Optimization method and device of industrial silicon production process, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an optimization method and device of an industrial silicon production process, electronic equipment and a storage medium. The method comprises the following steps: constructing a knowledge graph of an industrial silicon production process based on data related to industrial silicon production, and in a subsequent industrial silicon production process, if at least one item of process data of industrial silicon production triggers production process optimization, generating an optimization strategy of an optimization target according to the knowledge graph and data related to the optimization target, the technical parameters of the industrial silicon production are adjusted according to the optimization strategy, and the generation efficiency of the optimization strategy is improved and the real-time requirement of the production process optimization is met through the support of the knowledge graph on the complex reasoning and decision-making of the production process, so that the real-time control and optimization of the industrial silicon production are supported, and the optimization target is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to an industrial silicon production process optimization method and device, electronic equipment and storage medium. BACKGROUND

[0002] Industrial silicon (also known as silicon metal) is a crucial raw material in modern industry, widely used in electronics, chemical industry, metallurgy and other fields. With the continuous development of science and technology and industry, the demand for high-purity and high-performance industrial silicon is increasing.

[0003] In recent years, researchers have improved the smelting process, optimized the reaction conditions, and improved the materials and equipment to optimize the production process of industrial silicon. The application of automation and intelligent technology in industrial silicon production has gradually increased. Machine learning and big data analysis in industrial silicon production mainly focus on optimizing production parameters, fault diagnosis and process prediction. Deep learning and neural network models can handle complex nonlinear systems in industrial silicon production, achieving high-precision process modeling and optimization.

[0004] However, the optimization algorithm of the industrial silicon production process is complex, requiring a large amount of computing resources and time, which is not sufficient to support real-time control and optimization of industrial silicon production. SUMMARY

[0005] Therefore, the present application provides an industrial silicon production process optimization method, device, electronic equipment and storage medium, which can quickly realize production process optimization based on a knowledge graph during industrial silicon production, thereby achieving the purposes of improving the production efficiency of industrial silicon, reducing production costs and reducing environmental impact.

[0006] According to a first aspect of an embodiment of the present application, an industrial silicon production process optimization method is provided, the method comprising: based on a data source related to industrial silicon production, constructing a knowledge graph corresponding to an industrial silicon production process; if at least one process data of industrial silicon production is detected to meet a trigger condition of production process optimization, generating an optimization strategy of an optimization target according to the knowledge graph and data related to the optimization target, the optimization target being used to indicate a target reached after optimizing the industrial silicon production process; and adjusting the process parameters of the industrial silicon production according to the optimization strategy.

[0007] According to a second aspect of the embodiments of the present application, an optimization device for an industrial silicon production process is provided, the device comprising: a graph construction module configured to construct a knowledge graph corresponding to an industrial silicon production process based on data sources related to the industrial silicon production; a strategy generation module configured to, if at least one process data of the industrial silicon production is detected to meet a trigger condition for production process optimization, generate an optimization strategy for an optimization target according to the knowledge graph and data related to the optimization target, the optimization target being used to indicate a target achieved after optimizing the industrial silicon production process; and a process optimization module configured to adjust process parameters of the industrial silicon production according to the optimization strategy.

[0008] According to a third aspect of the embodiments of the present application, an electronic device is provided, the electronic device comprising: a processor, a communication interface, a memory and a communication bus, the processor, the communication interface and the memory being capable of communicating with each other through the communication bus; the memory being configured to store at least one executable instruction, the executable instruction causing the processor to perform operations corresponding to the method of the first aspect.

[0009] According to a fourth aspect of the embodiments of the present application, a computer readable storage medium is provided, the computer readable storage medium storing a computer program, the computer program being executed by a processor to implement the method of the first aspect.

[0010] As can be seen from the above optimization scheme for the industrial silicon production process, based on data related to the industrial silicon production, a knowledge graph of the industrial silicon production process is constructed, and if at least one process data of the industrial silicon production triggers production process optimization during the subsequent industrial silicon production, an optimization strategy for an optimization target is generated according to the knowledge graph and data related to the optimization target, and process parameters of the industrial silicon production are adjusted according to the optimization strategy, through the support of the knowledge graph for complex reasoning and decision-making of the production process, the generation efficiency of the optimization strategy is improved, the real-time requirement of the production process optimization is met, to support real-time control and optimization of the industrial silicon production, and to achieve the optimization target. BRIEF DESCRIPTION OF DRAWINGS

[0011] The preferred embodiments of the present application will be described in detail below with reference to the accompanying drawings, so that the above and other features and advantages of the present application can be more clearly understood by those skilled in the art.

[0012] Figure 1 A flowchart of an optimization method for an industrial silicon production process is provided for an exemplary embodiment of the present application.

[0013] Figure 2 A flowchart of a cause positioning method for an industrial silicon production anomaly is provided for an exemplary embodiment of the present application.

[0014] Figure 3A process diagram of knowledge graph construction provided for an exemplary embodiment of the present application.

[0015] Figure 4 A schematic diagram of an optimization device of an industrial silicon production process provided for an exemplary embodiment of the present application.

[0016] Figure 5 A structural schematic diagram of an electronic device provided for an exemplary embodiment of the present application.

