MES-based solar cell production and manufacturing execution method and system

Abstract

The invention provides an MES-based solar cell production and manufacturing execution method and system, and the method comprises the steps: collecting equipment states and technological parameters in real time, combining a bar code system to achieve the single-chip-level metadata association, constructing a standardized data architecture, building a multi-dimensional index, converting the expert experience into reasonable rules, and carrying out the calculation of the reasonable rules. The method comprises the following steps: simulating the influence of parameter combination on product performance, triggering an edge computing node to call a knowledge graph to perform parameter optimization and production simulation based on a dynamic threshold early warning mechanism, identifying process improvement opportunities through a batch tracing report and clustering analysis, forming knowledge base iteration update, predicting inventory risks in combination with a production plan and equipment load, and performing production management. According to the method, intelligent purchasing is automatically triggered, raw material distribution is optimized, technological parameter self-optimization, abnormity self-healing, quality traceability and supply chain collaboration are achieved through data driving, the production efficiency, the product consistency and the resource utilization rate are improved, and a whole-process digital solution is provided for photovoltaic manufacturing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of solar cell production control, and particularly relates to a solar cell production manufacturing execution method and system based on MES. BACKGROUND

[0002] Current solar cell production is faced with process optimization lag and insufficient supply chain response, etc. Traditional MES systems rely on discrete data collection and manual experience decision-making, and it is difficult to realize real-time linkage of equipment state, process parameters and product quality in the whole process, resulting in low equipment utilization, difficult traceability of process deviation and long abnormal processing period. In the prior art, production data is often scattered in different systems, lacking unified standardized processing and multi-dimensional indexing, restricting the transparency and intelligentization of the production process. Process parameter adjustment relies on expert experience, lacks knowledge graph reasoning based on historical data, and it is difficult to dynamically simulate the influence of parameter combination on product performance, such as the relationship between coating thickness and cell efficiency. At the same time, inventory management relies on static prediction model, without combining dynamic factors such as equipment load, supplier stability and volatility, which is easy to cause raw material shortage or redundancy. SUMMARY

[0003] The present application relates to the technical field of solar cell production control, and particularly relates to a solar cell production manufacturing execution method and system based on MES.

[0004] The present application relates to the technical field of solar cell production control, and particularly relates to a solar cell production manufacturing execution method and system based on MES.

[0005] A solar cell production manufacturing execution method based on MES, the technical scheme adopted is as follows: S1: Deploying Internet of Things sensors and edge computing devices on the production line, collecting equipment state data and process parameter data in real time, recording the metadata of each cell based on bar code, and associating with equipment data; S2: Based on equipment sensor data, image data and text data, unified conversion to standardized format supported by MES system, using time series database to store equipment state data, using relational database to store process parameters and batch metadata, establishing multi-dimensional index based on batch number + equipment ID + timestamp; S3: When the MES system detects that the process parameter deviates from the threshold, triggering the decision-making process of the edge computing node, calling the process knowledge graph to match the optimal parameter combination, based on the optimal parameter combination to simulate production, when the simulation result is not up to standard, triggering the secondary optimization process, and when the simulation result is up to standard, sending adjustment instructions to the equipment; S4: Configuring rule-based early warning, for the abnormal events that have occurred, the MES automatically generates a processing work order and analyzes the abnormal root cause, updates the process knowledge graph and the early warning rule library; S5: Record the production data of each batch of battery pieces based on the MES system, associate the equipment operation log, operator record, generate a visual traceability report by batch number, compare the performance data of different batches of products, and identify process improvement opportunities through cluster analysis; S6: Based on the production plan and equipment load, predict the risk of inventory shortage, automatically trigger procurement orders, and dynamically allocate solar cell production raw materials.

[0006] Further, the real-time collection of equipment state data and process parameter data in S1, based on the barcode record each piece of battery metadata, and associate with equipment data, including: Install industrial Internet of Things sensors including temperature sensors, vibration sensors, and current sensors on the equipment of the solar cell production line, including laser cutting machines, coating machines, and sintering furnaces. Real-time collection of equipment state data including temperature, vibration frequency, and motor current. Deploy edge computing devices such as industrial computers for local data preprocessing and real-time decision-making; Based on the real-time collection of equipment operation state data including sintering furnace temperature curve, coating machine speed, and cutting machine vibration amplitude, record the data by timestamp, collect process parameters including coating thickness, electrode width, and sintering time, and read the device set value and actual value through programmable logic controller. Based on each piece of solar cell, a unique barcode is assigned, containing batch number, production time, and operator number metadata.

[0007] Further, the S2 is uniformly converted to the standardized format supported by the MES system, the time series database is used to store the equipment state data, the relational database is used to store the process parameters and batch metadata, and the multi-dimensional index based on batch number + equipment ID + timestamp is established, including: Based on the OPC UA protocol, the raw data collected by the industrial Internet of Things sensors including temperature and vibration sensors is converted to the JSON format supported by the MES system. The defect detection image of the silicon wafer surface is labeled with defect types including cracks and bubbles. The operation log is extracted by extracting key fields including time, operator, and parameter modification value, and converted to structured table data. The converted data is de-duplicated, outlier removed, and field integrity verified; Time series database is used to store equipment state data including sintering furnace temperature curve and coating machine vibration frequency, indexed by timestamp. MySQL is used to store batch metadata including batch number, production time, operator ID, and process parameters including coating thickness and sintering time. SQL query is used to implement complex condition filtering; Create a composite index in the database as batch number + equipment ID + timestamp. Based on the batch number field, bind the metadata including operator record and equipment state data including sintering furnace temperature.

