TWI898831B - Solar panel recycling method and system - Google Patents

Abstract

The present invention relates to a solar panel recycling method and recycling system. The recycling steps include: S1: Remove the outer frame and junction box, crush the solar panel, and produce a plurality of fragments within a set size range. S2: Establish an AI image recognition system. This AI image recognition system takes photos of a plurality of fragments and performs AI recognition, distinguishing between different identified objects, such as battery cells, glass and battery cell composites, plastic backplates, and copper-tin conductors.S3: Set up a robotic arm system that corresponds to the identified objects, extracting and placing them into respective collection boxes for recycling. Through this invention, precise recycling and improved environmental benefits can be achieved.

Description

Solar panel recycling method and recycling system

This invention relates to a solar panel recycling method and system, primarily a method and system that can distinguish and recognize objects and automatically recycle solar panels.

With the trend toward energy transition and net-zero carbon emissions, the future energy mix will be dominated by solar and wind power. While solar technology continues to advance, silicon-based solar panels still account for the majority of production. Given the approximately 20-30 year lifespan of a silicon-based solar panel, a significant number of panels are expected to be discarded at the end of their useful life. Therefore, solar panel recycling has become an urgent issue.

The main components of a solar panel are glass, aluminum frame, battery cells, copper-tin conductors, plastic backsheet, and junction box. The aluminum frame and junction box can be easily disassembled and recycled, while the glass, battery cells, and plastic backsheet are bonded together by packaging materials and are difficult to decompose. Currently, solar panels are typically crushed and decomposed by thermal or chemical methods.

However, the aforementioned thermal cracking process will produce smoke and harmful substances, and the chemicals are also polluting, making the recycling process environmentally unfriendly.

The purpose of this invention is to provide a solar panel recycling method and system that can be automatically and accurately recycled using AI image recognition.

The recycling steps of the solar panel recycling method of the present invention include:

The S1 removes the outer frame and junction box, crushes the solar panels, and produces fragments within a range of predefined sizes.

S2 establishes an AI image recognition system. The AI image recognition system takes photos of multiple pieces of the broken grains and performs AI recognition, distinguishing the broken grains as different recognition objects.

S3 is equipped with a robotic arm system to remove the corresponding crushed particles according to the identified object and place them in the corresponding storage box for recycling.

In one embodiment of the present invention, the S1 step comprises a crushing device and a screening device. The crushing device crushes the solar panels, and the screening device selects the crushed particles within a set size range.

In one embodiment of the present invention, there are four types of identification objects: battery cells; glass and battery cell composites; plastic backsheets; and copper-tin conductors.

In one embodiment of the present invention, the AI image recognition system includes a camera and a data processing device. After the camera takes a picture, the data processing device processes the image, creates an AI data model for recognition, and outputs the recognized object and its location data to the robotic arm system.

One embodiment of the present invention further includes a feeder and a vibrating plate. The feeder and vibrating plate are electrically connected and integrated with the AI image recognition system and the robotic arm system. The crumbs produced in step S1 are placed in the feeder, which then deposits the crumbs onto the surface of the vibrating plate. The camera then takes a picture of the surface to generate the identification object.

In one embodiment of the present invention, the robotic arm system is activated when the identified object has a set take-out quantity, and executes the following judgment logic:

1. When the recognized quantity is outside the error range and is less than the set first quantity of the set take-out quantity, the feeder is notified to start.

2. If the identification error is within the range and the identified quantity is greater than the first quantity and less than the removed quantity, the vibrating disk is notified to activate.

3. When the only objects identified are the battery cell and the glass and battery cell composite, the vibrating disk is notified to activate.

The solar panel recycling system of the present invention includes a crushing device, a screening device, an AI image recognition device, and a robotic arm system. The crushing device breaks the solar panels into a plurality of fragments. The screening device has a screen that vibrates to select fragments within a set size range. The AI image recognition device has a camera and a data generation device. The camera captures the fragments, and the data processing device processes the images, creates an AI data model for identification, and generates a plurality of different identification objects and their location data. The robotic arm system receives the identification objects and location data output by the AI image recognition device and removes the identified objects.

In one embodiment of the present invention, the recycling system further includes a feeder and a vibrating plate. The feeder and vibrating plate are electrically connected and integrated with the AI image recognition system and the robotic arm system. The crushed particles screened out by the screening device are placed in the feeder, which then deposits the crushed particles onto the surface of the vibrating plate. The camera then takes a picture of the plate surface.

In one embodiment of the present invention, the crushing device is a claw-knife type crushing device; the screening device has a mesh size of 8 mm and can screen out objects with a size of 1.9-4 mm.