[0017] List of reference signs:

[0018] 100: optimization method of industrial silicon production process; 101-103: steps of optimization method of industrial silicon production process; 200: cause positioning method of industrial silicon production abnormality; 201-202: steps of cause positioning method of industrial silicon production abnormality; 301: top-down construction process; 302: bottom-up construction process; 10: first data; 11-19: steps of knowledge graph construction; 20: mode layer; 21: first concept; 22: second concept; 23: third concept; 24: fourth concept; 25: first relationship; 25: second relationship; 27: third relationship; 31: first entity; 32: second entity; 33: third entity; 34: first attribute; 35: fourth entity; 36: second attribute; 40: second data; 41: industrial history database; 42: structured data; 43: semi-structured data; 44: unstructured data; 50: knowledge graph; 400: optimization device of industrial silicon production process; 401: graph construction module; 402: strategy generation module; 403: process optimization module; 502: processor; 504: communication interface; 506: memory; 508: communication bus; 510: program. DETAILED DESCRIPTION

[0019] In order to make the objectives, technical solutions and advantages of the present application clearer, the following embodiments are further described in detail.

[0020] Reference should be made to Figure 1 which shows an optimization method 100 of an industrial silicon production process provided by an embodiment of the present application, the method is applied to a master control device of an industrial silicon production system, the method comprises:

[0021] Step 101, based on data sources related to industrial silicon production, a knowledge graph corresponding to the industrial silicon production process is constructed.

[0022] The master control device acquires data sources related to industrial silicon production, and constructs a knowledge graph corresponding to the industrial silicon production process based on these data sources. The knowledge graph is used to represent the knowledge structure and interconnection model of the industrial silicon production process, which contains entities (such as equipment, raw materials, process steps) and their relationships in the industrial silicon production process.

[0023] If it is detected that at least one process data of the industrial silicon production meets the trigger condition of the production process optimization, an optimization strategy of the optimization target is generated according to the knowledge graph and data related to the optimization target.

[0024] The optimization target is used to indicate a target achieved after the optimization of the industrial silicon production process. For example, the optimization target includes at least one of the following: reducing equipment wear, reducing energy consumption, improving production efficiency, reducing production cost, reducing impurity generation, controlling carbon oxygen ratio, reducing carbon excess, improving silicon recovery rate, improving silicon quality, reducing electric furnace power consumption, and stabilizing furnace condition reaction.

[0025] One or more fixed optimization targets are set in advance. Each time the optimization of the industrial silicon production process is triggered, an optimization strategy is generated for the fixed optimization target. For example, the fixed optimization target is set as improving production efficiency and reducing production cost. After the optimization of the industrial silicon production process is triggered, an optimization strategy of the process parameters for improving the production efficiency and the production cost is generated according to the knowledge graph and data related to the production efficiency and the production cost.

[0026] One or more optimization targets corresponding to the trigger condition can also be flexibly set according to the trigger condition. Each time the optimization of the industrial silicon production process is triggered, one or more optimization targets corresponding to the trigger condition of the current optimization are obtained, and the optimization target of the current optimization is obtained. Then, an optimization strategy is generated according to the knowledge graph and data related to the optimization target of the current optimization. For example, the trigger condition is raw material composition change, and the optimization target corresponding to the trigger condition is set as improving silicon recovery rate, reducing impurity generation, reducing power consumption, and reducing carbon excess. After the optimization of the industrial silicon production process is triggered, an optimization strategy of the process parameters for improving the silicon recovery rate, the impurity generation, the power consumption, and the carbon excess is generated according to the knowledge graph and data related to the silicon recovery rate, the impurity generation, the power consumption, and the carbon excess.

[0027] The two methods of setting fixed optimization targets and flexibly setting optimization targets corresponding to trigger conditions can also be combined. Each time the optimization of the industrial silicon production process is performed, one or more fixed optimization targets and one or more optimization targets corresponding to the trigger condition are optimized. For example, the fixed optimization target is improving production efficiency and reducing production cost. The trigger condition is raw material composition change, and the optimization target corresponding to the trigger condition is set as improving silicon recovery rate, reducing impurity generation, reducing power consumption, and reducing carbon excess. After the optimization of the industrial silicon production process is triggered, an optimization strategy of the process parameters for improving the production efficiency, the production cost, the silicon recovery rate, the impurity generation, the power consumption, and the carbon excess is generated according to the knowledge graph and data related to the production efficiency, the production cost, the silicon recovery rate, the impurity generation, the power consumption, and the carbon excess.

[0028] The trigger condition is set according to the process parameters that have an impact on the production of industrial silicon. For example, the trigger condition includes at least one of the following:

[0029] The raw material composition changes, for example, the raw material composition of the current batch of raw materials is different from that of the previous batch of raw materials, and / or the raw material composition of the current batch of raw materials is the same as that of the previous batch of raw materials, but the proportion of each component is different;

[0030] The furnace condition in the electric furnace changes, for example, the temperature gradient in the electric furnace changes from a first gradient to a second gradient, and the second gradient meets the temperature requirement of the production of industrial silicon, and / or the electrode position in the electric furnace changes from a first position to a second position, and the second position meets the position requirement of the production of industrial silicon, and / or the humidity gradient in the electric furnace changes from a third gradient to a fourth gradient, and the fourth gradient meets the humidity requirement of the production of industrial silicon;

[0031] The carbon balance in the electric furnace changes, for example, the carbon in the electric furnace is excessive, such as the carbon content per unit in the furnace is higher than the carbon content threshold; and / or the carbon-oxygen ratio in the electric furnace is too high or too low, such as the carbon-oxygen ratio is higher than a first ratio, then the carbon-oxygen ratio is too high, and the carbon-oxygen ratio is lower than a second ratio, then the carbon-oxygen ratio is too low.

[0032] Optionally, if it is checked that the raw material composition of the current batch of raw materials for the production of industrial silicon is different from that of the previous batch of raw materials, at least one of the following is performed:

[0033] 1) Obtain the first raw material composition and the first raw material ratio of a plurality of raw materials in the best yield batch of the production of industrial silicon, analyze the first raw material composition and the first raw material ratio according to the knowledge graph, obtain the corresponding relationship between the raw material composition and the raw material configuration, and generate the raw material ratio optimization strategy corresponding to the raw material composition of the current batch of raw materials according to the corresponding relationship and the raw material composition of the current batch of raw materials.