[0008] Further, the S3 calls the process knowledge graph to match the optimal parameter combination, and based on the optimal parameter combination, the production simulation is performed. When the simulation result is not up to standard, a secondary optimization process is triggered, including: The dynamic threshold range of the process parameters is preset in the MES system, and is self-adaptively adjusted according to historical data. When the sensor data exceeds the threshold, the MES pushes the abnormal event to the edge computing node through the OPC UA protocol, and automatically allocates the processing priority according to the abnormal type and the influence degree; The edge computing node calls the process knowledge graph, matches the historical successful cases, generates multiple parameter combinations based on the generated parameter combinations, calculates the theoretical product performance of each combination, embeds expert experience into the simulation process, dynamically adjusts the parameter weight, compares the simulation result with the historical data in the knowledge graph, and selects the optimal parameter combination scheme; The optimal parameter combination scheme is input to simulate the production process. When the simulation result is not up to standard, a secondary optimization is triggered to adjust the parameter combination and re-simulate until the simulation result is up to standard; The up-to-standard scheme is fed back to the MES system to generate an adjustment instruction, which is issued to the equipment through the OPC UA protocol. The edge computing node sends the adjustment instruction to the sintering furnace PLC through the Modbus TCP protocol, and monitors the execution state in real time.

[0009] Further, the S4 configures a rule-based early warning, updates the process knowledge graph and the early warning rule library, including: The MES system presets multi-level early warning rules based on the process influence degree, sets the rule priority, dynamically adjusts the threshold based on the historical data of the knowledge graph, divides the abnormal events into device fault type, process deviation type and operation error type, and binds the corresponding processing flow; When the MES detects an abnormal event, an electronic work order is automatically created, including event description, associated data and emergency degree information. The work order is distributed to the corresponding person in charge according to the abnormal classification and the processing personnel role; The root cause report is generated based on the cross-analysis positioning of the root cause of the time series database, the relational database and the knowledge graph in S2; The historical early warning records are analyzed regularly, the false positive rate and the false negative rate of the rules are calculated, the potential rules are mined from the historical data based on the knowledge graph, and the feasibility of the mined rules is verified through trial operation.

[0010] Further, the S5 generates a visual traceability report according to the batch number, compares the product performance data of different batches, identifies the process improvement opportunity through cluster analysis, including: Generate structured reports based on batch number, including production profile, process parameter trend chart, quality index comparison and abnormal event record, production profile includes batch number, production time, equipment ID, operator, process parameter trend chart includes time series curve of parameters, quality index comparison includes efficiency, crack rate index of this batch of batteries compared with historical batches, abnormal event record includes warning work order and root cause analysis results associated through S4; Based on the multi-dimensional index of S2, extract production data by batch number, compare key performance indicators with process parameters, group historical batch data to output high, medium and low efficiency batch groups, identify parameters affecting efficiency, combine clustering results with process knowledge graph to generate optimization suggestions; Push the generated optimization suggestions to the edge computing nodes of S3, trigger parameter adjustment process and simulation verification, combine clustering analysis results with root cause analysis results of S4, update process knowledge graph, update S4's warning rule library according to abnormal patterns found by clustering analysis.

[0011] Further, the S6 predicts inventory shortage risk based on production plan and equipment load, automatically triggers procurement order, including: S61: Obtain current and future N-day production plan from MES system, real-time collect equipment utilization through S1's Internet of Things sensors , combined with the batch traceability report and clustering analysis in S5, obtain the amount of raw materials required for producing a unit product and the current raw material inventory , safety stock threshold , historical consumption volatility , historical average consumption , integrate supplier delivery cycle and supplier stability score , Based on historical on-time delivery rate calculation; S62: Based on inventory shortage risk = demand gap / supply capacity x uncertainty penalty factor, establish inventory prediction formula , where is the inventory shortage risk index at the current time t, is the production plan output on the i-th day, is the volatility weight coefficient, take 0.3, through numerical calculation, the difference between total demand and current inventory in the future N days is obtained, combined with supplier stability and historical volatility, the shortage risk is quantified, when , it means that the inventory may not meet the demand, procurement is needed, based on the root cause analysis results of S4, real-time correct production plan , based on knowledge graph reasoning, dynamically update raw material consumption rate ; S63: When When this occurs, the procurement process is triggered to calculate the required replenishment quantity. , , A safety stock factor of 1.75 is used to cover volatility buffering, based on supplier stability scores. and delivery cycle Preferred selection and If no supplier meets the criteria, the shortest supplier is selected. If no supplier meets the criteria, the S4 warning process is triggered, a purchase order is automatically generated, and the order is pushed to the supplier through the system. The order status is synchronized to the MES system in real time. S64: High-priority orders are allocated raw materials first, and production lines with low equipment load are allocated raw materials first to balance capacity.

