A method and system for controlling the disassembly process of decommissioned photovoltaic modules
By employing temperature zone detection and multi-head attention processing technology, the problem of incomplete removal of encapsulant film during photovoltaic module dismantling has been solved, enabling efficient dismantling and recycling of photovoltaic modules.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ANHUI ZHENHAO ENVIRONMENTAL PROTECTION TECH CO LTD
- Filing Date
- 2026-06-08
- Publication Date
- 2026-07-31
AI Technical Summary
Existing technologies cannot perform localized temperature control during the disassembly of photovoltaic modules, resulting in incomplete removal of the encapsulant film and reduced utilization of the photovoltaic modules.
By employing temperature zone detection and multi-head attention processing technology, the photovoltaic modules are dynamically controlled by performing temperature zone detection on the photovoltaic modules and combining multi-head attention processing and observation and adjustment models to dynamically adjust the heating power and air volume valve opening.
It achieves complete removal of the encapsulant film from photovoltaic modules, avoiding physical damage and improving the dismantling, recycling, and utilization rates of photovoltaic modules.
Smart Images
Figure CN122480071A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of intelligent recycling and dismantling technology, specifically relating to a method and system for controlling the dismantling process of retired photovoltaic modules. Background Technology
[0002] With the large-scale development of the photovoltaic industry, the stock of retired photovoltaic modules continues to rise, making the demand for module recycling and dismantling increasingly urgent. Existing recycling technologies are gradually evolving from extensive manual dismantling to thermal debinding and mechanical screening, but intelligent control and fine sorting technologies are still relatively weak, making it difficult to meet the development needs of industrialized, efficient, and high-purity recycling.
[0003] Existing technology (Publication No.: CN119502030A) discloses a photovoltaic module dismantling device and method. The photovoltaic module dismantling device includes a control component, a cutting component, and a driving component. The control component includes an electrically connected power supply module and a transmitting coil. The cutting component includes a receiving coil and a hot-cutting head, which are electrically connected to the receiving coil. The hot-cutting head is used to thermally cut the adhesive. The driving component is used to control the movement of the cutting component in the gap between the photovoltaic module and the fixing body during hot cutting. This solution allows for flexible arrangement of the photovoltaic module dismantling device. Using hot cutting technology to cut the adhesive enables non-destructive dismantling of the photovoltaic module, avoiding damage to both the photovoltaic module and the fixing body, improving the reusability of the photovoltaic module, reducing production costs, and increasing the maintenance and replacement efficiency of the photovoltaic module.
[0004] The above-mentioned thermal cutting technology can cut the colloid to achieve non-destructive disassembly of photovoltaic modules. However, in the process of temperature control of photovoltaic modules, the photovoltaic modules are heated as a whole, and local zoning control is not possible. As a result, the colloid film is not completely removed, which reduces the utilization rate of photovoltaic modules. Summary of the Invention
[0005] The purpose of this invention is to solve the problem that in the process of temperature control of photovoltaic modules, the photovoltaic modules are heated as a whole and cannot be controlled locally, resulting in incomplete removal of the encapsulant film and reduced utilization of the photovoltaic modules. Therefore, this invention proposes a method and system for controlling the dismantling process of retired photovoltaic modules.
[0006] In a first aspect of this invention, a method for controlling the dismantling process of decommissioned photovoltaic modules is first proposed, the method comprising:
[0007] The target photovoltaic module is subjected to pyrolysis to remove the encapsulant film from the photovoltaic components, thus obtaining the individual photovoltaic components.
[0008] Photovoltaic components whose size is smaller than a preset size threshold are selected to obtain each valid photovoltaic component.
[0009] Collect images of each valid photovoltaic component, and call the preset photovoltaic component recognition model to identify the photovoltaic component images to obtain the photovoltaic component category;
[0010] Based on the category of photovoltaic components, a pre-defined database is searched to determine the disassembly scheme for the photovoltaic components;
[0011] The execution process through pyrolysis includes:
[0012] Temperature zoning detection was performed on the photovoltaic modules to be pyrolyzed to obtain corresponding observation information;
[0013] Multi-head attention processing is applied to the observation information of each temperature zone to obtain fused global collaborative features;
[0014] The integrated global collaborative features are substituted into the observation and regulation model to obtain the corresponding comprehensive regulation parameters; the comprehensive regulation parameters include the comprehensive heating power regulation amount and the comprehensive air volume valve opening value regulation amount.