This invention utilizes the three aforementioned logical judgment methods to accurately classify and recycle waste, enabling continuous operation of the recycling system and achieving improved recycling efficiency. Furthermore, this invention uses AI imaging to accurately distinguish and recycle solar panel fragments with different identification targets, achieving more environmentally friendly recycling than traditional thermal decomposition or chemical methods.

S1: Remove the outer frame and junction box, crush the solar panels, and generate fragments within a range of set sizes.

S2: Establish an AI image recognition system. The AI image recognition system takes photos of the multiple pieces of the broken grains and performs AI recognition to distinguish between the multiple different recognition objects in the broken grains.

S3: Set up a robotic arm system to remove the identified object and place it in the corresponding storage box for recycling.

1: Crushing device

2: Screening device

21: Screen

3: AI Image Recognition System

31: Camera

32: Data processing device

T1: Battery Cell

T2: Glass and battery cell composite

T3: Plastic back panel

T4: Copper-tin conductor

4: Feeder

5: Vibrating plate

51: Market

6: Robotic Arm System

Figure 1 is a schematic diagram of the solar panel recycling steps of the present invention.

Figure 2 is a schematic diagram of the solar panel recycling system of the present invention.

Figure 3 is a schematic diagram of the identification object of the solar panel recycling system of the present invention.

Please refer to Figures 1-3, which respectively illustrate the solar panel recycling steps, the solar panel recycling system, and the identification object of the present invention. The solar panel recycling steps of the present invention include:

The S1 removes the outer frame and junction box, crushes the solar panels, and produces fragments within a range of predefined sizes.

S2 establishes an AI image recognition system. The AI image recognition system takes photos of multiple pieces of the broken grains and performs AI recognition, distinguishing whether the broken grains contain multiple different recognition objects.

S3 is equipped with a robotic arm system to remove the identified object and place it in the corresponding storage box for recycling.

In step S1, the aluminum frame and junction box, which can be directly recycled, are removed from the solar panels. The composite solar panels, consisting of battery cells, glass, plastic backsheet, and copper-tin conductors, are then crushed by crushing device 1, which can be a claw-knife type crusher and produces mostly 1.19 to 4 mm fragments.

The aforementioned crushed particles are placed in a screening device 2 for screening. The screening device 2 has a screen with an 8mm mesh and screens out crushed particles of 1.9-4mm.

The AI image recognition system 3 in step S2 includes a camera 31 and a data processing device 32. The camera 31 takes photos of the filtered particles and inputs the photo data into the data processing device 32, which then creates an AI data model. The AI model then distinguishes the particles into different recognition objects and their locations. In this embodiment, the recognition objects are four types: battery cells T1 with a size between 1.9 and 4 mm, a glass and battery cell composite T2, a plastic backsheet T3, and copper-tin conductors T4.

In this embodiment, the shooting light source can use colored white light, and the software is developed using Visual Studio, with the model language being Python. The AI image recognition system 3 is integrated with the robotic arm system 6, using the same package, such as CV2, tkinter, glob, or geopandas. After the AI image recognition system 3 takes a photo, the data processing device 32 performs image pre-processing on the photo, including Gaussian filtering and dedistortion. Two image enhancement methods are then applied: cropping and rotation.

This embodiment further includes a feeder 4 and a vibrating plate 5. Crumbs screened by the screening device 2 are placed into the feeder 4. The vibrating plate 5 is located on one side of the feeder 4 and can pour the crushed grains onto the vibrating plate 5. The vibrating plate 5 vibrates to evenly distribute the crushed grains on the plate surface 51, allowing the camera 31 to take photos of the plate surface 51 of the vibrating plate 5. Figure 3 shows an enlarged schematic diagram of the identified object. In this embodiment, approximately 50 crushed grains are distributed on the plate surface 51 of the vibrating plate 5 each time. Crumbs larger than 4 mm are returned to the crushing device 1 for further crushing. A small amount of crushed grains smaller than 1.9 mm are recovered by thermochemical means.

This example uses Yolo as the foundation for the AI recognition model. The image is cropped into six frames (5488 x 3762 to 640 x 640). The AI recognition model is loaded and recognition begins on the six small images. The recognition results and positions of the six images are integrated into a folder, and the image positions are converted into arm coordinates.

The robotic arm system 6 at position S3 is electrically connected and integrated with the AI image recognition system 3, the feeder 4, and the vibrating plate 5, communicating via Modbus. The robotic arm system 6 receives arm coordinates from the AI image recognition system 3 via Modbus and executes the program logic written for the arm system. The arm program logic in this embodiment undergoes three judgments for a single image. When a recognized object has a set removal quantity (eight in this embodiment), the robotic arm system 6 removes eight of the recognized objects from the vibrating plate 5 and places them in the corresponding storage bin.