[0012] Furthermore, a MES-based solar cell manufacturing execution system is provided to implement the MES-based solar cell manufacturing execution method described above. The MES-based solar cell manufacturing execution system includes: an equipment status module, a process parameter tracking module, a process optimization module, and a supply chain collaboration module. Equipment Status Module: Install IoT sensors to collect data in real time, deploy edge computing devices for local preprocessing, and store the data in a time-series database by timestamp. The PLC directly reads the device's set values ​​and actual values, and combines a barcode system to assign unique metadata to each battery. Image data and text data are converted into structured tables and stored in a relational database. Process parameter tracking module: When MES detects that a parameter deviates from the threshold, it triggers the edge computing node to call the knowledge graph, match historical successful cases, dynamically adjust the parameter weights based on expert experience, simulate the production process and select the optimal solution, configure rule-based early warning, and for abnormal events that have occurred, MES automatically generates processing work orders and analyzes the root cause of the abnormality, and updates the process knowledge graph and early warning rule base. Process optimization module: Generates structured reports by batch number, including production overview, process parameter trend chart, quality index comparison and abnormal event records. Based on multidimensional index, it compares data from different batches, groups high / low efficiency batches, combines clustering results with knowledge graph to generate optimization suggestions, and pushes them to the process line management module for simulation verification. Supply chain collaboration module: Obtain production plans and equipment load from MES, combine historical consumption volatility, supplier delivery cycle and stability score to calculate shortage risk index. When the shortage risk index is >1, trigger the procurement process, calculate the replenishment quantity, prioritize suppliers with stability score ≥0.8 and shortest supplier delivery cycle, generate purchase orders and push them. High priority orders are allocated raw materials first, and production lines with low equipment utilization are allocated first to balance capacity. The allocation status is monitored in real time. Based on order priority and delivery date, and combined with capacity data from the production planning management module, the scheduling is dynamically adjusted. Based on inventory forecast results, raw material procurement is planned in advance. Based on equipment utilization and inventory forecast results, the production plan is dynamically adjusted. Combined with the early warning rule base, the production plan is corrected. Based on sensor data, the equipment status is monitored in real time, and tasks are prioritized for equipment with low load.

[0013] The beneficial effects of this invention are: based on knowledge graph-based process rule reasoning and simulation verification, it can quickly match the optimal parameter combination and adjust equipment parameters in real time, reduce rework or scrap rates caused by parameter deviations, shorten the trial and error cycle, and improve output per unit time. Through batch performance comparison and cluster analysis, it can quickly identify key process parameters that affect quality and optimize the process accordingly. When process parameters deviate from the threshold, the system automatically triggers the decision-making process of the edge computing node, combines the knowledge graph to simulate and optimize the solution, and generates early warning work orders and root cause analysis reports to shorten the problem handling time and reduce the defect rate. Through the construction of the knowledge graph and the rule mining of historical data, the expert experience and historical success cases are structured to form a dynamic process knowledge base, continuously optimize production rules, and improve product quality stability. By combining production planning and inventory shortage risk models with historical volatility and supplier stability scores, we can achieve precise replenishment and high-priority allocation of raw materials, reduce the risk of inventory backlog, ensure production continuity, and reduce equipment idle time and resource waste by combining dynamic scheduling and equipment load balancing, thereby improving the overall operating efficiency of the production line. Attached Figure Description

[0014] Figure 1 This is a flowchart of a MES-based manufacturing execution method for solar cell production. Figure 2 This is a module diagram of a manufacturing execution system (MES) for the production of solar cells. Detailed Implementation

[0015] The present invention will be further described clearly and completely below, but the scope of protection of the present invention is not limited thereto.

[0016] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0017] Example 1 A manufacturing execution method for solar cell production based on MES, the technical solution of which is as follows: S1: Deploy IoT sensors and edge computing devices on the production line to collect equipment status data and process parameter data in real time, record metadata of each battery based on barcodes, and associate it with equipment data; S2: Based on equipment sensor data, image data, and text data, it is uniformly converted into a standardized format supported by the MES system. Time-series database is used to store equipment status data, and relational database is used to store process parameters and batch metadata. A multi-dimensional index based on batch number + equipment ID + timestamp is established. S3: When the MES system detects that the process parameters deviate from the threshold, it triggers the decision-making process of the edge computing node, calls the process knowledge graph to match the optimal parameter combination, performs production simulation based on the optimal parameter combination, and triggers a secondary optimization process when the simulation result does not meet the standard, and sends an adjustment instruction to the equipment when the simulation result meets the standard. S4: Configure rule-based early warning. For abnormal events that have occurred, MES automatically generates processing work orders, analyzes the root cause of the abnormality, and updates the process knowledge graph and early warning rule base. S5: Based on the MES system, record the production data of each batch of battery cells, and link it with equipment operation logs and operator records. Generate a visual traceability report by batch number, compare the product performance data of different batches, and identify process improvement opportunities through cluster analysis. S6: Based on production plans and equipment load, predict inventory shortage risks, automatically trigger purchase orders, and dynamically allocate raw materials for solar cell production.

[0018] refer to Figure 1 The diagram shown is a flowchart of a MES-based manufacturing execution method for solar cell production.

[0019] Furthermore, in S1, real-time acquisition of equipment status data and process parameter data is performed, metadata for each battery is recorded based on barcodes, and associated with equipment data, including: Industrial IoT sensors, including temperature sensors, vibration sensors, and current sensors, are installed on equipment such as laser cutting machines, coating machines, and sintering furnaces in the solar cell production line to collect real-time equipment status data, including temperature, vibration frequency, and motor current. Edge computing devices and industrial control computers are deployed for local data preprocessing and real-time decision-making. The system collects real-time equipment operating status data based on sensors, including sintering furnace temperature curves, coating machine speed, and cutting machine vibration amplitude, and records the data with timestamps to ensure data timeliness. The process parameters collected include coating thickness, electrode width, and sintering time. The set values ​​and actual values ​​of the equipment are directly read by the programmable logic controller to ensure data accuracy. Each solar cell is assigned a unique barcode containing metadata such as batch number, production time, and operator number.