[0015] The actual control parameters are obtained by correcting the comprehensive adjustment parameters and the historical control parameters of each temperature zone.
[0016] The actual control parameters of all temperature zones are transmitted to the control terminal for pyrolysis control.
[0017] Optionally, multi-head attention processing can be applied to the observation information of each temperature zone to obtain fused global collaborative features, including:
[0018] The observation information for each temperature zone is vectorized to obtain the corresponding feature vector;
[0019] Linear projection of all feature vectors yields the query vector, key vector, and value vector for each temperature partition;
[0020] The corresponding attention weights are calculated for the query vector and key vector of each temperature partition;
[0021] The global collaborative features are obtained by multiplying the attention weights of each temperature zone with the feature vectors corresponding to each temperature zone.
[0022] The formulas for calculating attention weights include: ,in, This represents the attention weight for the i-th temperature partition. It is the mean of the attention across all temperature zones. This represents the transpose of the query vector for the i-th temperature partition. Let i represent the key vector of the j-th temperature zone, where i ≠ j; Represents the key vector of the i-th temperature zone;
[0023] The calculation process for integrating global collaborative features is as follows:
[0024]
[0025] in, This represents the collaborative characteristics of the i-th temperature zone; Let represent the feature vector of the i-th temperature partition, and C represent the fusion of global collaborative features.
[0026] Optionally, the calculation process of the observation-based regulation model includes:
[0027] By substituting the integrated global collaborative features into the pyrolysis control function calculation, a comprehensive control score is obtained; the comprehensive control score evaluates the control compliance effect of all heating power adjustment and valve opening adjustment.
[0028] The comprehensive control parameters are obtained by mapping based on the comprehensive control score;
[0029] The specific formula for calculating the pyrolysis control function is as follows:
[0030]
[0031]
[0032] Where r is the overall control score, This indicates the total number of temperature zones. ( ) is the pyrolysis temperature barrier penalty function, T i It is the temperature value of the i-th temperature zone. This represents the standard pyrolysis temperature value. This indicates the allowable temperature fluctuation threshold. It is the standard deviation of the real-time temperature calculation for each temperature zone; Indicates the balanced power score; This represents the output heating power of the i-th temperature zone. It is the average output heating power of all temperature zones; This indicates the maximum output heating power.
[0033] Optionally, the process of target recognition of photovoltaic component images using a photovoltaic component recognition model includes:
[0034] The photovoltaic component identification model includes the main structure, neck structure, and detection head;
[0035] Substituting the photovoltaic component images into the main structure yields the first photovoltaic component features, the second photovoltaic component features, the third photovoltaic component features, and the fourth photovoltaic component features;
[0036] Substituting the features of the first photovoltaic component, the second photovoltaic component, the third photovoltaic component, and the fourth photovoltaic component into the neck structure yields the first photovoltaic component enhancement features, the second photovoltaic component enhancement features, the third photovoltaic component enhancement features, and the fourth photovoltaic component enhancement features;
[0037] Substituting the enhancement features of the first, second, third, and fourth photovoltaic components into the detection head yields the photovoltaic component category.
[0038] Optionally, the image processing of photovoltaic components through the backbone structure includes;
[0039] The photovoltaic component image is processed sequentially through the first Conv module, the second Conv module, and the spatially separable convolution module to obtain the first photovoltaic component feature;
[0040] Substitute the features of the first photovoltaic component into the C2f module to obtain the features of the second photovoltaic component;
[0041] The second photovoltaic component features are processed sequentially through the third Conv module, the spatially separable convolution module, and the C2f module to obtain the third photovoltaic component features.
[0042] The third photovoltaic component features are processed sequentially through the fourth Conv module, the spatially separable convolution module, the C2f module, and the SPPF module to obtain the fourth photovoltaic component features.
[0043] The execution process of the spatially separable convolution module is as follows: the input feature map is uniformly segmented into channels to obtain a first feature map, a second feature map, a third feature map, and a fourth feature map; the first feature map, the second feature map, the third feature map, and the fourth feature map are fused to obtain a fused feature map; and the fused feature map is convolved to obtain an output feature map.