[0034] The optimization target includes improving the silicon recovery rate and reducing the generation of impurities; the data related to the optimization target includes the data of the best yield batch of the production of industrial silicon. The data of the best yield batch includes the first raw material composition and the first raw material ratio of a plurality of raw materials in the batch; the knowledge graph covers the raw material ratio rules, such as the correlation between the silica / reducing agent ratio and the carbon content, etc. The first raw material composition and the first raw material ratio can be analyzed according to the raw material ratio rules covered by the knowledge graph to obtain the corresponding relationship between the raw material composition and the raw material ratio. Then, the best raw material ratio corresponding to the raw material composition of the current batch of raw materials is determined according to the corresponding relationship, and the raw material ratio optimization strategy is generated according to the best raw material ratio corresponding to the current batch of raw materials.

[0035] 2) Obtain the real-time furnace condition of the electric furnace for producing industrial silicon; determine the optimal power curve corresponding to the real-time furnace condition according to the knowledge graph, and generate the power distribution optimization strategy of the electric furnace according to the optimal power curve.

[0036] The optimization target includes reducing the power consumption of the electric furnace and stabilizing the furnace reaction; the data related to the optimization target includes the real-time furnace conditions of the electric furnace, such as the furnace temperature gradient, electrode position, and furnace humidity, etc. The real-time furnace conditions of the electric furnace are obtained; the knowledge graph covers the correlation between the furnace condition data, and an energy consumption model of the real-time furnace conditions is constructed according to the knowledge graph, and then the optimal power curve corresponding to the real-time furnace conditions can be determined according to the energy consumption model, and the optimal power curve is the relationship curve between voltage and power; the knowledge graph also covers the transformer gear rules of the electric furnace, and the power distribution optimization strategy that meets the transformer gear rules is generated according to the optimal power curve.

[0037] 3) Obtain real-time carbon emission data in the electric furnace; perform carbon balance calculation on the real-time carbon emission data according to the knowledge graph to determine the target input amount of the reducing agent required to achieve carbon balance in the electric furnace, and generate an optimization strategy for the input amount of the reducing agent according to the target input amount.

[0038] The optimization target includes controlling the carbon oxygen ratio and reducing carbon excess; the data related to the optimization target includes real-time carbon emission data, which is real-time data in the electric furnace collected by sensors, such as carbon monoxide (CO) concentration, flue gas composition, etc. The real-time carbon emission data such as CO concentration and flue gas composition in the electric furnace are obtained; the knowledge graph covers the correlation between various carbon emission data in carbon balance, and a carbon balance model corresponding to the real-time carbon emission data is constructed according to the knowledge graph, and the target input amount of the reducing agent required to achieve carbon balance in the electric furnace is calculated according to the carbon balance model, and then an optimization strategy for the input amount of the reducing agent is generated according to the target input amount, and after the optimization strategy is executed, the dynamic correction of the input amount of the reducing agent in the electric furnace can be realized, so as to accurately control the carbon oxygen ratio in the electric furnace and reduce carbon excess.

[0039] The above optimization strategy includes a strategy for optimizing the process parameters corresponding to the optimization target, such as an optimization strategy for optimizing the raw material ratio, adjusting the transformer gear of the electric furnace, and correcting the input amount of the reducing agent in the electric furnace, etc.

[0040] The generation of the optimization strategy can use a query language such as SPARQL Protocol and RDF Query Language (SPARQL) for knowledge retrieval and reasoning based on the knowledge graph. For example, using the SPARQL query language, retrieve the identifiers, yield, and recovery rate of all electric furnaces with a yield greater than 1000, and the program is as follows:

[0041] PREFIX ex:<http: / / XXXXXX#> / / Defines a namespace prefix "ex:", which is used to simplify the uniform resource identifiers used in subsequent queries. The uniform resource identifier "http: / / XXXXXX#" of the namespace points to a specific Resource Description Framework (RDF) dataset.

[0042] SELECT?furnace?yield?recoveryRate / / Specifies the variables to be returned in the query results. "?furnace" represents the identifier of the circuit, "?yield" represents the yield of silicon, and "?recoveryRate" represents the recovery rate of silicon.

[0043] WHERE{ / / Defines the pattern of the query, i.e., the conditions that need to be matched in the data.

[0044] ?furnace ex:hasYield?yield. / / Specifies the first matching condition, i.e., find all furnaces "?furnace" with yield "ex:hasYield" and assign the value of these yields to the variable "?yield".

[0045] ?furnace ex:hasRecoveryRate?recoveryRate. / / Specifies the second matching condition, i.e., find all furnaces "?furnace" with recovery rate "ex:hasRecoveryRate" and assign the value of these yields to the variable "?recoveryRate".

[0046] FILTER(?yield>1000) / / Uses the FILTER statement to filter the results, retaining records with yield "?yield" greater than 1000.

[0047] }

[0048] Step 103: Adjust the process parameters of industrial silicon production according to the optimization strategy.

[0049] Execute the optimization strategy to adjust the process parameters of industrial silicon production, such as adjusting the raw material ratio corresponding to the batch of raw materials according to the raw material ratio optimization strategy; for example, adjusting the power distribution of the electric furnace according to the power distribution optimization strategy, further adjusting the transformer gear of the electric furnace; for example, adjusting the amount of reducing agent input into the electric furnace according to the optimization strategy of the amount of reducing agent input.