[0020] Furthermore, in S2, the data is uniformly converted to a standardized format supported by the MES system. A time-series database is used to store device status data, and a relational database is used to store process parameters and batch metadata. A multi-dimensional index based on batch number + device ID + timestamp is established, including: Based on the OPC UA protocol, raw data collected by industrial IoT sensors, including temperature and vibration sensors (e.g., sintering furnace temperature 850℃), is converted into JSON format supported by the MES system. Defect types, such as cracks and bubbles, are labeled on silicon wafer surface defect detection images. Operation logs, such as "Operator A changed the coating machine speed to 500rpm on 2025-07-06 10:00", are processed by extracting key fields including time, operator, and parameter modification value, and converted into structured table data. The converted data is then deduplicated, outliers are removed (e.g., data points where temperature changes exceed a threshold), and field integrity is verified (e.g., missing batch numbers are automatically marked). The system uses a time-series database to store device status data, including sintering furnace temperature curves and coating machine vibration frequencies, indexed by timestamps. It also uses MySQL to store batch metadata, including batch number, production time, operator ID, and process parameters, including coating thickness and sintering time. Complex condition filtering is achieved based on SQL queries. Create a composite index in the database with batch number + device ID + timestamp, such as batch number: 20250706-123456, device ID: sintering furnace-L1-01, timestamp: 2025-07-06 10:00:00; Based on the batch number field, metadata including operator records is bound to equipment status data including sintering furnace temperature. For example, when querying batch number 20250706-123456, the sintering furnace temperature curve, coating machine speed record and operator modification log of that batch are automatically associated.

[0021] Furthermore, in step S3, the process knowledge graph is invoked to match the optimal parameter combination, and production simulation is performed based on the optimal parameter combination. When the simulation results do not meet the requirements, a secondary optimization process is triggered, including: The MES system presets dynamic threshold ranges for process parameters, such as sintering temperature 830℃±10℃, and adaptively adjusts them based on historical data. For example, by analyzing the process window of a batch of batteries through a knowledge graph, when sensor data such as sintering furnace temperature exceeds the threshold, the MES pushes the abnormal event to the edge computing node, i.e., the industrial gateway deployed near the sintering furnace, through the OPC UA protocol. Based on the type of abnormality, such as excessively high temperature vs. excessively low pressure, and the degree of impact, such as slight deviation vs. severe abnormality, the system automatically assigns processing priorities. Severe abnormalities directly trigger an emergency shutdown. Edge computing nodes call upon the process knowledge graph to match historical successful cases, identifying high-efficiency batteries corresponding to the combination of sintering temperature 830℃ and coating thickness 600nm. Based on multiple parameter combinations such as sintering temperatures of 825℃, 830℃, and 835℃, the theoretical product performance of each combination, including efficiency and crack rate, is calculated. Expert experience that the crack rate decreases by 0.5% for every 5℃ decrease in sintering temperature is embedded into the simulation process, dynamically adjusting parameter weights. The simulation results are compared with historical data in the knowledge graph to select the optimal parameter combination scheme, such as choosing the combination with the highest efficiency and a crack rate of 0. The optimal parameter combination scheme is input to simulate the production process, such as the effect of temperature curves on silicon wafer cracks. When the simulation results are unsatisfactory and the efficiency is lower than the target value, secondary optimization is triggered. The parameter combination is adjusted to increase the coating thickness to 620nm to compensate for the efficiency loss. The simulation is repeated until the simulation results meet the target, that is, the efficiency is ≥20%. The compliance solution is fed back to the MES system, generating adjustment instructions such as changing the temperature setpoint of sintering furnace L1-01 to 830℃. These instructions are then sent to the equipment via the OPC UA protocol. The edge computing node sends the adjustment instructions to the sintering furnace PLC via the Modbus TCP protocol to set the temperature to 830℃ and monitors the execution status in real time to ensure the current temperature is stable.

