Monocrystal photovoltaic automatic doping and batching method and system based on intelligent AI

Abstract

The invention discloses a single-crystal photovoltaic automatic doping and batching method and system based on intelligent AI, and belongs to the technical field of photovoltaic production and manufacturing. Currently, manual batching in the photovoltaic industry has the problems of inaccurate inventory, low doping precision and the like, which seriously restricts the improvement of production efficiency and product quality. The system is composed of an automatic weighing unit, a data acquisition and preprocessing module, an AI intelligent decision module, an optimization algorithm module and an execution control module. The method comprises the steps that data of a coil base, materials and the like are collected and preprocessed, then an optimal doping formula is generated through an AI intelligent decision and optimization algorithm, and finally accurate weighing, doping agent adding, stirring and mixing are conducted through the execution control module. According to the invention, the doping precision is obviously improved, the error is controlled within an extremely small range, and the quality of monocrystalline silicon is improved; the production efficiency is greatly improved, and manual intervention is reduced; the production cost is effectively reduced, and manpower and raw material waste is reduced; the product quality stability is guaranteed, the market competitiveness is enhanced, and powerful support is provided for the development of the photovoltaic industry.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic production and manufacturing, and particularly relates to an innovative method and system for deeply applying intelligent AI technology to the single-crystal photovoltaic production process to achieve automatic doping batching. Background Art

[0002] With the continuous growth of the global demand for clean energy, the photovoltaic industry, as an important renewable energy field, has achieved rapid development in recent years. Among various types of solar cells, solar cells based on the semiconductor photovoltaic effect dominate the market due to their good photoelectric conversion performance and stability. Currently, most mainstream solar cells use silicon crystal materials, and crystalline silicon and recycled silicon are their main production raw materials.

[0003] During the subsequent use of recycled silicon, due to the differences in its internal impurity distribution and crystal structure, it will exhibit different resistance characteristics. To ensure the stable quality and consistent performance of the produced monocrystalline silicon, it is necessary to store and use recycled silicon in different grades according to its resistance. However, in the current photovoltaic production workshops, the on-site inventory of materials and the doping batching process mainly rely on manual operations.

[0004] When manually inventorying materials, due to the complex environment in the photovoltaic production workshop, the large quantity and scattered storage of materials, it is easy for staff to miss or misjudge, resulting in inaccurate inventory data. During the batching process, it is difficult to accurately control the dosage of dopants by manual operation, and problems such as under-dosing and mis-dosing exist. These human errors will not only affect the electrical properties of monocrystalline silicon, reduce its photoelectric conversion efficiency, but also may lead to unstable product quality, increase the defective rate, and thus increase the production cost.

[0005] In addition, with the wide application of intelligent manufacturing technology in various industries, the photovoltaic industry also urgently needs to improve the intelligent level of production lines. Achieving real-time monitoring, accurate prediction, and adaptive adjustment of production lines can effectively improve production efficiency, reduce energy consumption, and meet the growing market demand for high-quality photovoltaic products. However, the existing manual batching mode severely restricts the development process of the intelligentization of photovoltaic production lines, unable to respond to various changes in the production process in a timely and accurate manner, and difficult to meet the requirements of modern large-scale production. Therefore, the research and development of an advanced single-crystal photovoltaic automatic doping batching technology based on intelligent AI has become the key to promoting the sustainable development of the photovoltaic industry. Summary of the Invention

[0006] The present invention aims to provide a method and system for automatic doping batching of single-crystal photovoltaic based on intelligent AI, so as to solve the problems of large error, low efficiency, high cost and inability to meet the intelligent requirements of the production line in traditional manual batching. By introducing intelligent AI technology, highly automated, precise and intelligent control of the doping batching process is achieved, the quality and production efficiency of single-crystal silicon are improved, the production cost is reduced, the competitiveness of photovoltaic products in the market is enhanced, and strong support is provided for the sustainable development of the photovoltaic industry.