[0044] Optionally, a target detection head is added to the detection head section, and a corresponding feature fusion module is added to the neck structure; the operation process of the feature fusion module includes:
[0045]
[0046] Where X1 represents the output feature of the C2f module in the neck structure, X2 represents the output feature of the spatially separable convolutional module in the trunk structure, and T1 and T2 represent the features generated during the operation. This represents the convolution operation. This represents the operations of the C2f module. This represents the operation of spatially separable convolutional modules. Indicates upsampling, T1 indicates channel splicing, and T3 indicates the feature output of the feature fusion module.
[0047] Optionally, during the training of the photovoltaic component recognition model, the loss function CIoU of the YOLOv8 model is replaced with the loss function InnerCIoU. The specific calculation process of the loss function InnerCIoU includes:
[0048]
[0049] in, The symbolic representation of the loss function, and These represent the Euclidean distances between the top-left and bottom-right corners of the ground truth bounding box and the predicted bounding box, respectively. , , , The coordinates of the true bounding box; , , , The coordinates of the predicted bounding box; It is the ratio of the Euclidean distance between the ground truth bounding box and the predicted bounding box; and These are the height and width of the target image.
[0050] In a second aspect of this invention, a control system for the dismantling process of decommissioned photovoltaic modules is proposed, the system comprising:
[0051] Pyrolysis module: Performs pyrolysis on the target photovoltaic module to remove the encapsulant film from the photovoltaic components, thus obtaining the individual photovoltaic components;
[0052] Preliminary screening module: Screens photovoltaic components whose size is smaller than a preset size threshold to obtain each valid photovoltaic component;
[0053] Small filtering module: Obtain images of each valid photovoltaic component, call the preset photovoltaic component recognition model to recognize the photovoltaic component images and obtain the photovoltaic component category;
[0054] Disassembly scheme module: Based on the category of photovoltaic components, search the preset database to determine the disassembly scheme of photovoltaic components.
[0055] The beneficial effects of this invention are:
[0056] This invention proposes a method and system for controlling the dismantling process of retired photovoltaic (PV) modules. Through pyrolysis, the encapsulant film can be removed and the individual components of the PV module can be completely separated, avoiding physical damage caused by dismantling. Valid PV components smaller than a preset threshold are screened, eliminating fragments or large residues that do not meet recycling standards, significantly improving the accuracy and efficiency of subsequent identification. A preset PV component identification model is used to automatically classify the acquired images, enabling rapid determination of component categories. Finally, dismantling schemes in a preset database are matched according to component categories, achieving a differentiated and standardized dismantling process for different PV components, improving the recycling rate. Dynamic zoning control is achieved by performing temperature zoning detection on the PV modules to be pyrolyzed, thereby eliminating the encapsulant film and further improving the dismantling and recycling rate of retired PV modules. Attached Figure Description
[0057] The invention will now be further described with reference to the accompanying drawings.
[0058] Figure 1 A flowchart of a method for controlling the dismantling process of decommissioned photovoltaic modules provided in an embodiment of the present invention;
[0059] Figure 2 This is a network structure diagram of a photovoltaic component identification model provided in an embodiment of the present invention;
[0060] Figure 3 This is a framework diagram of a control system for the dismantling process of retired photovoltaic modules, provided as an embodiment of the present invention. Detailed Implementation
[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and B can represent: A alone, A and B simultaneously, and B alone. Furthermore, descriptions involving "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" can explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.
[0062] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0063] This invention provides a method for controlling the dismantling process of retired photovoltaic modules. See also... Figure 1 The method includes the following steps:
[0064] The target photovoltaic module is subjected to pyrolysis to remove the encapsulant film from the photovoltaic components, thus obtaining the individual photovoltaic components.
[0065] Photovoltaic components whose size is smaller than a preset size threshold are selected to obtain each valid photovoltaic component.