[0050] In summary, the method for optimizing the industrial silicon production process provided in the embodiment is based on data related to the production of industrial silicon, constructs a knowledge graph of the industrial silicon production process, and if at least one process data of the industrial silicon production triggers the optimization of the production process, generates an optimization strategy for the optimization target according to the knowledge graph and data related to the optimization target, adjusts the process parameters of the industrial silicon production according to the optimization strategy, improves the generation efficiency of the optimization strategy through the support of the knowledge graph for the complex reasoning and decision-making of the production process, meets the real-time requirements of the optimization of the production process, supports the real-time control and optimization of the production of industrial silicon, and achieves the optimization target.

[0051] In some embodiments, production abnormality diagnosis can also be performed according to the knowledge graph, for example, Figure 2 which shows a flowchart of a method 200 for locating the cause of the abnormality in the production of industrial silicon provided in an example embodiment of the present application, the method comprising:

[0052] Step 201: Based on the data sources related to the production of industrial silicon, a knowledge graph corresponding to the production process of industrial silicon is constructed.

[0053] The host device acquires data sources related to the production of industrial silicon, and constructs a knowledge graph corresponding to the production process of industrial silicon based on these data sources. The knowledge graph is used to represent the knowledge structure and interconnection model of the production process of industrial silicon, which contains entities (such as equipment, raw materials, process steps) and their relationships in the production process of industrial silicon.

[0054] Step 202: If at least one process data is detected to have production abnormality data, the historical cases associated with the production abnormality data are queried according to the knowledge graph, and the cause of the production abnormality is inferred based on the historical cases associated with the production abnormality data.

[0055] Production abnormality includes the situation of the decline of the quality and / or production efficiency of industrial silicon on the production line caused by factors such as production process, operation method, machine equipment, research and design, and materials, and the delay of progress or production shutdown phenomenon in the production process. Production abnormality data refers to data reflecting production abnormality. For example, production abnormality data includes abnormal data of temperature in the furnace, abnormal data of humidity in the furnace, abnormal data of equipment (such as abnormal working voltage of the electric furnace, jump of the transformer gear of the electric furnace), abnormal data of carbon emission in the furnace, etc.

[0056] The knowledge graph is associated with the historical case library. After detecting the production abnormality data, the historical cases associated with the production abnormality data are queried from the historical case library according to the knowledge graph, and the cause of the production abnormality is inferred based on the historical cases associated with the production abnormality data.

[0057] Optionally, if it is detected that there is abnormal data of the furnace temperature in the at least one process data, a historical case related to the temperature drop is queried according to the knowledge graph, and it is inferred that the reason for the temperature drop is the electrode fracture based on the historical case related to the temperature drop. After the reason for the temperature drop is inferred, it is determined that the emergency plan corresponding to the electrode fracture is the load reduction maintenance. Subsequently, the emergency plan of the load reduction maintenance is also pushed.

[0058] If it is detected that the difference between two adjacent furnace temperature values in the continuous multiple furnace temperature values is greater than the temperature difference threshold value, there is abnormal data of the furnace temperature in the at least one process data.

[0059] Optionally, if it is detected that there is abnormal data of the working voltage of the electric furnace in the at least one process data, a historical case related to the abnormal working voltage is queried according to the knowledge graph, and it is inferred that the reason for the abnormal working voltage is the power supply problem, or the electrode and the conducting ring problem, or the coil fault, or the capacitor problem based on the historical case related to the abnormal working voltage. After the reason for the temperature drop is inferred, it is determined that the emergency plan corresponding to the power supply problem is to check whether the power supply is normal; it is determined that the emergency plan corresponding to the electrode and the conducting ring problem is to check and clean the electrode and the conducting ring; it is determined that the emergency plan corresponding to the coil fault is to check whether the coil is intact, and to repair or replace the damaged coil if damaged; and it is determined that the emergency plan corresponding to the capacitor problem is to check the state of the capacitor, and to maintain or replace the capacitor if damaged. Subsequently, the emergency plans corresponding to the abnormal reasons are also pushed.

[0060] Optionally, if it is detected that there is abnormal jump data of the transformer gear of the electric furnace in the at least one process data, a historical case related to the abnormal jump of the transformer gear is queried according to the knowledge graph, and it is inferred that the reason for the abnormal jump of the transformer gear is the transformer damage based on the historical case related to the abnormal jump of the transformer gear. After the transformer damage is inferred, it is determined that the corresponding emergency plan is to check the transformer, and to repair or replace the transformer if damaged. Subsequently, the emergency plan corresponding to the abnormal jump of the transformer gear is also pushed.

[0061] In summary, the method for positioning the reason of the abnormality in the production of industrial silicon provided in the embodiment is based on the data related to the production of industrial silicon, constructs the knowledge graph of the production process of industrial silicon, and subsequently, if it is detected that there is abnormal production data in the at least one process data in the production process of industrial silicon, a historical case associated with the abnormal production data is queried according to the knowledge graph, and the reason for the abnormal production can be accurately positioned based on the historical case associated with the abnormal production data. In addition, a solution to the abnormal production can be provided, and problems occurring in the production process can be found and solved in a timely manner.

[0062] The method can update the proportion of raw materials, adjust the proportion of various raw materials, such as adjusting the proportion of reducing agents to balance impurities, and accurately adjust the proportion of raw materials, so as to improve the reduction rate of silicon, increase the yield of a single furnace, and optimize the material balance rate. In the method, the power distribution adjustment can maintain a more ideal reaction temperature, reduce silicon crystallization abnormalities, and improve the recovery rate; the power distribution strategy matches the furnace condition, and can also reduce invalid heat loss. In the method, the carbon balance is controlled in real time, the carbon balance in the furnace is better maintained, excessive reducing agents are avoided, and invalid energy consumption is reduced.