[0022] Furthermore, S4 configures rule-based early warning, updates the process knowledge graph and early warning rule base, including: The MES system pre-sets multi-level early warning rules: sintering temperature > 850℃ → Level 1 warning, coating thickness < 500nm → Level 2 warning. The priority of the rules is set based on the degree of process impact. The risk of cracking is higher than the risk of efficiency decline. Combined with historical data from the knowledge graph, such as the optimal sintering temperature range of a certain batch of batteries, the threshold is dynamically adjusted. The temperature tolerance is automatically relaxed according to the aging trend of the equipment. Abnormal events are classified into equipment failures such as sensor failure, process deviations such as parameter drift, and operational errors such as human input errors, and are bound to corresponding processing procedures. Equipment failures directly trigger shutdown. When the MES detects an abnormal event, such as the sintering temperature exceeding the standard in S3, it automatically creates an electronic work order, which includes an event description, related data, and urgency information. The event description is, for example, the temperature of sintering furnace L1-01 is 855℃, which exceeds the threshold of 850℃. The related data includes batch number, equipment ID, and timestamp. The urgency level is, for example, high, which requires immediate processing, and low, which can be postponed. The work order is assigned to the corresponding responsible person according to the abnormality classification and the role of the handling personnel, such as process engineer or equipment maintenance personnel. Based on the time-series database, relational database, and knowledge graph described in S2, such as the sintering furnace temperature curve, the relational database such as process parameters, and the knowledge graph such as excessive temperature → crack risk, cross-analysis is performed to locate the root cause and generate a root cause report, such as: Problem: Sintering temperature exceeds the standard → Cause 1: PLC control program error → Cause 2: Software version not updated → Cause 3: Not implementing the latest process specifications. Regularly analyze historical early warning records, such as all work orders in the past 30 days, and calculate the false alarm rate and false negative rate of the rules. The false alarm rate is when a rule is triggered but there is no actual anomaly, and the false negative rate is when an anomaly occurs but the rule is not triggered. Based on the knowledge graph, potential rules are mined from historical data. For example, if a negative correlation is found between sintering temperature and coating thickness, an early warning rule is automatically generated that the coating thickness needs to be increased by 10nm when the temperature increases by 5℃. The feasibility of the mined rules is verified by trial operation.

[0023] Furthermore, in step S5, a visual traceability report is generated by batch number, and product performance data from different batches is compared. Cluster analysis is used to identify opportunities for process improvement, including: A structured report is generated based on the batch number, which includes production overview, process parameter trend chart, quality index comparison and abnormal event record. The production overview includes batch number, production time, equipment ID and operator. The process parameter trend chart includes time series curves of key parameters such as sintering temperature and coating thickness. The quality index comparison includes the comparison of the efficiency and crack rate of this batch of batteries with historical batches. The abnormal event record includes early warning work orders and root cause analysis results linked through S4. Based on the S2 multidimensional index, production data is extracted by batch number. Key performance indicators such as efficiency and crack rate are compared with process parameters such as sintering temperature and coating time. Historical batch data are grouped and output as high, medium and low efficiency batch groups. Parameters affecting efficiency are identified. The clustering results are combined with the process knowledge graph to generate optimization suggestions. The generated optimization suggestions are pushed to the edge computing node of S3, triggering parameter adjustment processes such as modifying the coating time and simulating verification. The cluster analysis results are merged with the root cause analysis results of S4, and the process knowledge graph is updated, such as adding a causal relationship that coating time of 35 seconds → efficiency improvement of 0.5%. Based on the abnormal patterns found by cluster analysis, the early warning rule base of S4 is updated, such as adding a warning that coating time <30 seconds → triggers efficiency decline.

[0024] Furthermore, in S6, based on production planning and equipment load, the risk of inventory shortage is predicted, and purchase orders are automatically triggered, including: S61: Obtain the current and N-day production plan from the MES system, and collect equipment utilization in real time through S1's IoT sensors. Combining the batch traceability report and cluster analysis described in S5, the amount of raw materials required for the production unit's products can be obtained. and current raw material inventory Safety stock threshold Historical consumption volatility Historical average consumption For example, the polysilicon consumption volatility over the past 30 days was 5%, and the delivery cycle of integrated suppliers was considered. Rate the time from order placement to delivery and supply stability. For example, a score of 0-1, where 1 is the most stable. Calculated based on historical on-time delivery rates, for example, if a supplier's on-time delivery rate is 95%, the corresponding... ; S62: Based on the formula for inventory shortage risk = demand gap / supply capacity × uncertainty penalty factor, establish an inventory forecasting formula. ,in This represents the inventory shortage risk index at time t; the higher the value, the higher the risk. Let i be the planned production output for day i. With a volatility weighting factor of 0.3, the difference between total demand and current inventory over the next N days is calculated by substituting numerical values. This difference is then weighted by supplier stability and historical volatility to quantify shortage risk. This indicates that inventory may not be able to meet demand, requiring procurement to be triggered, and the production plan should be adjusted in real time based on the root cause analysis results from S4. Based on knowledge graph reasoning, the raw material consumption rate is dynamically updated. Automatic adjustment when equipment aging leads to increased material loss. ;

[0025] If 300 tons of silicon material are needed in the next 3 days, and the current inventory is only 200 tons, supplier stability is a concern. Historical consumption volatility ,but Historical average consumption tons / day, then This indicates insufficient inventory, requiring an immediate purchase order to be triggered; S63: When When this occurs, the procurement process is triggered to calculate the required replenishment quantity. , , A safety stock factor of 1.75 is used to cover volatility buffering, based on supplier stability scores. and delivery cycle Preferred selection and If no suitable supplier is found, the S4 alert process is triggered, such as generating an emergency supplier sourcing work order and automatically generating a purchase order containing material number and quantity. Delivery time The system pushes the information to the supplier, and the order status is synchronized to the MES system in real time, such as "order placed", "in transit", and "in stock". S64: High-priority orders, such as urgent customer orders, are allocated raw materials first; production lines with low equipment load, such as those monitored by S1, have low equipment utilization. <70%, prioritize the allocation of raw materials to balance production capacity.