[0007] In a first aspect, an embodiment of the present invention provides a system for automatic doping batching of single-crystal photovoltaic based on intelligent AI, including:

[0008] An automatic weighing unit, the automatic weighing unit is provided with a weighing platform connected to the system, the weighing platform is equipped with a high-precision electronic scale, and the high-precision electronic scale is provided with a pressure sensor and a strain gauge sensor, which are used to convert the change in the weight of the dopant into an electrical signal, and transmit the converted digital signal to the data processing module of the system through a high-speed data transmission line;

[0009] A data acquisition and preprocessing module, which is used to collect the operation parameters of the workshop furnace platform, production plan instructions, material inventory information and equipment status data, and uses data cleaning algorithms to remove noise, outliers and duplicate data, and uses normalization algorithms to unify the data into a standard range;

[0010] An AI intelligent decision-making module, which integrates a deep learning neural network and an expert system. The deep learning neural network trains a large amount of historical production data, extracts and analyzes features of real-time data, and generates a preliminary doping formula suggestion plan. The expert system verifies and adjusts the preliminary plan based on industry knowledge and experience rules;

[0011] An optimization algorithm module, based on the preliminary formula generated by the AI intelligent decision-making module, uses genetic algorithms or particle swarm optimization algorithms, takes production cost, product quality indicators and production efficiency as multi-objective optimization functions, and determines the final optimal doping formula under the condition of meeting production process constraints;

[0012] An execution control module, according to the optimal doping formula determined by the optimization algorithm module, controls the high-precision electronic scale to weigh the dopant, uses the PID control algorithm to adjust the feeding speed and start-stop state of the feeding equipment, controls the automatic feeding equipment to add the dopant to the corresponding barrel after weighing, and coordinates the mixing device to perform uniform mixing.

[0013] In some embodiments of the present invention, in the above data acquisition and preprocessing module, the data cleaning algorithm uses the 3σ criterion based on statistical principles to identify and remove outliers, and uses the moving average filtering algorithm to remove noise data; for continuous data, the normalization algorithm uses the linear normalization method to map it to the [0,1] interval, and for discrete data, it uses the one-hot encoding method for processing.

[0014] In some embodiments of the present invention, the deep learning neural network of the above AI intelligent decision-making module adopts a convolutional neural network (CNN) or a recurrent neural network (RNN) architecture, and the expert system stores industry knowledge and experience rules in the form of a rule base.

[0015] In some embodiments of the present invention, the genetic algorithm of the above optimization algorithm module simulates the selection, crossover, and mutation operations in the process of biological evolution, encodes the doping formulation parameters as chromosomes for iterative optimization; the particle swarm optimization algorithm simulates the foraging behavior of a flock of birds, regards each possible doping formulation as a particle, and searches for the optimal solution through information sharing and cooperation among particles.

[0016] In some embodiments of the present invention, the feeding equipment of the above execution control module includes a screw conveyor and a vibrating feeder, and the stirring device includes a paddle stirrer and a screw stirrer.

[0017] Second, the embodiments of the present application provide a single-crystal photovoltaic automatic doping and batching method based on intelligent AI, including the following steps:

[0018] Data acquisition and preprocessing: The workshop furnace platform starts the material reporting program according to the production plan, sends the production-related data to the data acquisition and preprocessing module through the industrial Internet, and the module performs data cleaning and normalization processing;

[0019] AI intelligent formula generation: The preprocessed data is transmitted to the AI intelligent decision-making module, the deep learning neural network extracts and analyzes the data features, generates a preliminary doping formula suggestion plan, the expert system verifies and adjusts the preliminary plan, and the optimization algorithm module performs multi-objective optimization based on the adjusted preliminary formula using the genetic algorithm or the particle swarm optimization algorithm to determine the final optimal doping formula;

[0020] Dopant metering and addition: The execution control module controls the high-precision electronic scale to weigh the dopant according to the optimized formula, uses the PID control algorithm to adjust the feeding speed and start / stop state of the feeding equipment, after weighing, controls the automatic feeding equipment to add the dopant to the corresponding hopper, and starts the stirring device for uniform mixing.

[0021] In some embodiments of the present invention, in the above data acquisition and preprocessing steps, 3σ criterion based on statistical principles is used for data cleaning to identify and remove outliers, and a moving average filtering algorithm is used to remove noise data; for normalization processing, linear normalization method is used to map continuous data to the interval [0, 1], and one-hot encoding method is used to process discrete data.

[0022] In some embodiments of the present invention, in the above AI intelligent formula generation step, a deep learning neural network uses a convolutional neural network (CNN) or a recurrent neural network (RNN) architecture to train historical production data, and an expert system stores industry knowledge and experience rules in the form of a rule base to correct the preliminary scheme.

