System and method for modelling and forecasting electricity demand

1. A method for modeling and forecasting electricity demand in a power generation and distribution network which includes a plurality of nodes each having cooling appliances, heating appliances, or both cooling and heating appliances, the method comprising:

building, using intraday temperature and electricity demand data, a per-capita parametric model of high-level features of an electric demand curve as a function of temperature and time for a geographic region in the power generation and distribution network, the electric demand curve comprising the high-level features and residual features, the high-level features comprising underlying base load and heating or cooling requirements;

modeling, using a residual parametric model of the residual features of the electric demand curve, residual intraday variations in electric demand with respect to intraday variations in temperature from a daily temperature mean;

determining, using the intraday electricity demand data and modeled high-level features of the electric demand curve, average residual variations in electric demand with respect to intraday time periods;

iteratively optimizing, by an electricity demand modeling device, a residual error between the modeled residual intraday variations in electric demand and the determined average residual variations in electric demand by updating parameters in the residual parametric model until the residual error is below a predetermined threshold;

repeating, by the electricity demand modeling device using an optimization method, the building, modeling, determining, and iteratively optimizing in a recursive fashion to optimize parameters of the per-capita and residual parametric models;

determining, using the optimized parametric models, electricity demand per-capita for the geographic region; and

applying the determined electricity demand per-capita for the geographic region to modify one or more electricity transmission parameters in a power generation station in order to adjust electric power distributed in the power generation and distribution network.

2. The method in claim 1 , further comprising determining the residual error based on at least one of:

a cooling demand function COOL(T,T avg );

a heating demand function HEAT(T,T avg );

an intraday temperature variation function ITV(T,T avg );

a humidity function HUM(H,H avg ,T avg );

a vacation demand function VAC(T,H,H avg ); and

a calendar effect function CAL t ,

where T is population-weighted temperature, T avg is daily average temperature, H is population-weighted humidity, and H avg is daily average humidity.

3. The method in claim 1 , further comprising modeling effects of humidity on the electric demand curve.

4. The method in claim 1 , further comprising modeling effects of temperature on the electric demand curve.

5. The method in claim 1 , further comprising:

determining the average residual variations in electric demand throughout a lunar calendar year or solar calendar year.

6. The method in claim 1 , further comprising:

determining a distinguishable effect in the residual intraday variations in electric demand;

generating a plurality of demand models; and

prioritizing the plurality of demand models based on a magnitude of contribution to residual demand.

7. The method in claim 6 , wherein generating the plurality of demand models comprises parametrically modelling the distinguishable effect in the residual intraday variations in electric demand to build at least one of the plurality of demand models.

8. The method in claim 1 , further comprising:

applying a seasonal variation function to the residual intraday variations in electric demand.

9. The method in claim 1 , further comprising:

transmitting the electricity demand per-capita for the geographic region to a communicating device.

10. The method in claim 1 , further comprising:

estimating an evolution of one or more parameters in the per-capita and residual parametric models over time.

11. The method in claim 10 , wherein the estimating the evolution of one or more parameters in the per-capita and residual parametric models comprises sliding a window of a fixed period length to optimize a value of a parameter of interest while keeping remaining parameters fixed.

12. The method in claim 1 , further comprising:

supplying the adjusted electric power to a node in the geographic region.

13. A system for modeling and forecasting electricity demand in a power generation and distribution network that includes a plurality of nodes having cooling appliances or heating appliances, the system comprising:

one or more computing devices, comprising:

an electricity demand modeling device configured to

build, using intraday temperature and electricity demand data, a per-capita parametric model of high-level features of an electric demand curve as a function of temperature and time for a geographic region in the power generation and distribution network, the electric demand curve comprising the high-level features and residual features, the high-level features comprising underlying base load and heating or cooling requirements;

model, using a residual parametric model of the residual features of the electric demand curve, residual intraday variations in electric demand with respect to intraday variations in temperature from a daily temperature mean;

determine, using the intraday electricity demand data and modeled high-level features of the electric demand curve, average residual variations in electric demand with respect to intraday time periods;

iteratively optimize a residual error between the modeled residual intraday variations in electric demand and the determined average residual variations in electric demand by updating parameters in the residual parametric model until the residual error is below a predetermined threshold; and

repeat, using an optimization method, the building, modeling, determining, and iteratively optimizing in a recursive fashion to optimize parameters of the per-capita and residual parametric models; and

an electricity demand forecasting device that determines, using the optimized parametric models, electricity demand per-capita for the geographic region and transmits the determined electricity demand per-capita model for the geographic region to a communicating device in the power generation and distribution network,

wherein the communicating device is configured to apply the transmitted electricity demand per-capita for the geographic region in order to modify one or more electricity transmission parameters in a power generation station and to adjust electric power distributed in the power generation and distribution network.