[0066] Collect images of each valid photovoltaic component, and call the preset photovoltaic component recognition model to identify the photovoltaic component images to obtain the photovoltaic component category;
[0067] Based on the category of photovoltaic components, a pre-defined database is searched to determine the disassembly scheme for the photovoltaic components;
[0068] The execution process via pyrolysis includes:
[0069] Temperature zoning detection was performed on the photovoltaic modules to be pyrolyzed to obtain corresponding observation information;
[0070] Multi-head attention processing is applied to the observation information of each temperature zone to obtain fused global collaborative features;
[0071] The integrated global collaborative features are substituted into the observation and regulation model to obtain the corresponding comprehensive regulation parameters; the comprehensive regulation parameters include the comprehensive heating power regulation amount and the comprehensive air volume valve opening value regulation amount.
[0072] The actual control parameters are obtained by correcting the comprehensive adjustment parameters and the historical control parameters of each temperature zone.
[0073] The actual control parameters of all temperature zones are transmitted to the control terminal for pyrolysis control.
[0074] Specifically, the mapping process for obtaining the actual control parameters by correcting the comprehensive adjustment parameters and the historical control parameters of each temperature zone is as follows: photovoltaic components include unpyrolyzed solar cells, copper solder strips, etc.; preset size thresholds and preset databases are set by relevant personnel.
[0075] The present invention provides a method for controlling the dismantling process of decommissioned photovoltaic modules. Through pyrolysis, the encapsulant film can be removed and the individual components of the photovoltaic module can be completely separated, avoiding physical damage caused by dismantling. Valid photovoltaic components smaller than a preset threshold are screened, eliminating fragments or large pieces of residue that do not meet recycling standards, significantly improving the accuracy and efficiency of subsequent identification. A preset photovoltaic component identification model is used to automatically classify the acquired images, quickly determining the component category. Finally, dismantling schemes in a preset database are matched according to the component category, achieving a differentiated and standardized dismantling process for different photovoltaic components, improving the recycling rate. Temperature zoning detection is performed on the photovoltaic modules to be pyrolyzed, achieving dynamic zoning control, thereby eliminating the encapsulant film and further improving the dismantling and recycling rate of decommissioned photovoltaic modules.
[0076] In one implementation, the observation information from each temperature zone is processed using multi-head attention to obtain fused global collaborative features, including:
[0077] The observation information for each temperature zone is vectorized to obtain the corresponding feature vector;
[0078] Linear projection of all feature vectors yields the query vector, key vector, and value vector for each temperature partition;
[0079] The corresponding attention weights are calculated for the query vector and key vector of each temperature partition;
[0080] The global collaborative features are obtained by multiplying the attention weights of each temperature zone with the feature vectors corresponding to each temperature zone.
[0081] The formulas for calculating attention weights include: ,in, This represents the attention weight for the i-th temperature partition. It is the mean of the attention across all temperature zones. This represents the transpose of the query vector for the i-th temperature partition. Let i represent the key vector of the j-th temperature zone, where i ≠ j; Represents the key vector of the i-th temperature zone;
[0082] The calculation process for integrating global collaborative features is as follows:
[0083]
[0084] in, This represents the collaborative characteristics of the i-th temperature zone; Let represent the feature vector of the i-th temperature partition, and C represent the fusion of global collaborative features.
[0085] In one implementation, a multi-head attention mechanism is introduced into the pyrolysis control. This mechanism vectorizes and linearly projects the observation information from each temperature zone, calculating cross-zone attention weights to obtain fused global collaborative features. This effectively detects the thermal conduction coupling effect and airflow disturbances between different zones within the pyrolysis furnace, avoiding localized overheating or underheating problems caused by traditional independent PID control. When a zone experiences an abnormal temperature rise due to carbon buildup or uneven feeding, the dynamic distribution of attention weights immediately strengthens the feature association between that zone and neighboring zones, prompting the controller to reduce the heating power of the relevant zones and adjust the airflow valves in advance, thereby suppressing the spread of thermal runaway.
[0086] In one implementation, the calculation process of the observation-regulation model includes:
[0087] By substituting the integrated global collaborative features into the pyrolysis control function calculation, a comprehensive control score is obtained; the comprehensive control score evaluates the control compliance effect of all heating power adjustment and valve opening adjustment.