[0023] In some embodiments of the present invention, in the above AI intelligent formula generation step, the genetic algorithm simulates the selection, crossover, and mutation operations in the biological evolution process, encodes the doping formula parameters as chromosomes for iterative optimization; the particle swarm optimization algorithm simulates the foraging behavior of bird flocks, regards each possible doping formula as a particle, and searches for the optimal solution through information sharing and cooperation among particles.

[0024] In some embodiments of the present invention, in the above dopant metering and addition step, the feeding equipment includes a screw conveyor and a vibrating feeder, and the stirring device includes a paddle stirrer and a screw stirrer, and the stirring device stirs according to the preset stirring speed and time.

[0025] The embodiments of the present invention have at least the following advantages or beneficial effects:

[0026] The present invention constructs a complete single-crystal photovoltaic automatic doping batching system based on intelligent AI, where each module works in coordination, bringing significant advantages in multiple aspects. The automatic weighing unit utilizes a high-precision electronic scale and advanced sensors to achieve accurate measurement of the weight of the dopant and real-time data transmission, providing reliable basic data for subsequent accurate batching and ensuring the accuracy of batching from the source. The data acquisition and preprocessing module cleans and normalizes multi-source data, effectively removing noise and outliers, making the data more standardized and accurate, providing high-quality data input for the AI intelligent decision-making module, and helping to improve the accuracy and reliability of decision-making. The AI intelligent decision-making module integrates a deep learning neural network and an expert system, combining data-driven intelligent analysis and industry experience knowledge, and can generate doping formula suggestions more comprehensively and accurately, improving the rationality and scientificity of the formula. The optimization algorithm module uses advanced optimization algorithms to deeply optimize the preliminary formula, comprehensively considering multi-objective factors such as production cost, product quality, and production efficiency, ensuring that the final obtained is the optimal doping formula under various constraints, and balancing multiple key production indicators. The execution control module realizes accurate metering, addition, and mixing of the dopant according to the optimized formula, adopts the PID control algorithm to ensure the high precision and stability of the weighing and feeding processes, coordinates the stirring device to ensure uniform distribution of the dopant, and further improves the consistency of product quality.

[0027] The present invention uses the 3σ criterion to identify and remove outliers, which can effectively exclude unreasonable data caused by factors such as sensor failures or external interferences, ensuring the authenticity and reliability of the data. The moving average filtering algorithm can smooth the data, remove the influence of high-frequency noise, and make the data more stable and accurate. For continuous data, the linear normalization method is used to map it to the [0,1] interval, and for discrete data, the one-hot encoding method is used for processing, unifying the magnitude and format of the data, enabling different types of data to be processed and analyzed under the same standard, improving the AI model's understanding and processing ability of the data, and thus enhancing the decision-making accuracy and stability of the entire system.

[0028] The present invention clarifies the specific architectures and forms of the deep learning neural network and the expert system in the AI intelligent decision-making module. Convolutional neural networks (CNNs) or recurrent neural networks (RNNs) have powerful feature extraction and pattern recognition capabilities, can dig out complex laws and relationships from a large amount of historical production data, accurately analyze and predict real-time data, and provide intelligent technical support for generating preliminary doping formula suggestions. The expert system stores industry knowledge and experience rules in the form of a rule base, can verify and adjust the preliminary solutions generated by the neural network, make up for the deficiencies that may exist in pure data-driven decision-making, combines artificial intelligence technology and the wisdom of industry experts, makes the generated doping formula more in line with actual production needs, and improves the scientificity and practicality of the formula.

[0029] The present invention iteratively optimizes the doping formulation parameters and can quickly find a better solution in a complex search space. The particle swarm optimization algorithm simulates the foraging behavior of a bird flock. Through information sharing and cooperation among particles, it continuously adjusts the particle positions to search for the optimal solution, and has strong global search ability and convergence speed. These two algorithms use the production cost, product quality indicators, and production efficiency as the multi-objective optimization function, and optimize under the condition of meeting the production process constraints. They can find the best balance among multiple interrelated and restrictive objectives, ensuring that the finally obtained doping formulation can not only reduce the production cost, but also improve the product quality and production efficiency, achieving the maximization of production benefits.

[0030] The screw conveyor and vibrating feeder of the present invention have the characteristics of stable conveying and accurate feeding. They can accurately convey and add the dopant according to the instructions of the execution control module, ensuring the accuracy and stability of the dopant addition amount. The paddle stirrer and screw stirrer can fully stir the materials, making the dopant evenly distributed in the materials, ensuring the consistency and stability of the quality of single crystal silicon products. Different types of equipment combinations can be flexibly selected and configured according to specific production requirements and material characteristics, improving the adaptability and versatility of the system, and being able to better meet the doping batching requirements in different production scenarios.

