefficients for a forecast error equation by performing a regression analysis using the historical data, the forecast equation being a function of day-ahead sparks ratio, day-ahead price profile, day-ahead profile of forecasted load, and hourly forecasted load relative to the forecasted profile;

calculating, by the one or more computers, forecast prediction errors by applying data from the obtained electrical grid information to the forecast error equation with the calculated coefficients; and

providing, by the one or more computers, one or more electronic reports containing forecast errors for the electric grid.

2. The method of claim 1, the method further comprising:

calculating, by the one or more computers, one or more residual terms as the differences between i) predicted errors from the historical data using the forecast error equation, and ii) actual errors from the historical data;

applying, by the one or more computers, the residual terms to an auto-regressive moving average analysis, so as to determine one or more disturbances at one or more time times; and

calculating, by the one or more computers, a new set of coefficients for the forecast error equation based on the auto-regressive moving average analysis.

3. The method of claim 1, wherein the forecast error equation is a multivariable fractional polynomial model.

4. The method of claim 1, wherein the forecast error equation comprises a Coefficient of Variation variable calculated over a day, using one or more computers, as the ratio of standard deviation of the day-ahead prices to the average hourly price.

5. The method of claim 1, wherein the forecast error equation comprises a Positive Skewness variable calculated over a day, using one or more computers, as values equal to the absolute value of the skewness in the day ahead prices.

6. The method of claim 1, wherein the forecast error equation comprises a Negative Skewness variable calculated over a day, using one or more computers, as values equal to the absolute value of the skewness in the day ahead prices.

7. The method of claim 1, wherein the forecast error equation comprises a Forecasted Load Peak Ratio variable calculated over a day, using one or more computers, as the ratio of the forecasted load to the forecasted peak load.

8. The method of claim 1, wherein the forecast error equation comprises a Forecasted Load Nadir Ratio variable calculated over a day, using one or more computers, as the ratio of the forecasted load to the forecasted minimum load.

9. The method of claim 1, wherein the forecast error equation comprises a Forecasted Load Peak variable equal to values of the forecasted peak hourly load for a day.

10. The method of claim 1, wherein the forecast error equation comprises a Forecasted Load Nadir variable equal to values of the forecasted minimum hourly load for a day.

11. The method of claim 1, wherein the forecast error equation comprises a Coefficient of Variation Forecasted Load Variable calculated over a day, using one or more computers, as the ratio of standard deviation of the day prices to the average hourly price.

12. The method of claim 1, wherein the forecast error equation comprises a Positive Skewness Forecasted Load Variable calculated, using one or more computers, as equal to the skewness in the day-ahead forecasted hourly load when the skewness is positive.

13. The method of claim 1, wherein the forecast error equation comprises a Negative Skewness Forecasted Load variable calculated, using one or more computers, as equal the absolute skewness in the day-ahead forecasted hourly load when the skewness is negative.

14. The method of claim 1, wherein the forecast error equation comprises a Forecasted Delta Load variable calculated over a day, using one or more computers, as equal to the ratio of a forecasted load at a first time to a forecasted load at previous time.

15. The method of claim 14, wherein the forecast error equation comprises a Geo Sigma Forecasted Load variable calculated over a day, using one or more computers, as the geometric standard deviation of the Forecasted Delta Load variable.

16. The method of claim 1, wherein the forecast error equation comprises a Peak Sparks Ratio variable equal to the ratio of the highest day-ahead price of electricity relative to the price of a fuel used to generate electricity.

17. The method of claim 1, wherein the forecast error equation comprises an Average Peak Forecasted Load variable calculated as, using one or more computers, as the average value of the forecasted load for the day relative to the peak level of the forecasted load for the day.

18. The method of claim 1, wherein the forecast error equation comprises a day-ahead hourly volume of energy traded variable.

19. The method of claim 1, wherein the forecast error equation comprises a mean daily volume of energy traded variable.

20. The method of claim 1, wherein the forecast error equation comprises a coefficient of variation in the daily volume of energy traded variable.

21. The method of claim 1, wherein the forecast error equation comprises a skewness in the volume of energy traded variable.

22. The method of claim 1, wherein the sparks ratio is the day-ahead electricity prices relative to the prices of the fuel used to generate electricity.

23. The method of claim 22, wherein the fuel is natural gas.

24. The method of claim 22, wherein the fuel is coal.

25. The method of claim 22, wherein the fuel is oil.