Induced Markov chain for wind farm generation forecasting
View Patent ↗Systems and methods for forecasting power generation in a wind farm are disclosed. The systems and methods utilize an induced Markov chain model to generate a forecast of power generation of the wind farm. The forecast is at least one of a point forecast or a distributional forecast. Additionally, the systems and methods modify at least one of: (i) a generation of electricity at a power plant coupled to a common power grid as the wind farm; or (ii) a distribution of electricity in the common power grid based on the forecast of power generation of the wind farm. In an exemplary approach, utilizing the induced Markov chain model to generate the forecast may include determining a series of time adjacent power output measurements based on historical wind power measurements and calculating a time series of difference values based on the series of time adjacent power output measurements.
1. A method for forecasting power generation in a wind farm, the method comprising:
utilizing, by a processor, an induced Markov chain model to generate a forecast of power generation of the wind farm, wherein the forecast is at least one of a point forecast or a distributional forecast; and
modifying at least one of: (i) a generation of electricity at a power plant coupled to a common power grid as the wind farm; or (ii) a distribution of electricity in the common power grid based on the forecast of power generation of the wind farm, wherein utilizing the induced Markov chain model to generate the forecast of the power generation of the wind farm comprises:
determining a series of time adjacent power output measurements based on historical wind power measurements of the wind farm;
transforming time adjacent power output measurements into discrete states, the discrete states comprising ranges of power, the transforming comprising determining at least one discrete state for each time adjacent power output measurement, wherein the discrete states comprise at least one overlapping state, the overlapping state having a first range of power overlapping with a second range of power of another state; and
calculating a time series of difference values based on the series of time adjacent power output measurements including calculating a difference value between adjacent power output measurements of the series of time adjacent power output measurements.
2. The method of claim 1 , wherein determining the series of time adjacent power output measurements and calculating the time series of difference values occurs before forecasting begins.
3. The method of claim 1 , further comprising providing the forecast to at least one of an electric utility or a customer of the electric utility.
4. The method of claim 1 , wherein the modifying the generation of electricity results in reduced greenhouse gas emissions associated with the generation of electricity.
5. The method of claim 1 , wherein the modifying the generation of electricity results in decreased costs associated with the generation of electricity.
6. The method of claim 1 , wherein the forecast of the power generation of the wind farm predicts power output of the wind farm for a period of between 5 seconds and 6 hours into the future.
7. A device for forecasting power generation in a wind farm, the device comprising a processor configured to be in electrical communication with a wind farm power output sensor, wherein the processor is configured to:
utilize an induced Markov chain model to generate a forecast of the power generation of the wind farm, wherein the forecast is at least one of a point forecast or a distributional forecast; and
modify at least one of: (i) a generation of electricity at a power plant coupled to a common power grid as the wind farm; or (ii) a distribution of electricity in the common power grid based on the forecast of the power generation of the wind farm,
wherein utilizing the induced Markov chain model to generate the forecast of the power generation of the wind farm comprises:
determining a series of time adjacent power output measurements based on historical wind power measurements of the wind farm; and
calculating a time series of difference values based on the series of time adjacent power output measurements including calculating a difference value between adjacent power output measurements of the series of time adjacent power output measurements,
wherein the processor is further configured to transform time adjacent power output measurements into discrete states, the discrete states comprising ranges of power, the transforming comprising determining at least one discrete state for each of the time adjacent power output measurements, and wherein the discrete states comprise at least one overlapping state, the overlapping state having a first range of power overlapping with a second range of power of another state.
8. The device of claim 7 , wherein determining the series of time adjacent power output measurements and calculating the time series of difference values occurs before forecasting begins.
9. The device of claim 7 , wherein the processor is further configured to provide the forecast to at least one of an electric utility or a customer of the electric utility.
10. The device of claim 7 , wherein the modifying the generation of electricity results in reduced greenhouse gas emissions associated with the generation of electricity.
11. The device of claim 7 , wherein the modifying the generation of electricity results in decreased costs associated with the generation of electricity.
12. The device of claim 7 , wherein the forecast of the power generation of the wind farm predicts power output of the wind farm for a period of between 5 seconds and 6 hours into the future.
13. A system for forecasting power generation in a wind farm, the system comprising:
a wind farm power output sensor; and
a processor configured to be in electrical communication with the wind farm power output sensor, wherein the processor is configured to:
utilize an induced Markov chain model to generate a forecast of the power generation of the wind farm, wherein the forecast is at least one of a point forecast or a distributional forecast; and
modify at least one of: (i) a generation of electricity at a power plant coupled to a common power grid as the wind farm; or (ii) a distribution of electricity in the common power grid based on the forecast of the power generation of the wind farm, wherein utilizing the induced Markov chain model to generate the forecast of the power generation of the wind farm comprises:
determining a series of time adjacent power output measurements based on historical wind power measurements of the wind farm; and
calculating a time series of difference values based on the series of time adjacent power output measurements including calculating a difference value between adjacent power output measurements of the series of time adjacent power output measurements,
wherein the processor is further configured to transform time adjacent power output measurements into discrete states, the discrete states comprising ranges of power, the transforming comprising determining at least one discrete state for each of the time adjacent power output measurements, and wherein the discrete states comprise at least one overlapping state, the overlapping state having a first range of power overlapping with a second range of power of another state.