[0031] The data acquisition and preprocessing steps of the present invention ensure the accuracy and standardization of the production-related data obtained, providing a reliable data basis for subsequent intelligent decision-making. The AI intelligent formulation generation step combines the intelligent analysis ability of the deep learning neural network and the empirical knowledge of the expert system, and generates the optimal doping formulation through the multi-objective optimization algorithm, improving the scientificity and rationality of the formulation, and being able to better balance multiple key indicators such as production cost, product quality, and production efficiency. The dopant metering and addition step realizes the precise operation of the dopant according to the optimized formulation, uses the PID control algorithm to ensure the high precision of weighing and feeding, and the stirring device ensures the uniform mixing of the dopant. The entire method process forms a closed-loop intelligent control process, effectively improving the accuracy and efficiency of single crystal photovoltaic doping batching, and enhancing the stability of product quality.

[0032] The present invention further refines the specific algorithms in the data acquisition and preprocessing steps. The application of the 3σ criterion and the moving average filtering algorithm can effectively remove the outliers and noise in the data, ensure the authenticity and reliability of the data, and avoid the interference of incorrect data on subsequent intelligent decision-making. The linear normalization and one-hot encoding methods unify the data format and magnitude, enabling different types of data to be processed and analyzed under the same standard, improving the processing efficiency and accuracy of the AI model for data, and thus providing strong support for generating a more reasonable and accurate doping formulation, which helps to improve the performance and effect of the entire batching method.

[0033] The training of the convolutional neural network (CNN) or recurrent neural network (RNN) of the present invention on historical production data can mine potential laws and patterns in the data, accurately analyze and predict real-time data, and provide an intelligent technical means for generating preliminary doping formula suggestions. The expert system stores industry knowledge and experience rules in the form of a rule base, can correct and improve the preliminary solutions generated by the neural network, combines artificial intelligence technology and the wisdom of industry experts, makes the generated doping formula more in line with the actual production situation, improves the feasibility and effectiveness of the formula, and helps to improve the quality and production efficiency of single-crystal silicon products.

[0034] The genetic algorithm of the present invention performs iterative optimization by simulating the biological evolution process, and can quickly search for better solutions in a complex parameter space; the particle swarm optimization algorithm continuously adjusts the particle positions to find the optimal solution through cooperation and information sharing among particles, and has strong global search capabilities. These two algorithms are guided by a multi-objective optimization function, optimize the doping formula under the constraints of production processes, can find the best balance among multiple objectives such as production cost, product quality, and production efficiency, ensure that the generated doping formula can maximize production benefits, and improve the scientificity and economy of the entire batching method.

[0035] The use of the screw conveyor and vibrating feeder of the present invention ensures the accuracy and stability of the doping agent transportation and addition, and can accurately add the doping agent according to the optimized formula. The paddle stirrer and screw stirrer stir according to the preset stirring speed and time, can make the doping agent fully and evenly mixed in the material, and ensure the consistency and stability of the quality of single-crystal silicon products. The reasonable selection and configuration of different types of equipment and clear stirring requirements improve the controllability and reliability of the doping agent metering and addition steps, and further enhance the effect of the entire batching method and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0037] Figure 1 is the principle block diagram of the present invention;

[0038] Figure 2 is the step flow chart of the single-crystal photovoltaic automatic doping batching method of the present invention;

[0039] Figure 3A structural block diagram of an electronic device provided by an embodiment of the present invention.

[0040] Explanation of reference numerals: 101, memory; 102, processor; 103, communication interface. Specific embodiments

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0042] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0043] It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0044] It should be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or device. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0045] In the description of the embodiments of the present invention, "a plurality of" represents at least two.