[0088] The comprehensive control parameters are obtained by mapping based on the comprehensive control score;
[0089] The specific formula for calculating the pyrolysis control function is as follows:
[0090]
[0091]
[0092] Where r is the overall control score, This indicates the total number of temperature zones. ( ) is the pyrolysis temperature barrier penalty function, T i It is the temperature value of the i-th temperature zone. This represents the standard pyrolysis temperature value. This indicates the allowable temperature fluctuation threshold. It is the standard deviation of the real-time temperature calculation for each temperature zone; Indicates the balanced power score; This represents the output heating power of the i-th temperature zone. It is the average output heating power of all temperature zones; This indicates the maximum output heating power.
[0093] In one implementation, a comprehensive control score is calculated by substituting the integrated global collaborative features into the pyrolysis control function, and the comprehensive heating power adjustment and airflow valve opening adjustment are obtained based on this score. The mapping process from score to parameters is as follows: a preset score-parameter mapping table is obtained, which is constructed using historical optimal control data; the comprehensive control score is substituted into the mapping table for lookup, and the corresponding combination of comprehensive heating power adjustment and comprehensive airflow valve opening adjustment is output. This approach unifies the conflicting objectives of temperature deviation control and heating power balance within a single scoring framework for optimization, rather than simply relying on fixed PID parameters. Specifically, the large deviation penalty term in the scoring function forces the system to respond quickly to severe over- or under-temperature conditions, while the Gaussian reward term allows for small fluctuations permissible by the process, avoiding over-adjustment; the balanced power scoring term penalizes uneven power distribution across zones, preventing a single heater from aging due to prolonged full load. This indicates the allowable temperature fluctuation threshold (ranging from 15° to 25°).
[0094] In one implementation, the pyrolysis mixture is initially screened to obtain large-sized recyclable products (such as wide weld strips). The initial screening process includes: the pyrolysis material in the tunnel furnace contains weld strips, broken glass, battery cells, ash, and some solid waste. The weld strips are long strips, ranging from 3cm to 30cm in size; the battery cells are irregularly shaped sheets, ranging from 1mm to 30mm in size; the glass is irregularly shaped sheets, ranging from 5mm to 100mm in size; and the other ash is smaller than 1mm. The pyrolysis material is fed into a linear vibrating screen via a conveyor. The vibrating screen has two layers of screens: the upper screen has a 30mm aperture, and the lower screen has a 1mm aperture. After screening, wide weld strips (over 30mm in diameter) and thin copper wires (less than 1mm in diameter) are obtained. Other mixtures (battery cells and glass) ranging from 1mm to 30mm are conveyed to the next air classification process via a belt conveyor. This process generates screening dust, small-diameter battery cell powder and glass powder, as well as noise. The dust is collected in a sealed environment and then treated by a bag filter to meet emission standards. After that, air separation and other operations are carried out.
[0095] In one implementation, see [link to implementation details]. Figure 2 This invention provides a network structure diagram for a photovoltaic component identification model; this model is an improvement upon the YOLOv8 model, specifically including:
[0096] In the backbone structure, each C2f module and the Conv module of the previous layer are connected by a spatially separable convolutional module; the SPPF module is replaced with a multi-scale module.
[0097] Add a target detection head to the detection head part and add a corresponding feature fusion module to the neck structure.
[0098] In one implementation, the process of target recognition of photovoltaic component images using a photovoltaic component recognition model includes:
[0099] The photovoltaic component identification model includes the main structure, neck structure, and detection head;
[0100] Substituting the photovoltaic component images into the main structure yields the first photovoltaic component features, the second photovoltaic component features, the third photovoltaic component features, and the fourth photovoltaic component features;
[0101] Substituting the features of the first photovoltaic component, the second photovoltaic component, the third photovoltaic component, and the fourth photovoltaic component into the neck structure yields the first photovoltaic component enhancement features, the second photovoltaic component enhancement features, the third photovoltaic component enhancement features, and the fourth photovoltaic component enhancement features;
[0102] Substituting the enhancement features of the first, second, third, and fourth photovoltaic components into the detection head yields the photovoltaic component category.
[0103] In one implementation, a photovoltaic component recognition model is used to identify photovoltaic component images, which systematically improves the visual screening capability of products after the dismantling of retired photovoltaic modules, especially the recognition accuracy and recall rate of high-value but small-sized targets (silver wires, small silicon wafer fragments, solder strips, etc.).