[0026] Furthermore, a MES-based solar cell manufacturing execution system is provided to implement the MES-based solar cell manufacturing execution method described above. The MES-based solar cell manufacturing execution system includes: an equipment status module, a process parameter tracking module, a process optimization module, and a supply chain collaboration module. Equipment Status Module: Install IoT sensors to collect data in real time, deploy edge computing devices for local preprocessing, and store the data in a time-series database by timestamp. The PLC directly reads the device's set values ​​and actual values, and combines a barcode system to assign unique metadata to each battery. Image data and text data are converted into structured tables and stored in a relational database. Process parameter tracking module: When MES detects that a parameter deviates from the threshold, it triggers the edge computing node to call the knowledge graph, match historical successful cases, dynamically adjust the parameter weights based on expert experience, simulate the production process and select the optimal solution, configure rule-based early warning, and for abnormal events that have occurred, MES automatically generates processing work orders and analyzes the root cause of the abnormality, and updates the process knowledge graph and early warning rule base. Process optimization module: Generates structured reports by batch number, including production overview, process parameter trend chart, quality index comparison and abnormal event records. Based on multidimensional index, it compares data from different batches, groups high / low efficiency batches, combines clustering results with knowledge graph to generate optimization suggestions, and pushes them to the process line management module for simulation verification. Supply chain collaboration module: Obtain production plans and equipment load from MES, combine historical consumption volatility, supplier delivery cycle and stability score to calculate shortage risk index. When the shortage risk index is >1, trigger the procurement process, calculate the replenishment quantity, prioritize suppliers with stability score ≥0.8 and shortest supplier delivery cycle, generate purchase orders and push them. High priority orders are allocated raw materials first, and production lines with low equipment utilization are allocated first to balance capacity. The allocation status is monitored in real time. Based on order priority and delivery date, and combined with capacity data from the production planning management module, the scheduling is dynamically adjusted. Based on inventory forecast results, raw material procurement is planned in advance. Based on equipment utilization and inventory forecast results, the production plan is dynamically adjusted. Combined with the early warning rule base, the production plan is corrected. Based on sensor data, the equipment status is monitored in real time, and tasks are prioritized for equipment with low load.

[0027] refer to Figure 2 The diagram shown is a module diagram of a MES-based solar cell manufacturing execution system.

[0028] Example 2 In the coating process of the PERC cell production line, infrared temperature sensors with an accuracy of ±0.5℃, vibration sensors with a range of 0-2000Hz, and edge computing industrial control computers are installed on the PECVD equipment. Each cell is assigned a unique barcode, such as PC20250727-185634, which is associated with metadata such as batch number, PECVD-Unit3 equipment ID, and OP089 operator ID. The raw sensor data is converted into JSON format via the OPC UA protocol. Defect images of silicon wafers are labeled with defect types including scratches / contamination. Process parameters are stored in a MySQL database, and equipment status time-series data is stored in InfluxDB. A multi-dimensional index is established: batch number + equipment ID + timestamp, PC20250727-185634_PECVD-Unit3_20250727103025.

[0029] When the PECVD chamber temperature exceeds the dynamic threshold (set value 480±10℃, historical adaptive adjustment range 475-490℃), the MES pushes the abnormal event to the edge node with high priority. It calls the process knowledge graph containing historical data and matches successful cases. Case 1: Temperature 485℃ + Pressure 1.8 Torr → Battery efficiency 23.15%; Case 2: Temperature 480℃ + Pressure 1.6 Torr → Crack rate 0.12%. Three sets of parameter combinations are generated and simulated. The combinations, temperature (℃), pressure (Torr), simulated efficiency, and simulated crack rate are: A: 482, 1.7, 23.08%, 0.09%; B: 478, 1.9, 22.95%, 0.07%; C: 485, 1.75, 23.12%, 0.05%. The optimal combination is C, with efficiency > 23.1% and crack rate < 0.1%. After the simulation results meet the standards, the edge node sends the command SET_PARAM to the PECVD via Modbus TCP. temp=485, pressure=1.75, real-time temperature monitoring stabilized at 485±2℃.

[0030] A report was generated based on batch number PC20250727-185634. Process parameter trends: coating thickness increased from 78nm to 82nm. Quality comparison: efficiency 23.12% vs. historical average 22.8%, crack rate 0.05% vs. historical average 0.15%. Clustering of 200 historical batches revealed characteristics of the high-efficiency group (efficiency > 23%): coating thickness 80±2nm + chamber cleanliness level ≤ 5. The main cause of the low-efficiency group was RF power fluctuation > 5%, originating from power module aging. Optimization suggestions were generated: upgrade the PECVD RF power module, calibrate power every 2000 wafers, and add a rule to the warning library: RF power fluctuation > 5% → trigger equipment maintenance work order.

[0031] Production plan for the next 7 days: D1 output 24,000 wafers, D2-D7 168,000 wafers, total output 192,000 wafers, polysilicon consumption 3.8g / wafer = 0.0038kg / wafer, current polysilicon inventory =680kg, safety stock =500kg, historical volatility =6%, Supplier A's stability =0.92, delivery cycle L=3 days, historical average consumption calculated based on average daily output of 27,428 pieces × 0.0038 kg / piece. =104.2kg, Total demand =192000×0.0038=729.6kg, Demand gap =729.6−680=49.6kg Without triggering a purchase, the system does not generate a purchase order; it only monitors inventory. =680>500, no intervention is needed.

