Rotation system and controller for photovoltaic tracker

Patent metadata

Field	Value
Publication number	US-12413178-B2
Application number	US-202318160163-A
Country	US
Kind code	B2
Filing date	Jan 26, 2023
Priority date	Jul 27, 2020
Publication date	Sep 9, 2025
Grant date	Sep 9, 2025

Abstract

An image/video capture apparatus is provided, to capture image data of a photovoltaic panel ( 101 ) and surroundings, and a controller ( 502 ) recognizes a category corresponding to the image data, and generates different instructions to control rotation of a photovoltaic tracker ( 102 ), to remove dust or snow accumulating on the photovoltaic panel ( 101 ). This does not need continuous heating, but is simple and efficient, saves electric energy, and resolves a technical problem that a conventional snow removal method of heating has low efficiency and wastes electric energy. In addition, a corresponding controller ( 1200 ) is further provided.

First claim

What is claimed is:

1. A method comprising:
obtaining image data related to a photovoltaic panel or environment around the photovoltaic panel, the photovoltaic panel being mounted on a photovoltaic tracker, and rotatable with the photovoltaic tracker, wherein obtaining the image data comprises:
obtaining daytime images and nighttime images;
selecting first images from the daytime images as a first verification set and using remaining daytime images as a first training set;
selecting second images from the nighttime images as a second verification set and using remaining nighttime images as a second training set;
training a first neural network using the first training set;
training a second neural network using the second training set; and
classifying a received image as a daytime image or nighttime image and processing the received image using the first neural network or the second neural network based on the classification; and

generating, based on a category corresponding to the processed image from the first neural network or the second neural network, an instruction of a plurality of instructions to control rotation of the photovoltaic tracker, to move the photovoltaic panel in a corresponding manner, the plurality of instructions corresponding to different image data categories.

2. The method according to claim 1 , wherein generating the instruction comprises:
based on the category corresponding to the image data being snowfall or hailing, generating a first instruction to instruct the photovoltaic tracker to adjust to a preset tilt angle.

3. The method according to claim 1 , wherein generating the instruction comprises:
based on the category corresponding to the image data being rainfall and dust accumulation, generating a second instruction to instruct the photovoltaic tracker to rotate to a preset angle or to rotate n times, and n is an integer greater than or equal to 1.

4. The method according to claim 1 , further comprising:
obtaining video data; and
extracting the image data from the video data at a preset time interval.

5. The method according to claim 1 , further comprising:
removing the image data when image brightness of the image data is less than a preset threshold.

6. The method according to claim 1 , wherein training the first neural network and the second neural network comprises:
classifying each image in the first training set and the second training set into one of a plurality of weather categories, wherein the weather categories comprise cloudy, sunny, rainy, snowy, and hailing; and
determining, for each classified image, a degree of condition coverage selected from a plurality of coverage levels.

7. The method according to claim 1 , wherein processing the image data comprises:
filtering the image data based on overall brightness and black part proportions to remove images having poor quality; and
performing image recognition using the trained first neural network or second neural network to determine a weather condition category for filtered images.

8. The method according to claim 1 , further comprising:
determining a time at which each image is captured;
classifying each image as daytime or nighttime based on the time, wherein daytime comprises a predetermined time period; and
selecting images for processing based on the daytime or nighttime classification.

9. The method according to claim 1 , wherein each different image data category corresponds to a specific tracker rotation angle and rotation sequence.

10. A method comprising:
obtaining image data related to a photovoltaic panel or environment around the photovoltaic panel, the photovoltaic panel being mounted on a photovoltaic tracker, and rotatable with the photovoltaic tracker, wherein obtaining the image data comprises:
obtaining a plurality of images using an image capture device;
classifying and marking the plurality of images to create a training set and a verification set;
uploading the training set and the verification set to a cloud training system;
training, by the cloud training system, a neural network model using the training set;
receiving the trained neural network model from the cloud training system at a local controller; and
processing subsequently captured images using the trained neural network model at the local controller; and

generating, based on a category corresponding to the processed image from the trained neural network model, an instruction of a plurality of instructions to control rotation of the photovoltaic tracker, to move the photovoltaic panel in a corresponding manner, the plurality of instructions corresponding to different image data categories.

11. The method according to claim 10 , wherein classifying and marking the plurality of images comprises: classifying each image into one of a plurality of weather categories, wherein the weather categories comprise cloudy, sunny, rainy, snowy, and hailing; and classifying each image into one of a plurality of temporal categories, wherein the temporal categories comprise daytime and nighttime based on a time at which each image is captured.

12. The method according to claim 10 , wherein the training set and verification set are created by: selecting a predetermined percentage of the plurality of images as the verification set; and using remaining images of the plurality of images as the training set.

13. The method according to claim 10 , further comprising: filtering the plurality of images prior to creating the training set and verification set by removing images based on overall brightness and black part proportions to remove images having poor quality and corresponding to nighttime and early morning conditions.

14. The method according to claim 10 , wherein obtaining the plurality of images comprises:
obtaining video data from the image capture device; and
extracting images from the video data at a preset time interval, wherein the preset time interval is selected to obtain a predetermined number of images per hour.

15. The method according to claim 10 , wherein generating the instruction comprises:
based on the category corresponding to the processed image being nighttime and snowfall, generating a first instruction to adjust the photovoltaic tracker to a maximum tilt angle in a downwind direction determined based on current wind direction data.

16. The method according to claim 10 , wherein generating the instruction comprises:
based on the category corresponding to the processed image being nighttime and rainfall with dust accumulation recognized during daytime, generating a second instruction to adjust the photovoltaic tracker to repeatedly rotate toward east and west at a maximum angle for a predetermined number of times.

17. A method comprising:
obtaining image data related to a photovoltaic panel or environment around the photovoltaic panel, the photovoltaic panel being mounted on a photovoltaic tracker and rotatable with the photovoltaic tracker, wherein obtaining the image data comprises:
obtaining a plurality of images of the photovoltaic panel;
filtering the plurality of images based on overall brightness and black part proportions to remove images having poor quality and corresponding to nighttime and early morning conditions;
performing transfer learning by:
selecting a predetermined number of images from each category of obtained images;
mixing the selected images with existing training data in equal proportion; and
updating a neural network model based on the mixed images; and

generating, based on a category corresponding to processe

Assignees

• Huawei Digital Power Tech Co Ltd

Inventors

• Wan Song
• Liu Qinwei
• Sun Wuben
• Zhang Yanzhong
• Liu Kai

Classifications

• H02S40/10

  Cleaning arrangements · CPC title

• F24S30/425

  Horizontal axis · CPC title

• F24S40/20

  Cleaning; Removing snow · CPC title

• G05D3/105

  Solar tracker · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title