[0063] In the embodiments of the present application, a knowledge graph is constructed by combining a top-down and bottom-up construction method, as shown in FIG. 1. Figure 3 The construction process of the knowledge graph includes two parts.

[0064] · Top-down construction process 301.

[0065] From the data source related to the production of industrial silicon, the ontology is extracted, and the mode layer of the knowledge graph is constructed based on the ontology. For example, the implementation steps of this process are as follows:

[0066] Step 11: Constructing an expert ontology.

[0067] The data source related to the production of industrial silicon includes first data 10, which includes expert experience, industry standards, and software models; and the expert ontology is constructed based on the first data 10.

[0068] The expert ontology defines the entity concept and entity type in the mode layer 20, the entity and relationship modeling, and the constraints and rules, for example, the mode layer 20 defines a first concept 21, a second concept 22, a third concept 23, and a fourth concept 24, and also defines a first relationship 25 between the first concept 21 and the second concept 22, a second relationship 26 between the first concept 21 and the third concept 23, and a third relationship 27 between the second concept 22 and the fourth concept 24.

[0069] Concepts are the most basic elements in the mode layer, representing different entities or objects in the production process of industrial silicon. For example, the first concept can be a "reactor", the second concept can be a "temperature sensor", and the third concept can be a "pressure vessel". Each concept has its specific attributes and relationships with other concepts.

[0070] Attributes are descriptive information associated with concepts. For example, if "reactor" is a concept, its attributes may include "capacity", "material", "operating temperature", etc. Attributes provide specific details about concepts, helping to more accurately describe and distinguish different instances.

[0071] Relationships define the connection between different concepts, for example, the first relationship between "reactor" and "temperature sensor" can be a "monitoring" relationship, indicating that the temperature sensor is used to monitor the temperature inside the reactor, and relationships help to express the logical and physical connections in the production process.

[0072] Step 18: Quality Assessment.

[0073] Quality assessment is performed on the expert ontology defined in the schema layer 20.

[0074] Step 19: Data Storage.

[0075] The quality-assessed expert ontology is stored in the knowledge graph 50.

[0076] • Bottom-up construction process 302.

[0077] From the data sources related to industrial silicon production, extract entities related to raw material ratio, carbon balance in the furnace, power curve of the electric furnace, and output indicators of the electric furnace (such as silicon production, energy consumption, and silicon recovery rate), extract the relationships between entities in the industrial silicon production process, and extract the attributes of entities in the industrial silicon production process; according to the relationship and attribute, the entity is associated, the data layer of the knowledge graph is constructed, and the entity, relationship and attribute are matched with the ontology of the schema layer, so that the entity, relationship and attribute conform to the entity concept and entity type defined by the ontology, entity and relationship modeling, and constraints and rules; if there are entities, relationships and attributes that cannot be matched with the ontology, update the schema layer according to the entities, relationships and attributes that cannot be matched with the ontology.

[0078] Entities are specific manifestations of concepts, representing actual objects in the production process. For example, if "reactor" is a concept, then specific reactors A and B are entities. Entities usually contain specific attribute values, such as the capacity of reactor A is 1000 liters, the material is stainless steel, etc.

[0079] Exemplarily, the commonly used visual knowledge map construction tools include: Protégé, OILed, WebODE and OntoEdit. Among them, Protégé is an open source software for ontology editing and knowledge acquisition, which is a Java-based tool mainly used for ontology construction in the semantic web, and is the core development tool for ontology construction in the semantic web. OILed is an ontology editing tool based on OIL, which is the abbreviation of Ontology Inference Layer, which is a language used to build ontology. WebODE is an ontology engineering workbench, which provides multiple functions including editors, documentation services, hybrid services and evaluation tools for ontology creation, management and maintenance, aiming to simplify the construction and maintenance of ontology. OntoEdit is a knowledge ontology engineering environment that supports graphical construction and maintenance of knowledge ontology.

[0080] Exemplarily, the implementation steps of the construction process include:

[0081] Step 12: Data integration.

[0082] The data source related to the production of industrial silicon also includes second data 40, which includes real-time operation data, historical operation data, process mechanism documents, equipment device lists, operation instructions, industrial silicon product brochures, use cases, financial reports and other data related to the production of industrial silicon.

[0083] The second data 40 includes data obtained from the industrial history database 41 and data obtained in other ways. The structured data stored in the industrial history database 41 includes the above-mentioned real-time operation data, historical operation data, etc.; the data obtained in other ways includes structured data 42, semi-structured data 43 and unstructured data 44.

[0084] The structured data stored in the industrial history database 41 includes at least one of the following:

[0085] Static parameters such as electrode diameter, carbon bin capacity, silicon bin capacity;

[0086] Environmental variables such as flue gas temperature, molten pool voltage;

[0087] Control variables such as current setting, transformer gear, coal addition amount;

[0088] Intermediate calculation values such as carbon content, electrode consumption;

[0089] Quality variables such as yield per furnace, expected yield, recovery rate per furnace, material balance rate.

[0090] The unstructured data 44 can be data publicly disclosed on the network, data scraped, process mechanism documents, device lists, operation instructions, industrial silicon product brochures, use cases, financial reports, and other data.

[0091] Exemplary data sources related to industrial silicon production include at least one of the following: furnace temperature, current, and voltage of the electric furnace (belonging to operation data); purity of silica and carbon content of the reducing agent (belonging to operation data); expert experience related to industrial silicon production, and raw material proportioning formula; structural diagram of the electric furnace for producing industrial silicon (belonging to device list); historical optimization strategy of industrial silicon production (belonging to operation data).