[0032] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0033] This invention provides a manufacturing execution method and system for solar cell production based on MES (Manufacturing Execution System). By collecting equipment status and process parameters in real time, and combining this with a barcode system to achieve single-cell-level metadata association, a standardized data architecture is constructed to establish a multi-dimensional index. Expert experience is transformed into inferable rules, simulating the impact of parameter combinations on product performance. Based on a dynamic threshold early warning mechanism, edge computing nodes are triggered to call the knowledge graph for parameter optimization and production simulation. Batch traceability reports and cluster analysis identify process improvement opportunities, forming an iteratively updated knowledge base. Combined with production plans and equipment load prediction of inventory risks, intelligent procurement is automatically triggered to optimize raw material allocation. Through data-driven approaches, process parameters are self-optimized, anomalies are self-healed, quality is traceable, and supply chain collaboration is achieved, improving production efficiency, product consistency, and resource utilization, providing a full-process digital solution for photovoltaic manufacturing.

Claims

1. A method for manufacturing solar cells based on MES, characterized in that, include: S1: Deploy IoT sensors and edge computing devices on the production line to collect equipment status data and process parameter data in real time, record metadata of each battery based on barcodes, and associate it with equipment data; S2: Based on equipment sensor data, image data, and text data, it is uniformly converted into a standardized format supported by the MES system. Time-series database is used to store equipment status data, and relational database is used to store process parameters and batch metadata. A multi-dimensional index based on batch number + equipment ID + timestamp is established. S3: When the MES system detects that the process parameters deviate from the threshold, it triggers the decision-making process of the edge computing node, calls the process knowledge graph to match the optimal parameter combination, performs production simulation based on the optimal parameter combination, and triggers a secondary optimization process when the simulation result does not meet the standard, and sends an adjustment instruction to the equipment when the simulation result meets the standard. S4: Configure rule-based early warning. For abnormal events that have occurred, MES automatically generates processing work orders, analyzes the root cause of the abnormality, and updates the process knowledge graph and early warning rule base. S5: Based on the MES system, record the production data of each batch of battery cells, and link it with equipment operation logs and operator records. Generate a visual traceability report by batch number, compare the product performance data of different batches, and identify process improvement opportunities through cluster analysis. S6: Based on production plans and equipment load, predict inventory shortage risks, automatically trigger purchase orders, and dynamically allocate raw materials for solar cell production.

2. The MES-based solar cell manufacturing execution method as described in claim 1, characterized in that, In S1, real-time acquisition of equipment status data and process parameter data is performed. Metadata for each battery is recorded based on barcodes and associated with equipment data, including: Industrial IoT sensors are installed on equipment in the solar cell production line to collect equipment status data in real time, and edge computing devices and industrial control computers are deployed. The system collects real-time equipment operating status data using sensors and records the data with timestamps. It also collects process parameters and directly reads the equipment's set values ​​and actual values ​​through a programmable logic controller. Each solar cell is assigned a unique barcode containing metadata.

3. The method for manufacturing solar cells based on MES as described in claim 1, characterized in that, In S2, the data is uniformly converted to a standardized format supported by the MES system. A time-series database is used to store device status data, and a relational database is used to store process parameters and batch metadata. A multi-dimensional index based on batch number + device ID + timestamp is established, including: The raw data collected by industrial IoT sensors is converted into a format supported by the MES system. Defect types are labeled on silicon wafer surface defect detection images. Operation logs are processed by extracting fields including time, operator, and parameter modification values, and then converted into structured table data. The converted data is deduplicated, outliers are removed, and field integrity is verified. A time-series database is used to store equipment status data, including sintering furnace temperature curves and coating machine vibration frequencies. The data is indexed by timestamps. MySQL is used to store batch metadata and process parameters. Complex condition filtering is implemented based on SQL queries. A composite index is created in the database as batch number + equipment ID + timestamp. The metadata is bound to the equipment status data based on the batch number field.

4. The MES-based solar cell manufacturing execution method as described in claim 1, characterized in that, In step S3, the process knowledge graph is called to match the optimal parameter combination, and production simulation is performed based on the optimal parameter combination. When the simulation results do not meet the standards, a secondary optimization process is triggered, including: The MES system presets a dynamic threshold range for process parameters and adaptively adjusts it based on historical data. When sensor data exceeds the threshold, the MES pushes the abnormal event to the edge computing node and automatically assigns processing priorities based on the type and degree of impact of the abnormality. Edge computing nodes call upon the process knowledge graph, match historical success cases, calculate the theoretical product performance of each combination based on the generated multiple sets of parameters, embed expert experience into the simulation process, dynamically adjust parameter weights, compare the simulation results with historical data in the knowledge graph, and select the optimal parameter combination scheme. The optimal parameter combination scheme is input to simulate the production process. When the simulation results are not satisfactory, secondary optimization is triggered, the parameter combination is adjusted, and the simulation is repeated until the simulation results are satisfactory. The compliance plan is fed back to the MES system, which generates adjustment instructions and sends them to the equipment. The edge computing node sends the adjustment instructions to the sintering furnace PLC and monitors the execution status in real time.