[0104] In one implementation, the image processing of photovoltaic components through the backbone structure includes:
[0105] The photovoltaic component image is processed sequentially through the first Conv module, the second Conv module, and the spatially separable convolution module to obtain the first photovoltaic component feature;
[0106] Substitute the features of the first photovoltaic component into the C2f module to obtain the features of the second photovoltaic component;
[0107] The second photovoltaic component features are processed sequentially through the third Conv module, the spatially separable convolution module, and the C2f module to obtain the third photovoltaic component features.
[0108] The third photovoltaic component features are processed sequentially through the fourth Conv module, the spatially separable convolution module, the C2f module, and the SPPF module to obtain the fourth photovoltaic component features.
[0109] The execution process of the spatially separable convolution module is as follows: the input feature map is uniformly segmented into channels to obtain a first feature map, a second feature map, a third feature map, and a fourth feature map; the first feature map, the second feature map, the third feature map, and the fourth feature map are fused to obtain a fused feature map; and the fused feature map is convolved to obtain an output feature map.
[0110] In one implementation, the input feature map is uniformly segmented into four sub-feature maps, which are then processed separately, fused together, and finally convolved for output. This significantly reduces the computational load and number of parameters of the model while maintaining relatively high detection accuracy, enabling its deployment on industrial control computers or edge computing devices commonly used in dismantling operations.
[0111] In one implementation, a target detection head is added to the detection head section, and a corresponding feature fusion module is added to the neck structure; the operation process of the feature fusion module includes:
[0112]
[0113] Where X1 represents the output feature of the C2f module in the neck structure, X2 represents the output feature of the spatially separable convolutional module in the trunk structure, and T1 and T2 represent the features generated during the operation. This represents the convolution operation. This represents the operations of the C2f module. This represents the operation of spatially separable convolutional modules. Indicates upsampling, T1 indicates channel splicing, and T3 indicates the feature output of the feature fusion module.
[0114] In one implementation, deep semantic features (X1, from the neck region, possessing high abstract expressive power) are organically fused with shallow high-resolution detail features (X2, from the trunk region, preserving fine spatial location information), enabling the small target detection head to simultaneously acquire semantic discrimination capabilities and precise localization capabilities. In a specific embodiment, it can detect tiny targets such as microcracks less than 0.1 mm wide and silver lines only 5-20 mm long in photovoltaic dismantling products.
[0115] In one implementation, during the training of the photovoltaic component identification model, the loss function CIoU of the YOLOv8 model is replaced with the loss function InnerCIoU. The specific calculation process of the loss function InnerCIoU includes:
[0116]
[0117] in, The symbolic representation of the loss function, and These represent the Euclidean distances between the top-left and bottom-right corners of the ground truth bounding box and the predicted bounding box, respectively. , , , The coordinates of the true bounding box; , , , The coordinates of the predicted bounding box; It is the ratio of the Euclidean distance between the ground truth bounding box and the predicted bounding box; and These are the height and width of the target image.
[0118] In one implementation, the loss function InnerCIoU is used to more precisely measure the localization deviation of the bounding box, which can more efficiently guide the optimization and convergence of the predicted box, enabling the detection box to quickly fit the real target area; the addition of precise constraints on the vertices of the bounding box makes the regression branch more accurate and stable in identifying small recyclable products.
[0119] Based on the same inventive concept, this invention also provides a control system for the dismantling process of decommissioned photovoltaic modules. See also Figure 3 The system includes the following modules:
[0120] Pyrolysis module: Performs pyrolysis on the target photovoltaic module to remove the encapsulant film from the photovoltaic components, thus obtaining the individual photovoltaic components;
[0121] Preliminary screening module: Screens photovoltaic components whose size is smaller than a preset size threshold to obtain each valid photovoltaic component;
[0122] Small filtering module: Obtain images of each valid photovoltaic component, call the preset photovoltaic component recognition model to recognize the photovoltaic component images and obtain the photovoltaic component category;
[0123] Disassembly scheme module: Based on the category of photovoltaic components, search the preset database to determine the disassembly scheme of photovoltaic components.