The three judgment formulas of this embodiment are:

1. When the recognized quantity is outside the recognition error range and is less than the first quantity of the set take-out quantity 8 (in this embodiment, the first quantity is between 0 and 5 (not including 5)), the feeder 4 is notified to start.

The aforementioned action enables the feeder 4 to pour the crumbs onto the plate surface 51 of the vibrating plate 5, providing a sufficient amount of crumbs for the camera 31 to take a picture.

2. If the identification error is within the range and the identified quantity is greater than the first quantity and less than the removed quantity (in this embodiment, between 5 and 8 (not including 8)), the vibrating disk 5 is notified to activate.

In the aforementioned situation, since the particle recognition confidence level in the photo is less than 70%, it is not detected, and therefore the object type and location are not assigned. In this case, the number of particles on the vibrating plate 5 may have exceeded 8. Therefore, vibrating the vibrating plate 5 can flip and rotate the object to be recognized. This allows objects with a confidence level of less than 70% to be recognized by the recognition system again, as the camera 31 captures them from a different angle, raising the confidence level and allowing them to be recognized. Once the number of objects recognized again exceeds 8, the robotic arm system 6 can be activated to remove the object.

3. When the only recognized objects are the battery cell T1 and the glass and battery cell composite T2, the vibrating disk 5 is notified to activate.

Because the shapes and proportions of the four types of identification objects vary, after repeated removal cycles by the robotic arm system 6, only the difficult-to-identify battery cell T1 and glass-and-battery cell composite T2 may remain. This is because the glass-and-battery cell composite T2, when photographed with the glass side down by the upper camera 31, bears a high resemblance to the battery cell T1. To avoid misidentification, the AI image recognition system 3 instructs the vibrating plate 5 to vibrate again and take another photo if it determines that only these two types of objects are present, thereby improving accuracy.

This invention enables precise classification and recycling through the three aforementioned logical judgment methods, and allows the recycling system to operate continuously to improve efficiency. Furthermore, the AI image recognition system 3 and robotic arm system 6 of the present invention are equipped with a graphical user interface (not shown) that allows real-time observation of the object pickup process and the activation status of the robotic arm system 6, camera 31, feeder 4, and vibrating plate 5, enhancing operational convenience.

This invention uses AI imaging to accurately identify different solar panel fragments and automatically recycle them, offering superior recycling and environmental protection compared to traditional thermal decomposition or chemical methods. The aforementioned embodiments are merely illustrative and not limiting of this invention. Any equivalent modifications based on the spirit of this invention are also within its scope.

S1: Remove the outer frame and junction box, crush the solar panels, and generate fragments within a range of set sizes.

S2: Establish an AI image recognition system. The AI image recognition system takes photos of the multiple pieces of the broken grains and performs AI recognition to distinguish between the multiple different recognition objects in the broken grains.

S3: Set up a robotic arm system to remove the identified object and place it in the corresponding storage box for recycling.

Claims (2)

A solar panel recycling method includes the following steps: S1: removing the outer frame and junction box, crushing the solar panel, and generating a plurality of fragments within a set size range; S2: establishing an AI image recognition system, which takes photos of the plurality of fragments and performs AI recognition, and distinguishes the fragments as having a plurality of different recognition objects; S3: setting up a robotic arm system to remove the corresponding recognition objects and place them in the corresponding storage box for recycling; the recognition objects are divided into four categories: battery cells; glass and battery cell composites; plastic backplanes; and copper-tin wires; the AI image recognition system includes a camera and a data processing device. After the camera takes a picture, the data processing device processes the image, creates an AI data model for identification, and outputs the identified object and its location data to the robotic arm system. The system further includes a feeder and a vibrating plate. The feeder and the vibrating plate are electrically connected and integrated with the AI image recognition system and the robotic arm system. The pellets generated in step S1 are placed in the feeder, which then places the pellets on the vibrating plate surface. The camera then takes a picture of the plate surface to generate the identified object. The robotic arm system is activated when the identified object has a set removal quantity and executes the following judgment logic: 1. 1. When the recognized quantity is outside the error range and is less than the set first quantity of the set take-out quantity, the feeder is notified to start; 2. When the recognized quantity is within the error range and is greater than the first quantity and less than the take-out quantity, the vibrating plate is notified to start; 3. When the recognized object is only the battery cell and the glass and battery cell composite, the vibrating plate is notified to start. The solar panel recycling method as described in claim 1, wherein the step S1 comprises a crushing device and a screening device, wherein the crushing device crushes the solar panels and the screening device screens out particles within a set size range.