[0046] In the description of the embodiments of the present invention, it should also be noted that unless otherwise clearly specified and limited, the terms "set", "installed", "connected", and "coupled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0047] As Figures 1-3 , the present invention provides a single-crystal photovoltaic automatic doping batching system based on intelligent AI, including:

[0048] An automatic weighing unit, the automatic weighing unit is provided with a weighing platform connected to the system, the weighing platform is equipped with a high-precision electronic scale, and the high-precision electronic scale is provided with a pressure sensor and a strain gauge sensor, which are used to convert the change in the weight of the dopant into an electrical signal and transmit the converted digital signal to the data processing module of the system through a high-speed data transmission line;

[0049] A data acquisition and preprocessing module, which is used to collect the operation parameters of the workshop furnace platform, production plan instructions, material inventory information, and equipment status data, and uses a data cleaning algorithm to remove noise, outliers, and duplicate data, and uses a normalization algorithm to unify the data into a standard range;

[0050] An AI intelligent decision-making module, which integrates a deep learning neural network and an expert system. The deep learning neural network trains a large amount of historical production data, extracts and analyzes the features of real-time data, and generates a preliminary doping formula suggestion plan. The expert system verifies and adjusts the preliminary plan based on industry knowledge and experience rules;

[0051] An optimization algorithm module, based on the preliminary formula generated by the AI intelligent decision-making module, uses a genetic algorithm or a particle swarm optimization algorithm, takes the production cost, product quality index, and production efficiency as multi-objective optimization functions, and determines the final optimal doping formula under the condition of meeting the production process constraints;

[0052] An execution control module, according to the optimal doping formula determined by the optimization algorithm module, controls the high-precision electronic scale to weigh the dopant, uses a PID control algorithm to adjust the feeding speed and start-stop state of the feeding equipment, controls the automatic feeding equipment to add the dopant to the corresponding barrel after weighing, and coordinates the mixing device to perform uniform mixing.

[0053] In the embodiments of the present invention, in the above data acquisition and preprocessing module, the data cleaning algorithm uses the 3σ criterion based on statistical principles to identify and remove outliers, and uses the moving average filtering algorithm to remove noise data; for continuous data, the normalization algorithm uses the linear normalization method to map it to the [0,1] interval, and for discrete data, it uses the one-hot encoding method for processing.

[0054] In the embodiments of the present invention, the deep learning neural network of the above AI intelligent decision-making module adopts a convolutional neural network (CNN) or a recurrent neural network (RNN) architecture, and the expert system stores industry knowledge and experience rules in the form of a rule base.

[0055] In the embodiments of the present invention, the genetic algorithm of the above optimization algorithm module simulates the selection, crossover, and mutation operations in the biological evolution process, encodes the doping formulation parameters as chromosomes for iterative optimization; the particle swarm optimization algorithm simulates the foraging behavior of bird flocks, regards each possible doping formulation as a particle, and searches for the optimal solution through information sharing and cooperation among particles.

[0056] In some embodiments of the present invention, the feeding equipment of the above execution control module includes a screw conveyor and a vibrating feeder, and the stirring device includes a paddle stirrer and a screw stirrer.

[0057] Furthermore, the embodiments of the present invention provide a single-crystal photovoltaic automatic doping and batching method based on intelligent AI, including the following steps:

[0058] Data acquisition and preprocessing: The workshop furnace platform starts the material reporting program according to the production plan, sends the production-related data to the data acquisition and preprocessing module through the industrial Internet, and the module cleans and normalizes the data;

[0059] AI intelligent formula generation: The preprocessed data is transmitted to the AI intelligent decision-making module, the deep learning neural network extracts and analyzes the features of the data, generates a preliminary doping formula suggestion plan, the expert system verifies and adjusts the preliminary plan, and the optimization algorithm module performs multi-objective optimization based on the adjusted preliminary formula using the genetic algorithm or the particle swarm optimization algorithm to determine the final optimal doping formula;

[0060] Dopant metering and addition: The execution control module controls the high-precision electronic scale to weigh the dopant according to the optimized formula, uses the PID control algorithm to adjust the feeding speed and start-stop state of the feeding equipment, after weighing, controls the automatic feeding equipment to add the dopant to the corresponding hopper, and starts the stirring device for uniform mixing.

[0061] In the embodiments of the present invention, in the above data acquisition and preprocessing steps, 3σ criterion based on statistical principles is used for data cleaning to identify and remove outliers, and the moving average filtering algorithm is used to remove noise data; for normalization processing, linear normalization method is adopted for continuous data to map it to the interval [0,1], and one-hot encoding method is adopted for discrete data.

[0062] In the embodiments of the present invention, in the above AI intelligent formula generation steps, the deep learning neural network uses a convolutional neural network (CNN) or a recurrent neural network (RNN) architecture to train historical production data, and the expert system stores industry knowledge and experience rules in the form of a rule base to correct the preliminary scheme.