[0092] The structured data obtained from the industrial history database is directly integrated; the structured data obtained from other data sources is also directly integrated. Data integration is a format conversion of structured data, so that the data after format conversion meets the format requirements of the construction of the knowledge graph. Common data integration methods include Direct Mapping (DM) and RDB to RDF Mapping Language (R2RML), and RDB is Redis Data Base. Common tools for data integration include D2RQ, MASTRO, Ultrawrap, and Morph-RDB.

[0093] Step 13: entity extraction.

[0094] Regarding entity extraction, if it is structured data, such as sensor variables in the database, entity variables are constructed based on rules; if it is semi-structured data 43 or unstructured data 44, such as industrial silicon product brochures and use cases, a deep learning method is used for named entity recognition. Exemplarily, a deep learning model constructed by using a Long Short-Term Memory (LSTM) and a Conditional Random Field (CRF), or a deep learning model constructed by using a Bidirectional Encoder Representations from Transformer (BERT) and a CRF.

[0095] Step 14: relationship and attribute extraction.

[0096] As to the extraction of relations and attributes, it can be based on rules or machine learning. For example, in relation extraction, if it is based on rule extraction, templates are obtained through manual editing or learning, and then the relations between entities are extracted and determined based on template matching. If it is based on machine learning extraction, a model for relation extraction is first trained through supervised learning or weakly supervised learning, and then the relations in the data are extracted through the model. For example, supervised learning includes kernel-based learning, and weakly supervised learning includes distance supervision-based learning.

[0097] In the field of industrial silicon, the relations and attributes of entities are extracted from the actual production operation data based on a supervised prediction model. The use conditions, algorithm flow, basic parameters and selected parameters of the control algorithm are extracted from the promotional brochure, use cases and financial reports.

[0098] Step 15: Manually construct entities, relations and attributes.

[0099] For semi-structured data 43 and unstructured data 44, entities, relations and attributes can also be extracted manually.

[0100] Step 16: Entity disambiguation.

[0101] According to the extracted entities, relations and attributes, entity disambiguation is performed.

[0102] Step 17: Expert verification.

[0103] The entities, relations and attributes after entity disambiguation are verified by experts.

[0104] According to the entities, relations and attributes after expert verification, a data layer is constructed. Subsequently, the entities, relations and attributes of the data layer are matched with the ontology of the schema layer, so that the entities, relations and attributes conform to the entity concepts and entity types, entity and relation modeling, and constraints and rules defined by the expert ontology, such as the association between the first entity 31, the second entity 32 and the third entity 33, the second entity 32 having the first attribute 34, the third entity 33 having the second attribute 36, and the association between the third entity 33 and the fourth entity 35. The entities, relations and attributes of the data layer are matched with the expert ontology of the schema layer.

[0105] Due to the complexity of the industrial silicon production process, the content of the schema layer may need to be verified by a domain expert to ensure that all concepts, attributes and relations conform to industry standards and actual operations.

[0106] Step 18: Quality evaluation.

[0107] The matched entities, relationships, and attributes are quality evaluated to ensure the accuracy and reliability of the data. When building the schema layer, the concepts, attributes, and relationships need to be quality evaluated to ensure their accuracy and to reflect the true situation of the production process. This may involve data cleaning, verification, and standardization.

[0108] Step 19: Data storage.

[0109] The quality evaluated entities, relationships, and attributes are stored in the knowledge graph 50 for subsequent querying, analysis, and application.

[0110] If there are entities, relationships, and attributes that cannot be matched with the expert ontology, the schema layer 20 is updated according to the entities, relationships, and attributes that cannot be matched with the expert ontology, and the updated data (such as the updated ontology, entities, relationships, and attributes) is quality evaluated and stored in the knowledge graph 50 for subsequent querying, analysis, and application.

[0111] Common knowledge graph storage methods include RDF table databases represented by Jena and 3store, and graph databases represented by Neo4j.

[0112] In some embodiments, the data related to the production of industrial silicon in the latest period is also obtained according to the update period of the knowledge graph; the entities, relationships, and attributes are re-extracted from the data related to the production of industrial silicon in the latest period; and the knowledge graph is updated according to the re-extracted entities, relationships, and attributes. Updating the knowledge graph includes updating the entities, relationships, attributes, and ontology (such as the entity concepts and entity types defined by the ontology, entity and relationship modeling, and constraints and rules) in the graph.

[0113] The knowledge graph can continuously learn new data through a time series model and update the weights of each node in the production process in the knowledge graph in real time. For example, the power distribution strategy can be automatically iterated as the transformer ages.

[0114] In some embodiments, the knowledge graph can also be displayed on the device in response to a triggered viewing operation of the knowledge graph on the device, that is, the visual display of the knowledge graph is realized. Gephi, D3.js, and other tools can be used for visual display of the knowledge graph.

[0115] In summary, the construction method of the knowledge graph provided in this embodiment can integrate heterogeneous data from different sources (such as sensor data, experimental data, literature data, etc.), provide a unified data view, and solve the problem of data silos. In the application process of the knowledge graph, by explicitly representing entities and their relationships, the model's explainability can be improved, helping engineers understand and trust the model's prediction and optimization results; complex reasoning and decision-making, such as rule-based reasoning and semantic queries, can be supported, helping to discover hidden relationships and patterns, and providing more intelligent optimization solutions; and the knowledge graph has the ability to dynamically update and expand, and can be continuously enriched and improved based on new data and knowledge, ensuring that the optimization model is always based on the latest information.

[0116] Regarding performance

[0117] Data integration and mining capabilities: Knowledge graph-based algorithms can more effectively integrate and mine multi-source heterogeneous data, and are more adaptable and flexible compared to traditional data processing methods.