5. The MES-based manufacturing execution method for solar cells as described in claim 1, characterized in that, S4 configures rule-based early warnings and updates the process knowledge graph and early warning rule base, including: In the MES system, multi-level early warning rules are preset, rule priorities are set based on the degree of process impact, and thresholds are dynamically adjusted by combining historical data from the knowledge graph. Abnormal events are classified into equipment failure, process deviation, and operational error categories, and corresponding processing procedures are bound to them. When MES detects an abnormal event, it automatically creates an electronic work order, which includes an event description, related data, and urgency information. The work order is then assigned to the corresponding responsible person based on the anomaly classification and the role of the person handling the event. Based on the time-series database, relational database, and knowledge graph described in S2, cross-analysis is performed to locate the root cause and generate a root cause report; Regularly analyze historical early warning records, calculate the false alarm rate and false negative rate of the rules, mine potential rules from historical data based on knowledge graphs, and conduct trial runs to verify the feasibility of the mined rules.

6. The MES-based manufacturing execution method for solar cells as described in claim 1, characterized in that, In S5, a visual traceability report is generated by batch number, and product performance data from different batches are compared. Cluster analysis is used to identify opportunities for process improvement, including: A structured report is generated based on the batch number, which includes production overview, process parameter trend chart, quality index comparison and abnormal event record. The production overview includes batch number, production time, equipment ID and operator. The process parameter trend chart includes the time series curve of the parameter. The quality index comparison includes the comparison of the efficiency and crack rate of this batch of batteries with historical batches. The abnormal event record includes the early warning work orders and root cause analysis results associated with S4. Based on the S2 multidimensional index, production data is extracted by batch number, key performance indicators and process parameters are compared, historical batch data are grouped into high, medium and low efficiency batch groups, parameters affecting efficiency are identified, and the clustering results are combined with the process knowledge graph to generate optimization suggestions. The generated optimization suggestions are pushed to the edge computing node of S3, triggering the parameter adjustment process and simulation verification. The cluster analysis results are merged with the root cause analysis results of S4 to update the process knowledge graph. Based on the abnormal patterns found by the cluster analysis, the early warning rule base of S4 is updated.

7. The MES-based manufacturing execution method for solar cells as described in claim 1, characterized in that, S6, based on production planning and equipment load, predicts inventory shortage risks and automatically triggers purchase orders, including: S61: Obtain the current and N-day production plan from the MES system, and collect equipment utilization in real time through S1's IoT sensors. Combining the batch traceability report and cluster analysis described in S5, the amount of raw materials required for the production unit's products can be obtained. and current raw material inventory Safety stock threshold Historical consumption volatility Historical average consumption Integrate supplier delivery cycles and supply stability score , Calculated based on historical on-time delivery rates; S62: Establishing an inventory forecasting formula ,in Let t be the inventory shortage risk index at the current time. Let i be the planned production output for day i. The volatility weighting coefficient is set to 0.

3. This requires triggering procurement, and production plans should be adjusted in real time based on the root cause analysis results from S4. Based on knowledge graph reasoning, the raw material consumption rate is dynamically updated. ; S63: When When this occurs, the procurement process is triggered to calculate the required replenishment quantity. , , A safety stock factor of 1.75 is used to cover volatility buffering, based on supplier stability scores. and delivery cycle Preferred selection and If no supplier meets the criteria, the shortest supplier is selected. If no supplier meets the criteria, the S4 warning process is triggered, a purchase order is automatically generated, and the order is pushed to the supplier through the system. The order status is synchronized to the MES system in real time. S64: High-priority orders are allocated raw materials first, and production lines with low equipment load are allocated raw materials first to balance capacity.

8. A MES-based manufacturing execution system for solar cell production, characterized in that, To implement the MES-based solar cell manufacturing execution method as described in any one of claims 1-7, the MES-based solar cell manufacturing execution system includes: an equipment status module, a process parameter tracking module, a process optimization module, and a supply chain collaboration module. Equipment Status Module: Install IoT sensors to collect data in real time, deploy edge computing devices for local preprocessing, and store the data in a time-series database by timestamp. The PLC directly reads the device's set values ​​and actual values, and combines a barcode system to assign unique metadata to each battery. Image data and text data are converted into structured tables and stored in a relational database. Process parameter tracking module: When MES detects that a parameter deviates from the threshold, it triggers the edge computing node to call the knowledge graph, match historical successful cases, dynamically adjust the parameter weights based on expert experience, simulate the production process and select the optimal solution, configure rule-based early warning, and for abnormal events that have occurred, MES automatically generates processing work orders and analyzes the root cause of the abnormality, and updates the process knowledge graph and early warning rule base. Process optimization module: Generates structured reports by batch number, including production overview, process parameter trend chart, quality index comparison and abnormal event records. Based on multidimensional index, it compares data from different batches, groups high / low efficiency batches, combines clustering results with knowledge graph to generate optimization suggestions, and pushes them to the process line management module for simulation verification. Supply chain collaboration module: Obtain production plans and equipment load from MES, combine historical consumption volatility, supplier delivery cycle and stability score to calculate shortage risk index. When the shortage risk index is >1, trigger the procurement process, calculate the replenishment quantity, prioritize suppliers with stability score ≥0.8 and shortest supplier delivery cycle, generate purchase orders and push them. High priority orders are allocated raw materials first, and production lines with low equipment utilization are allocated first to balance capacity. The allocation status is monitored in real time. Based on order priority and delivery date, and combined with capacity data from the production planning management module, the scheduling is dynamically adjusted. Based on inventory forecast results, raw material procurement is planned in advance. Based on equipment utilization and inventory forecast results, the production plan is dynamically adjusted. Combined with the early warning rule base, the production plan is corrected. Based on sensor data, the equipment status is monitored in real time, and tasks are prioritized for equipment with low load.

Images