[0124] The decomposition process control system for retired photovoltaic modules provided by this invention can remove the encapsulant film and completely separate the various parts of the photovoltaic module through pyrolysis, avoiding physical damage caused by disassembly; it can screen valid photovoltaic parts with a size smaller than a preset threshold, and remove fragments or large residues that do not meet recycling standards, significantly improving the accuracy and processing efficiency of subsequent identification; it can automatically classify the acquired images using a preset photovoltaic part identification model, and quickly determine the part category; finally, it matches the disassembly scheme in the preset database according to the part category, realizing a differentiated and standardized disassembly process for different photovoltaic parts, improving the recycling rate; and it can achieve dynamic zoning control by performing temperature zoning detection on the photovoltaic modules to be pyrolyzed, thereby eliminating the encapsulant film of the photovoltaic modules and improving the disassembly and recycling rate of retired photovoltaic modules.
[0125] The foregoing has described one embodiment of the present invention in detail, but this content is merely a preferred embodiment and should not be considered as limiting the scope of the present invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the scope of the claims of this invention.
Claims
1. A method for controlling the dismantling process of retired photovoltaic modules, characterized in that, The method includes: The target photovoltaic module is subjected to pyrolysis to remove the encapsulant film from the photovoltaic components, thus obtaining the individual photovoltaic components. Photovoltaic components whose size is smaller than a preset size threshold are selected to obtain each valid photovoltaic component. Collect images of each valid photovoltaic component, and call the preset photovoltaic component recognition model to identify the photovoltaic component images to obtain the photovoltaic component category; Based on the category of photovoltaic components, a pre-defined database is searched to determine the disassembly scheme for the photovoltaic components; The execution process of the pyrolysis operation includes: Temperature zoning detection was performed on the photovoltaic modules to be pyrolyzed to obtain corresponding observation information; Multi-head attention processing is applied to the observation information of each temperature zone to obtain fused global collaborative features; The integrated global collaborative features are substituted into the observation and regulation model to obtain the corresponding comprehensive regulation parameters; the comprehensive regulation parameters include the comprehensive heating power regulation amount and the comprehensive air volume valve opening value regulation amount. The actual control parameters are obtained by correcting the comprehensive adjustment parameters and the historical control parameters of each temperature zone. The actual control parameters of all temperature zones are transmitted to the control terminal for pyrolysis control.
2. The method for controlling the dismantling process of decommissioned photovoltaic modules according to claim 1, characterized in that, Multi-head attention processing is applied to the observation information from each temperature zone to obtain fused global collaborative features, including: The observation information for each temperature zone is vectorized to obtain the corresponding feature vector; Linear projection of all feature vectors yields the query vector, key vector, and value vector for each temperature partition; The corresponding attention weights are calculated for the query vector and key vector of each temperature partition; The global collaborative features are obtained by multiplying the attention weights of each temperature zone with the feature vectors corresponding to each temperature zone.
3. The method for controlling the dismantling process of decommissioned photovoltaic modules according to claim 2, characterized in that, The formula for calculating the attention weight includes: in, This represents the attention weight for the i-th temperature partition. It is the mean of the attention across all temperature zones. This represents the transpose of the query vector for the i-th temperature partition. Let i represent the key vector of the j-th temperature zone, where i ≠ j; Represents the key vector of the i-th temperature zone; The calculation process for the fusion of global collaborative features is as follows: in, This represents the collaborative characteristics of the i-th temperature zone; Let represent the feature vector of the i-th temperature partition, and C represent the fusion of global collaborative features.
4. The method for controlling the dismantling process of decommissioned photovoltaic modules according to claim 1, characterized in that, The observation adjustment model's processing of fused global collaborative features includes: By substituting the integrated global collaborative features into the pyrolysis control function, a comprehensive control score is obtained; the comprehensive control score evaluates the control compliance effect of all heating power adjustment and valve opening adjustment. The comprehensive control parameters are obtained by mapping based on the comprehensive control score.
5. The method for controlling the dismantling process of decommissioned photovoltaic modules according to claim 4, characterized in that, The specific calculation formula for the pyrolysis control function is as follows: Where r is the overall control score, This indicates the total number of temperature zones. ( ) is the pyrolysis temperature barrier penalty function, T i It is the temperature value of the i-th temperature zone. This represents the standard pyrolysis temperature value. This indicates the allowable temperature fluctuation threshold. It is the standard deviation of the real-time temperature calculation for each temperature zone; Indicates the balanced power score; This represents the output heating power of the i-th temperature zone. It is the average output heating power of all temperature zones; This indicates the maximum output heating power.