[0118] Real-time optimization and decision support: Knowledge graphs can be quickly updated and queried, allowing process optimization to be close to real-time, thereby improving response speed and decision quality.

[0119] Knowledge graphs can uniformly represent and process data from different systems and formats. By reducing repetitive work through natural language, efficiency is improved. The natural language and graphical topology inherent to knowledge graph-based optimization systems have the potential to be combined with large models and multi-modal technologies, thereby demonstrating greater value in industrial applications.

[0120] Regarding efficiency and cost

[0121] Knowledge reuse: Knowledge graph-based algorithms enable a higher degree of automation, reducing human intervention and potential errors. Knowledge graphs allow for the reuse and sharing of knowledge, reducing redundant work and improving overall efficiency.

[0122] Real-time optimization and decision support: Effective data integration and automated processing reduce the time and labor costs associated with data cleaning and processing. Optimized processes can reduce equipment wear and energy consumption, extending equipment life and improving operational efficiency.

[0123] By effectively integrating and automating data processing, the time and labor costs associated with data cleaning and processing are reduced. Optimized processes can reduce equipment wear and energy consumption, extend equipment life, and improve operational efficiency. Knowledge graphs use a unified standard representation method, making it easy to store, manage, and reuse knowledge.

[0124] Please refer to Figure 4 , which shows a schematic diagram of an optimization device 400 for an industrial silicon production process provided by an example embodiment of the present application. The device comprises:

[0125] The atlas construction module 401 is configured to construct a knowledge graph corresponding to the industrial silicon production process based on a data source related to the industrial silicon production; the strategy generation module 402 is configured to, if it is detected that at least one process data of the industrial silicon production meets a trigger condition of production process optimization, generate an optimization strategy of an optimization target according to the knowledge graph and data related to the optimization target, the optimization target being used to indicate a target reached after optimizing the industrial silicon production process; and the process optimization module 403 is configured to adjust a process parameter of the industrial silicon production according to the optimization strategy.

[0126] In some embodiments, the strategy generation module 402 is configured to, if it is detected that the raw material of the current batch of industrial silicon is different from the raw material composition of the last batch of raw material, perform at least one of the following: obtain a first raw material composition and a first raw material ratio of a plurality of raw materials in a batch of industrial silicon production with the best yield, analyze the first raw material composition and the first raw material ratio according to the knowledge graph, obtain a corresponding relationship between the raw material composition and the raw material ratio, and generate a raw material ratio optimization strategy corresponding to the raw material of the current batch according to the corresponding relationship and the raw material composition of the current batch of raw material; obtain a real-time furnace condition of an electric furnace for producing industrial silicon; determine an optimal power curve corresponding to the real-time furnace condition according to the knowledge graph, the optimal power curve being a curve of voltage and power; generate a power distribution optimization strategy of the electric furnace according to the optimal power curve; obtain real-time carbon emission data in the electric furnace; perform carbon balance calculation on the real-time carbon emission data according to the knowledge graph, determine a target input amount of reducing agent required to achieve carbon balance in the electric furnace, and generate an optimization strategy of the input amount of the reducing agent according to the target input amount.

[0127] In some embodiments, the strategy generation module 402 is configured to, if it is detected that there is production abnormal data in at least one process data, query a historical case associated with the production abnormal data according to the knowledge graph, infer a cause of the production abnormality based on the historical case associated with the production abnormal data, and determine an emergency scheme corresponding to the cause of the production abnormality.

[0128] In some embodiments, the strategy generation module 402 is configured to, if it is detected that a difference between two adjacent in-furnace temperature values in a plurality of continuous in-furnace temperature values is greater than a temperature difference threshold value, query a historical case related to temperature drop according to the knowledge graph, infer that the cause of the temperature drop is electrode fracture based on the historical case related to the temperature drop, and determine that a corresponding emergency scheme is load reduction maintenance.

[0129] In some embodiments, the graph construction module 401 is configured to extract an ontology from a data source related to industrial silicon production, construct a schema layer of a knowledge graph based on the ontology, extract entities related to raw material proportioning, in-furnace carbon balance, power curve of an electric furnace, and output indicators of the electric furnace from the data source related to industrial silicon production, extract relationships between entities in the industrial silicon production process, and extract attributes of the entities in the industrial silicon production process; associate the entities, the relationships, and the attributes according to the relationships and the attributes, construct a data layer of the knowledge graph, and match the entities, the relationships, and the attributes with the ontology of the schema layer, so that the entities, the relationships, and the attributes conform to the entity concepts and entity types defined by the ontology, entity and relationship modeling, and constraints and rules; and if there are entities, relationships, and attributes that cannot be matched with the ontology, update the schema layer according to the entities, the relationships, and the attributes that cannot be matched with the ontology.

[0130] In some embodiments, the data source related to industrial silicon production includes at least one of the following:

[0131] furnace temperature, current, and voltage of the electric furnace, purity of silica in the raw material, and carbon content of the reducing agent, expert experience related to industrial silicon production, raw material proportioning formula, structure diagram of the electric furnace for producing industrial silicon, and historical optimization strategy of industrial silicon production.

[0132] In some embodiments, the graph construction module 401 is configured to obtain data related to industrial silicon production in a latest period according to an update period of the knowledge graph, re-extract entities, relationships, and attributes from the data related to industrial silicon production in the latest period, and update the knowledge graph according to the re-extracted entities, relationships, and attributes.

[0133] Figure 5 is a schematic block diagram of an electronic device provided by an embodiment of the present application, and embodiments of the present application do not limit the specific implementation of the electronic device. As shown in Figure 5 the electronic device can include a processor 502, a communications interface 504, a memory 506, and a communications bus 508. Among them:

[0134] The processor 502, the communications interface 504, and the memory 506 complete mutual communication through the communications bus 508.