14. The system in claim 13 , wherein the electricity demand modeler modeling device is further configured to determine the residual error based on at least one of:

a cooling demand function COOL(T,T avg );

a heating demand function HEAT(T,T avg );

an intraday temperature variation function ITV(T,T avg );

a humidity function HUM(H,H avg ,T avg );

a vacation demand function VAC(T,H,H avg ); and

a calendar effect function CAL t ,

where T is population-weighted temperature, T avg is daily average temperature, H is population-weighted humidity, and H avg is daily average humidity.

15. The system in claim 13 , wherein the electricity demand modeling device is further configured to model effects of humidity on the electric demand curve.

16. The system in claim 13 , wherein the electricity demand modeling device is further configured to model effects of temperature on the electric demand curve.

17. The system in claim 13 , wherein the electricity demand modeling device is further configured to determine the average residual variations in electric demand throughout a lunar calendar year or solar calendar year.

18. The system in claim 13 , wherein the electricity demand modeler modeling device is further configured to:

determine a distinguishable effect in the residual intraday variations in electric demand;

generate a plurality of demand models; and

prioritize the plurality of demand models based on a magnitude of contribution to residual demand.

19. The system in claim 18 , wherein the electricity demand modeling device is further configured to generate the plurality of demand models by parametrically modelling the distinguishable effect in the residual intraday variations in electric demand to build at least one of the plurality of demand models.

20. The system in claim 13 , wherein the electricity demand modeling device is further configured to apply a seasonal variation function to the residual intraday variations in electric demand.

21. A non-transitory computer readable storage medium storing electricity demand modeling and forecasting instructions for causing per-capita electricity demand for a geographic region in a power generation and distribution network that includes a plurality of nodes having cooling appliances or heating appliances to be modeled and forecasted, the program instructions comprising the steps of:

building, using intraday temperature and electricity demand data, a per-capita parametric model of high-level features of an electric demand curve as a function of temperature and time for a geographic region in the power generation and distribution network, the electric demand curve comprising the high-level features and residual features, the high-level features comprising underlying base load and heating or cooling requirements;

modeling, using a residual parametric model of the residual features of the electric demand curve, residual intraday variations in electric demand with respect to intraday variations in temperature from a daily temperature mean;

determining, using the intraday electricity demand data and modeled high-level features of the electric demand curve, average residual variations in electric demand with respect to intraday time periods;

iteratively optimizing a residual error between the modeled residual intraday variations in electric demand and the determined average residual variations in electric demand by updating parameters in the residual parametric model until the residual error is below a predetermined threshold;

repeating, using an optimization method, the building, modeling, determining, and iteratively optimizing in a recursive fashion to optimize parameters of the per-capita and residual parametric models;

determining, using the optimized parametric models, electricity demand per-capita for the geographic region; and

applying the determined electricity demand per-capita for the geographic region to modify one or more electricity transmission parameters in a power generation station in order to adjust electric power distributed in the power generation and distribution network.

22. The non-transitory computer readable storage medium in claim 21 , the program instructions comprising the further step of determining the residual error based on at least one of:

a cooling demand function COOL(T,T avg );

a heating demand function HEAT(T,T avg );

an intraday temperature variation function ITV(T,T avg );

a humidity function HUM(H,H avg ,T avg );

a vacation demand function VAC(T,H,H avg ); and

a calendar effect function CAL t ,

where T is population-weighted temperature, T avg is daily average temperature, H is population-weighted humidity, and H avg is daily average humidity.

23. The non-transitory computer readable storage medium in claim 21 , the program instructions comprising the further step of:

modeling effects of temperature or humidity on the electric demand curve.

24. The non-transitory computer readable storage medium in claim 21 , the program instructions comprising the further step of:

supplying the adjusted electric power to a node in the geographic region.