[0063] In the embodiments of the present invention, in the above AI intelligent formula generation steps, the genetic algorithm simulates the selection, crossover and mutation operations in the biological evolution process, encodes the doping formula parameters as chromosomes for iterative optimization; the particle swarm optimization algorithm simulates the foraging behavior of bird flocks, regards each possible doping formula as a particle, and searches for the optimal solution through information sharing and cooperation among particles.

[0064] In the embodiments of the present invention, in the above dopant metering and addition steps, the feeding equipment includes a screw conveyor and a vibrating feeder, and the stirring device includes a paddle stirrer and a screw stirrer. The stirring device stirs according to the preset stirring speed and time.

[0065] Example 1:

[0066] System construction and operation:

[0067] Automatic weighing unit: A high-precision electronic scale (accuracy up to ±0.005g) is selected, equipped with high-precision pressure sensors and strain gauge sensors, which can quickly and accurately convert the weight change of the dopant into an electrical signal and transmit it to the data processing module in real time through a high-speed data transmission line.

[0068] Data acquisition and preprocessing module: It is connected to the furnace platforms, material storage equipment, etc. in the workshop through industrial Ethernet to collect various data in real time. Advanced data cleaning algorithms are used to effectively identify and remove outliers and noise data caused by sensor failures or external interferences; normalization algorithms are used to unify data of different magnitudes and units into a standard range.

[0069] AI intelligent decision-making module: The deep learning neural network uses a convolutional neural network (CNN) architecture to deeply train a large amount of historical production data. The expert system stores the experience and knowledge of industry senior experts in the form of a rule base, and can verify and adjust the preliminary formula generated by the neural network.

[0070] Optimization algorithm module: The particle swarm optimization algorithm is adopted, with production cost, product quality (such as resistivity uniformity and minority carrier lifetime of monocrystalline silicon), and production efficiency as the multi-objective optimization function. Under the condition of meeting production process constraints, the optimal doping formula is searched for.

[0071] Execution control module: Controls a high-precision electronic scale to weigh the dopant, and uses the PID control algorithm to dynamically adjust the feeding speed and start-stop state of the feeding equipment (screw conveyor). After weighing is completed, it controls the automated feeding equipment to accurately add the dopant to the corresponding hopper and coordinates the paddle stirrer for uniform stirring.

[0072] Batching process

[0073] Data acquisition and preprocessing: The workshop furnace plans to produce a certain specific type of monocrystalline silicon, and the material reporting program is started. The data acquisition and preprocessing module quickly collects furnace operation parameters (such as temperature 1430°C, pressure 1.04 atm), production plan instructions, and material inventory information, etc. The 3σ criterion is used to remove outliers in the temperature data, and the moving average filtering algorithm is used to smooth the pressure data. Linear normalization is performed on continuous data, and one-hot encoding is performed on discrete data.

[0074] AI intelligent formula generation: The preprocessed data is transmitted to the AI intelligent decision-making module. The CNN extracts and analyzes the features of the data, initially determines that phosphorus (P) and boron (B) need to be used as dopants, and gives the approximate dosage. The expert system evaluates and adjusts the preliminary plan according to the rule base to generate a preliminary doping formula. The particle swarm optimization algorithm of the optimization algorithm module determines the final optimal doping formula through multiple rounds of iteration, that is, the dosage of phosphorus is 32 grams and the dosage of boron is 28 grams.

[0075] Dopant metering and addition: The execution control module controls the high-precision electronic scale for weighing according to the optimized formula. During the weighing process, the PID control algorithm dynamically adjusts the feeding speed of the screw conveyor according to the weight data real-time feedback by the electronic scale. When the target weight is reached, the feeding is accurately stopped. After weighing is completed, it controls the automated feeding equipment to add phosphorus and boron dopants to the corresponding hoppers in sequence, and starts the paddle stirrer to stir at a speed of 320 revolutions per minute for 18 minutes to ensure that the dopant is fully mixed with the silicon raw material.

[0076] Product detection and results

[0077] The produced monocrystalline silicon is comprehensively detected, and the results show that: the average resistivity is 1.15 Ω·cm, the resistivity fluctuation range is controlled within ±0.03 Ω·cm, the minority carrier lifetime reaches 130 μs, and the photoelectric conversion efficiency is as high as 23%. The total time taken for this batch of production is 5.5 hours, and the deviation of the raw material usage from the theoretical value is controlled within ±0.3%.

