Modeling and calculating normalized aggregate power of renewable energy source stations

Patent metadata

Field	Value
Publication number	US-11204591-B2
Application number	US-201715815790-A
Country	US
Kind code	B2
Filing date	Nov 17, 2017
Priority date	Nov 17, 2017
Publication date	Dec 21, 2021
Grant date	Dec 21, 2021

Abstract

The present invention provides a method, system, and computer program product of modeling and calculating aggregate power of a set of renewable energy source stations using power output from representative renewable energy source stations. In an embodiment, the present invention includes receiving location, power output time series, and weather time series data of renewable energy source stations in a geographic region and aggregate power output time series data for the geographic region, for each cluster of stations, normalizing the aggregate power value to a representative renewable energy source station, learning a regression model, and de-normalizing a normalized aggregate output power model with respect to a maximum possible power value, and applying a combined model to the received data and power output of representative renewable energy source stations for a particular day, resulting in a total aggregate power value of the renewable energy source stations for the particular day.

First claim

What is claimed is:

1. A computer implemented method comprising:
receiving, by a computer system, location data of a plurality of renewable energy source stations in a geographic region, power output time series data of each of the plurality of renewable energy source stations, weather time series data of weather parameters corresponding to the plurality of renewable energy source stations, and aggregate power output time series data for the geographic region;
clustering, by the computer system, the plurality of renewable energy source stations according to a clustering metric, resulting in clusters of renewable energy source stations;
building, by the computer system, a machine learning model based on one or more historical relationships between the location data, the power output time series data, the weather time series data, and the aggregate power output time series data,
wherein the building comprises deriving one or more algorithms capable of making data-driven decisions to select a representative renewable energy source station through predictive analytics;

for each of the clusters, designating, by the machine learning model in the computer system, a renewable energy source station among the plurality of renewable energy source stations to be the representative renewable energy source station of the each of the clusters based on at least an aggregate power value among the aggregate power output time series data, wherein the machine learning model designates the representative renewable energy source station based on a prediction that a future power value for the representative renewable energy source station will be the best predictor for a future estimation;
for the each of the clusters, determining, by the computer system, a set of features corresponding to the representative renewable energy source station of the each of the clusters, resulting in a determined set of features of the each of the clusters;
for the each of the clusters, executing, by the computer system, a set of logical operations normalizing the aggregate power value based on the power output from the representative renewable energy source station of the each of the clusters with respect to a maximum of aggregate power values in the aggregate power output time series data, resulting in a normalized aggregate output power value of the representative renewable energy source station of the each of the clusters;
for the each of the clusters, executing, by the computer system, a set of logical operations for learning a regression model of the each of the clusters based on the determined set of features of the each of the clusters and the normalized aggregate output power value of the representative renewable energy source station of the each of the clusters, resulting in learned regression models of the clusters of renewable energy source stations;
for the each of the clusters, executing, by the computer system, a set of logical operations de-normalizing a normalized aggregate output power model of the each of the clusters, among the learned regression models, with respect to a maximum possible power value, resulting in a de-normalized aggregate output power model of the each of the clusters;
executing, by the computer system, a set of logical operations combining the learned regression models of the clusters of renewable energy source stations with respect to the de-normalized aggregate output power model of the each of the clusters, resulting in a combined model for aggregate power of the plurality of renewable energy source stations;
executing, by the computer system, a set of logical operations applying the combined model to the location data, the power output time series data, the weather time series data, and power output of representative renewable energy source stations for a particular day, among the power output time series data, resulting in a predicted total aggregate power production value of the plurality of renewable energy source stations for the particular day, and
applying the predicted total aggregate power production value of the plurality of renewable energy source stations to integrate the power output of the plurality of renewable energy source stations into an electrical grid for the particular day.

2. The method of claim 1 wherein the location data describe locations of the plurality of renewable energy source stations.

3. The method of claim 1 wherein the clustering comprises choosing the clustering metric based on the location data, the power output time series data, and the weather time series data.

4. The method of claim 1 wherein the designating comprises designating the representative renewable energy source station based on representative metrics including at least a center of the corresponding cluster and a correlation value indicating a correlation of the renewable energy source station to the aggregate power value.

5. The method of claim 1 wherein the set of features comprises a power output value, among the power output time series data, of the representative renewable energy source station, weather data, among the weather time series data, describing weather at the representative renewable energy source station, and time data at the representative renewable energy source station.

6. The method of claim 1 wherein the normalizing comprises calculating, by the computer system, a mapping from a power output value, among the power output time series data, of the representative renewable energy source station of the each of the clusters and the determined set of features of the each of the clusters, to the aggregate power value.

7. The method of claim 6 wherein the normalizing comprising executing, by the computer system, a set of logical operations normalizing the aggregate power value to the maximum possible power value.

8. The method of claim 7 wherein the normalizing the aggregate power value to the maximum possible power value comprises executing, by the computer system, a set of logical operations normalizing the aggregate power output time series data by dividing each power value P t , among the aggregate power output time series data, at time t, by the maximum possible power value, Pmax,t , at time t.

9. The method of claim 7 wherein the plurality of renewable energy source stations comprises a plurality of photovoltaic stations.

10. The method of claim 9 further comprising:
in response to detecting by the computer system a lower aggregate power value of the plurality of photovoltaic stations, executing, by the computer system, a set of logical operations scaling the lower aggregate power value to a scaled power value on a clear-sky day based on the weather time series data of weather parameters of representative photovoltaic stations among the plurality of photovoltaic stations; and
using, by the computer system, the scaled power value as the aggregate power value in the normalizing the aggregate power value to the maximum possible power value.

11. The method of claim 9 further comprising detecting, by the computer system, at least one change to an installed capacity of the plurality of photovoltaic stations.

12. The method of claim 11 wherein the detecting comprises:
receiving, by the computer system, solar irradiance yield data of correlated cluster regions of the plurality of photovoltaic stations;
for each of the correlated cluster regions, calculating, by the computer system, a mapping from the solar irradiance yield data, E, to an actual aggregate power value, P t,act , in a sliding window, W, with a coefficient, c(t), at time t; and
in response to determining that a regression coefficient differential, |c(t)-c(W)|, of the sliding window, W, ex

Assignees

• IBM

Inventors

• Devi Umamaheswari
• Singhee Amith
• Sinn Mathieu
• Lonij Vincent
• Azad Amar P

Classifications

• H02J2103/30

  Simulating, planning, modelling, reliability check or computer assisted design [CAD] of electric power networks · CPC title

• H02J2101/28

  Wind energy · CPC title

• H02J2101/24

  Photovoltaics · CPC title

• H02J2101/22

  Solar energy · CPC title

• H02J2101/20

  Dispersed power generation using renewable energy sources · CPC title