Method and device for controlling an energy-generating system which can be operated with a renewable energy source

Patent metadata

Field	Value
Publication number	US-9853592-B2
Application number	US-201314764102-A
Country	US
Kind code	B2
Filing date	Dec 3, 2013
Priority date	Feb 5, 2013
Publication date	Dec 26, 2017
Grant date	Dec 26, 2017

Abstract

A method and a device for controlling an energy-generating system are operated with a renewable energy source. In the method, a prediction about an energy yield of the energy-generating system is made for a predefined prediction time period, and a predefined area, using a learning system with an input vector and an output vector. The output vector includes operating variables for a multiplicity of successive future times of the time period. The input vector includes variables, influencing the operating variables, for a point in time from a multiplicity of points in time of a predefined observation time period. The input variables include at least three items of information for the observation time period and the predefined area. The energy-generating system is controlled on the basis of the generated prediction such that weather-conditioned fluctuations in the energy yield of the energy-generating system are reduced.

First claim

The invention claimed is:

1. A method for controlling an energy-generating system which can be operated with a renewable energy source, comprising:
generating a prediction of an energy yield of the energy-generating system for a predefined prediction period and a predefined area using a learning system formed by a number n of neural networks each with an input vector and an output vector,
wherein the output vector comprises one or more operating variables of the energy-generating system for a plurality of consecutive future times of the predefined prediction period, and
wherein the input vector comprises one or more input variables influencing the operating variable or operating variables for one time from a plurality of times of a predefined observation period,
wherein each of the n neural networks comprises an artificial neural feed-forward network with a plurality of interconnected layers which comprise an input layer, a plurality of hidden layers and an output layer,
wherein the input layer contains a number of input neurons to describe the input vectors,
wherein a respective hidden layer contains a number of hidden neurons and wherein the output layer contains a number of output neurons to describe the output vectors,
wherein the output layer comprises a plurality of output clusters corresponding to the plurality of hidden layers and in each case comprising one or more output neurons,
wherein each output cluster describes the same output vector and is connected to a different hidden layer,
wherein the input variables comprise at least three of the following data for the predefined observation period and the predefined area:
weather data,
first image data of a cloud drift provided by a satellite,
second image data of the cloud drift provided by a ground camera, and
simulation data generated by a physical model for simulating the energy yield of the energy-generating system using the weather data, and

controlling the energy-generating system on the basis of the generated prediction such that weather-related fluctuations in the energy yield of the energy-generating system are reduced.

2. The method as claimed in claim 1 ,
wherein the input vector is compressed before the generation of the prediction through a principal component analysis of the components of the learning system.

3. The method as claimed in claim 1 ,
wherein the input vector for an i-th neural network, with iε[1, . . . , n], comprises the output vector of the (i−1)th neural network in addition to the input variables.

4. The method as claimed in claim 1 ,
wherein the input variables are provided individually for each of the n neural networks.

5. The method as claimed in claim 1 ,
wherein a sequence of the n neural networks is predefinable.

6. The method as claimed in claim 1 ,
further comprising multiple execution of the step of generating a prediction in order to generate a plurality of predictions, wherein a different prediction period and/or a different observation period is specified in each case for the generation of a respective prediction.

7. The method as claimed in claim 6 ,
wherein the plurality of generated predictions are amalgamated to form an amalgamated prediction.

8. The method as claimed in claim 7 ,
wherein the plurality of generated predictions are amalgamated by a weighted summation.

9. The method as claimed in claim 7 ,
wherein the plurality of generated predictions are amalgamated by a further neural network.

10. The method as claimed in claim 1 ,
wherein the first and/or the second image data comprise image features provided by a pattern recognition.

11. A computer program product embodied on a non-transitory computer-readable media,
adapted to implement the method as claimed in claim 1 on a program-controlled device.

12. A non-transitory data medium with a stored computer program thereon
adapted to implement the method as claimed in claim 1 on a program-controlled device.

13. The method of claim 1 , wherein one of the at least three of said following data comprises the first image data of the cloud drift provided by the satellite.

14. A device for controlling an energy-generating system which can be operated with a renewable energy source, comprising:
a prediction generator adapted to generate a prediction relating to an energy yield of the energy-generating system for a predefined prediction period and a predefined area using a learning system formed by a number n of neural networks each with an input vector and an output vector,
wherein the output vector comprises one or more operating variables of the energy-generating system for a plurality of consecutive future times of the predefined prediction period, and
wherein the input vector comprises one or more input variables influencing the operating variable or operating variables for one time from a plurality of times of a predefinable observation period,
wherein each of the n neural networks comprises an artificial neural feed-forward network with a plurality of interconnected layers which comprise an input layer, a plurality of hidden layers and an output layer,
wherein the input layer contains a number of input neurons to describe the input vectors,
wherein a respective hidden layer contains a number of hidden neurons and wherein the output layer contains a number of output neurons to describe the output vectors,
wherein the output layer comprises a plurality of output clusters corresponding to the plurality of hidden layers and in each case comprising one or more output neurons,
wherein each output cluster describes the same output vector and is connected to a different hidden layer,
wherein the input variables comprise at least three of the following data for the predefined observation time period and the predefined area:
weather data,
first image data of a cloud drift provided by a satellite,
second image data of the cloud drift provided by a ground camera, and
simulation data generated by a physical model for simulating the energy yield of the energy-generating system using the weather data, and

a controller adapted to control the energy-generating system on the basis of the generated prediction in such a way that weather-related fluctuations in the energy yield of the energy-generating system are reduced.

15. The device of claim 14 , wherein one of the at least three of said following data comprises the first image data of the cloud drift provided by the satellite.

Assignees

• Siemens Ag

Inventors

• Bischoff Martin
• Chen Terrence
• Grothmann Ralph
• Hennig Oliver
• Kim Johann
• Ritzhaupt-Kleissl Eberhard

Classifications

• G05B13/027

  using neural networks only · CPC title

• Y04S10/56

  Cross-Sectional Technologies · mapped topic

• G06Q10/08

  Logistics, e.g. warehousing, loading or distribution; Inventory or stock management · CPC title

• H02S10/00 Primary

  PV power plants; Combinations of PV energy systems with other systems for the generation of electric power · CPC title

• G05B17/00

  Systems involving the use of models or simulators of said systems (G05B13/00, G05B15/00, G05B19/00 take precedence) · CPC title