[0078] Example 2:

[0079] System setup and operation

[0080] Automatic weighing unit: A high-precision electronic scale (accuracy of ±0.01 g) is used. Its sensor performance is stable, and it can accurately capture changes in the dopant weight and transmit data in a timely manner.

[0081] Data acquisition and preprocessing module: Connects with more production equipment for data docking to collect more comprehensive production data. An improved data cleaning algorithm is adopted to enhance the ability to identify and process outliers; the normalization algorithm is optimized according to the data characteristics to make data processing more accurate.

[0082] AI intelligent decision-making module: The deep learning neural network adopts a long short-term memory network (LSTM) architecture, which can better process production data with time series characteristics. The expert system continuously updates the rule base and incorporates the latest industry research results and production experience.

[0083] Optimization algorithm module: Combines genetic algorithm and simulated annealing algorithm, and synthesizes their advantages for formula optimization to achieve more efficient multi-objective search.

[0084] Execution control module: Controls the high-precision electronic scale and vibrating feeder to weigh and add dopants, and precisely adjusts the feeding speed through the PID control algorithm. The stirring device uses a spiral stirrer, and the stirring parameters can be adjusted according to the material characteristics.

[0085] Batching process

[0086] Data acquisition and preprocessing: The workshop furnace plans to produce another type of single crystal silicon and starts feeding materials. The data acquisition and preprocessing module collects data such as the furnace temperature of 1440 °C and pressure of 1.05 atm. The data is cleaned using the improved 3σ criterion and filtering algorithm, and targeted normalization methods are adopted for different types of data.

[0087] AI intelligent formula generation: The preprocessed data is input into the AI intelligent decision-making module. The LSTM network deeply analyzes the data, predicts that arsenic (As) and gallium (Ga) need to be used as dopants, and gives preliminary dosage suggestions. The expert system optimizes the plan in combination with the latest rule base to generate a preliminary doping formula. The optimization algorithm module, through the synergistic effect of genetic algorithm and simulated annealing algorithm, after multiple iterations, determines the final optimal doping formula, that is, the arsenic dosage is 35 grams and the gallium dosage is 25 grams.

[0088] Dopant metering and addition: The execution control module controls a high-precision electronic scale for weighing according to the optimized formula, and the PID control algorithm adjusts the feeding speed of the vibrating feeder in real time. After weighing, arsenic and gallium dopants are added to the barrel, and the screw agitator is started to agitate at a speed of 250 revolutions per minute for 20 minutes to ensure uniform distribution of the dopants.

[0089] Product detection and results

[0090] The produced single-crystalline silicon is detected. The average resistivity is 1.2 Ω·cm, the resistivity fluctuation range is within ±0.04 Ω·cm, the minority carrier lifetime reaches 125 μs, and the photoelectric conversion efficiency is 22.8%. The production of this batch takes 6 hours, and the deviation of the raw material usage from the theoretical value is controlled within ±0.4%.

[0091] Such as Figure 2 In addition, the embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by the processor 102, the system described in any one of the above first aspects is implemented. If the above functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0092] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0093] For those skilled in the art, it is obvious that the present application is not limited to the details of the above-mentioned exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or basic characteristics of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present application is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. A single crystal photovoltaic automatic doping and batching system based on intelligent AI, characterized in that: include: An automatic weighing unit, wherein the automatic weighing unit is provided with a weighing platform connected to the system, the weighing platform is equipped with a high-precision electronic scale, and the high-precision electronic scale is provided with a pressure sensor and a strain gauge sensor for converting the weight change of the dopant into an electrical signal, and transmitting the converted digital signal to the data processing module of the system through a high-speed data transmission line; The data acquisition and preprocessing module is used to collect workshop furnace operation parameters, production plan instructions, material inventory information and equipment status data, and use data cleaning algorithms to remove noise, outliers and duplicate data, and use normalization algorithms to unify the data into a standard range; AI intelligent decision-making module, integrating deep learning neural network and expert system. The deep learning neural network trains massive historical production data, extracts and analyzes features of real-time data, and generates preliminary doping formula recommendations. The expert system verifies and adjusts the preliminary plan based on industry knowledge and empirical rules; The optimization algorithm module uses genetic algorithm or particle swarm optimization algorithm based on the preliminary formula generated by the AI ​​intelligent decision-making module, and takes production cost, product quality index and production efficiency as multi-objective optimization functions to determine the final optimal doping formula under the conditions of meeting the production process constraints; The execution control module controls the high-precision electronic scale to weigh the dopant according to the optimal doping formula determined by the optimization algorithm module, and uses the PID control algorithm to adjust the feeding speed and start and stop status of the feeding equipment. After the weighing is completed, the automatic feeding equipment is controlled to add the dopant to the corresponding barrel, and the stirring device is coordinated to mix evenly.