6. The method for controlling the dismantling process of decommissioned photovoltaic modules according to claim 1, characterized in that, The process of target recognition of photovoltaic component images using the photovoltaic component recognition model includes: The photovoltaic component identification model includes a main structure, a neck structure, and a detection head; Substituting the photovoltaic component images into the main structure yields the first photovoltaic component features, the second photovoltaic component features, the third photovoltaic component features, and the fourth photovoltaic component features; Substituting the first photovoltaic component feature, the second photovoltaic component feature, the third photovoltaic component feature, and the fourth photovoltaic component feature into the neck structure yields the first photovoltaic component enhancement feature, the second photovoltaic component enhancement feature, the third photovoltaic component enhancement feature, and the fourth photovoltaic component enhancement feature; Substituting the enhancement features of the first, second, third, and fourth photovoltaic components into the detection head yields the photovoltaic component category.
7. The method for controlling the dismantling process of decommissioned photovoltaic modules according to claim 6, characterized in that, The image processing of photovoltaic components through the main structure includes: The photovoltaic component image is processed sequentially through the first Conv module, the second Conv module, and the spatially separable convolution module to obtain the first photovoltaic component feature; Substitute the features of the first photovoltaic component into the C2f module to obtain the features of the second photovoltaic component; The second photovoltaic component features are processed sequentially through the third Conv module, the spatially separable convolution module, and the C2f module to obtain the third photovoltaic component features. The third photovoltaic component features are processed sequentially through the fourth Conv module, the spatially separable convolution module, the C2f module, and the SPPF module to obtain the fourth photovoltaic component features. The execution process of the spatially separable convolution module is as follows: the input feature map is uniformly segmented into channels to obtain a first feature map, a second feature map, a third feature map, and a fourth feature map; the first feature map, the second feature map, the third feature map, and the fourth feature map are fused to obtain a fused feature map; and the fused feature map is convolved to obtain an output feature map.
8. The method for controlling the dismantling process of decommissioned photovoltaic modules according to claim 6, characterized in that, A target detection head is added to the detection head section, and a corresponding feature fusion module is added to the neck structure; The operation process of the feature fusion module includes: Where X1 represents the output feature of the C2f module in the neck structure, X2 represents the output feature of the spatially separable convolutional module in the trunk structure, and T1 and T2 represent the features generated during the operation. This represents the convolution operation. This represents the operations of the C2f module. This represents the operation of spatially separable convolutional modules. Indicates upsampling, T1 indicates channel splicing, and T3 indicates the feature output of the feature fusion module.
9. The method for controlling the dismantling process of decommissioned photovoltaic modules according to claim 6, characterized in that, During the training of the photovoltaic component identification model, the loss function CIoU of the YOLOv8 model is replaced with the loss function InnerCIoU. The specific calculation process of the loss function InnerCIoU includes: in, The symbolic representation of the loss function, and These represent the Euclidean distances between the top-left and bottom-right corners of the ground truth bounding box and the predicted bounding box, respectively. , , , The coordinates of the true bounding box; , , , The coordinates of the predicted bounding box; It is the ratio of the Euclidean distance between the ground truth bounding box and the predicted bounding box; and These are the height and width of the target image.
10. A process control system for dismantling retired photovoltaic modules, used to implement the process control method for dismantling retired photovoltaic modules as described in any one of claims 1-9, characterized in that, The system includes: Pyrolysis module: Performs pyrolysis on the target photovoltaic module to remove the encapsulant film from the photovoltaic components, thus obtaining the individual photovoltaic components; Preliminary screening module: Screens photovoltaic components whose size is smaller than a preset size threshold to obtain each valid photovoltaic component; Small filtering module: Obtain images of each valid photovoltaic component, call the preset photovoltaic component recognition model to recognize the photovoltaic component images and obtain the photovoltaic component category; Disassembly scheme module: Based on the category of photovoltaic components, search the preset database to determine the disassembly scheme of photovoltaic components.