2. The single crystal photovoltaic automatic doping and batching system based on intelligent AI according to claim 1 is characterized in that: In the data acquisition and preprocessing module, the data cleaning algorithm uses the 3σ criterion based on statistical principles to identify and remove outliers, and uses a sliding average filtering algorithm to remove noise data; The normalization algorithm uses linear normalization to map continuous data to the [0,1] interval, and uses one-hot encoding to process discrete data.

3. The single crystal photovoltaic automatic doping and batching system based on intelligent AI according to claim 1 is characterized in that: The deep learning neural network of the AI ​​intelligent decision-making module adopts a convolutional neural network (CNN) or a recurrent neural network (RNN) architecture, and the expert system stores industry knowledge and empirical rules in the form of a rule base.

4. The single crystal photovoltaic automatic doping and batching system based on intelligent AI according to claim 1 is characterized in that: The genetic algorithm of the optimization algorithm module simulates the selection, crossover and mutation operations in the biological evolution process, and encodes the doping formula parameters into chromosomes for iterative optimization; the particle swarm optimization algorithm simulates the foraging behavior of bird flocks, regards each possible doping formula as a particle, and searches for the optimal solution through information sharing and collaboration between particles.

5. The single crystal photovoltaic automatic doping and batching system based on intelligent AI according to claim 1 is characterized in that: The feeding equipment of the execution control module includes a screw conveyor and a vibrating feeder, and the stirring device includes a paddle stirrer and a screw stirrer.

6. A method for automatic doping and batching of single crystal photovoltaics based on intelligent AI, characterized in that: The following steps are involved: Data collection and preprocessing: The workshop furnace starts the material reporting procedure according to the production plan, and sends the production-related data to the data collection and preprocessing module through the industrial Internet. The module cleans and normalizes the data; AI intelligent formula generation: The preprocessed data is transmitted to the AI ​​intelligent decision-making module, and the deep learning neural network extracts and analyzes the features of the data to generate a preliminary doping formula recommendation plan. The expert system verifies and adjusts the preliminary plan. The optimization algorithm module uses genetic algorithm or particle swarm optimization algorithm based on the adjusted preliminary formula for multi-objective optimization to determine the final optimal doping formula; dopant metering and addition: The execution control module controls the high-precision electronic scale to weigh the dopant according to the optimized formula, and uses the PID control algorithm to adjust the feeding speed and start and stop status of the feeding equipment. After weighing, the automatic feeding equipment is controlled to add the dopant to the corresponding barrel, and the stirring device is started for uniform mixing.

7. The single crystal photovoltaic automatic doping and batching method based on intelligent AI according to claim 6, characterized in that: In the data collection and preprocessing steps, data cleaning uses the 3σ criterion based on statistical principles to identify and remove outliers, and uses a sliding average filtering algorithm to remove noise data; Normalization processing uses linear normalization method to map continuous data to the [0,1] interval, and uses one-hot encoding method to process discrete data.

8. The method for automatic doping and batching of single crystal photovoltaics based on intelligent AI according to claim 6, characterized in that: In the AI ​​intelligent formula generation step, the deep learning neural network uses a convolutional neural network (CNN) or a recurrent neural network (RNN) architecture to train historical production data, and the expert system stores industry knowledge and experience rules in the form of a rule base to revise the preliminary plan.

9. The method for automatic doping and batching of single-crystal photovoltaics based on intelligent AI according to claim 6, characterized in that: In the AI ​​intelligent formula generation step, the genetic algorithm simulates the selection, crossover and mutation operations in the biological evolution process, and encodes the doping formula parameters into chromosomes for iterative optimization; the particle swarm optimization algorithm simulates the foraging behavior of bird flocks, regards each possible doping formula as a particle, and searches for the optimal solution through information sharing and collaboration between particles.

10. The method for automatic doping and batching of single crystal photovoltaics based on intelligent AI according to claim 6, characterized in that: In the dopant metering and adding step, the feeding equipment includes a screw conveyor and a vibrating feeder, and the stirring device includes a paddle stirrer and a screw stirrer. The stirring device stirs according to a preset stirring speed and time.

Images