[0135] The communications interface 504 is configured to communicate with other electronic devices or servers.

[0136] The processor 502 is configured to execute the program 510, and specifically can execute the related steps in any of the preceding embodiments.

[0137] Specifically, the program 510 can include program code including computer operation instructions.

[0138] The processor 502 can be a CPU, or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present application. The one or more processors of the smart device can be the same type of processor, such as one or more CPUs; or different types of processors, such as one or more CPUs and one or more ASICs.

[0139] RISC-V is an open-source instruction set architecture based on the principle of reduced instruction set (RISC), which can be applied to various aspects such as single-chip microcomputers and FPGA chips. Specifically, it can be applied in the fields of Internet of Things security, industrial control, mobile phones, personal computers, etc. Due to the consideration of small size, fast speed, and low power consumption in the design, it is especially suitable for modern computing devices such as warehouse-scale computers, high-end mobile phones, and small embedded systems. With the rise of artificial intelligence Internet of Things (AIoT), RISC-V instruction set architecture has received more and more attention and support, and is expected to become the next generation of widely used CPU architecture.

[0140] The computer operation instructions in the embodiments of the present application can be computer operation instructions based on the RISC-V instruction set architecture, and correspondingly, the processor 502 can be designed based on the RISC-V instruction set. Specifically, the chip of the processor in the electronic device provided by the embodiments of the present application can be a chip designed based on the RISC-V instruction set, which can execute executable code based on the configured instructions, and further implement the optimization method of the industrial silicon production process in the above embodiments.

[0141] The memory 506 is used to store the program 510. The memory 506 can include a high-speed RAM memory, and can also include a non-volatile memory such as at least one disk memory.

[0142] The program 510 can be specifically used to enable the processor 502 to execute the method in any of the preceding embodiments.

[0143] The specific implementation of each step in the program 510 can refer to the corresponding description in the corresponding steps and units of any of the preceding method embodiments, and will not be described here. Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the device and the module described above can refer to the corresponding process description in the preceding method embodiments, which will not be described here.

[0144] The present application also provides a computer-readable storage medium storing instructions for causing a machine to perform the optimization method of the industrial silicon production process as described herein. Specifically, a system or apparatus equipped with a storage medium on which a software program code for realizing the functions of any of the above-described embodiments is stored, and causing a computer (or CPU or MPU) of the system or apparatus to read out and execute the program code stored in the storage medium can be provided.

[0145] In this case, the program code read from the storage medium itself realizes the functions of any of the above-described embodiments, and thus the program code and the storage medium storing the program code constitute a part of the present application.

[0146] Embodiments of the storage medium for providing the program code include a floppy disk, a hard disk, a magneto-optical disk, an optical disk (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), a magnetic tape, a non-volatile memory card, and a ROM. Alternatively, the program code can be downloaded from a server computer via a communication network.

[0147] Embodiments of the present application also provide a computer program product including computer instructions instructing a computing device to perform any corresponding operations of the above-described method embodiments.

[0148] It should be noted that, according to the needs of implementation, each component / step described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or parts of the operation of the components / steps can be combined into a new component / step, to achieve the purpose of the embodiments of the present application.

[0149] The above-described method according to the embodiments of the present application can be implemented in hardware, firmware, or as software or computer code stored in a recording medium such as a CD ROM, a RAM, a floppy disk, a hard disk, or a magneto-optical disk, or computer code downloaded over a network and originally stored in a remote recording medium or a non-transitory machine-readable medium and to be stored in a local recording medium, so that the method described herein can be processed by such software using a general-purpose computer, a special-purpose processor, or programmable or special-purpose hardware such as an ASIC or an FPGA. It can be understood that the computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the method described herein. Furthermore, when a general-purpose computer accesses code for implementing the method shown herein, the execution of the code converts the general-purpose computer into a special-purpose computer for executing the method shown herein.

[0150] Those skilled in the art can realize the units and method steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of the present application.

[0151] The pronouns and pronouns in the patent application are not limited to a specific gender.

[0152] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method (100) for optimizing an industrial silicon production process, characterized in that, The method comprises: constructing a knowledge graph corresponding to the industrial silicon production process based on a data source related to industrial silicon production; if it is detected that at least one process data of the industrial silicon production meets a trigger condition of production process optimization, generating an optimization strategy of an optimization target according to the knowledge graph and data related to the optimization target, the optimization target being used to indicate a target reached after optimizing the industrial silicon production process; adjusting the process parameters of the industrial silicon production according to the optimization strategy.

2. The method of claim 1, wherein, The method further comprises: if it is detected that at least one process data of the industrial silicon production meets a trigger condition of production optimization, calling the knowledge graph to generate an optimization strategy of an optimization target based on data related to the optimization target, the optimization target being used to indicate a target reached after optimizing the industrial silicon production process. if it is detected that the raw material composition of the current batch of raw materials of the industrial silicon is different from that of the last batch of raw materials, at least one of the following is performed: obtaining a first raw material composition and a first raw material ratio of a plurality of raw materials in a batch of industrial silicon production with the best yield, analyzing the first raw material composition and the first raw material ratio according to the knowledge graph to obtain a corresponding relationship between the raw material composition and the raw material ratio, and generating a raw material ratio optimization strategy corresponding to the current batch of raw materials according to the corresponding relationship and the raw material composition of the current batch of raw materials; obtaining real-time furnace conditions of an electric furnace for producing industrial silicon; determining an optimal power curve corresponding to the real-time furnace conditions according to the knowledge graph, the optimal power curve being a curve of voltage versus power; and generating a power distribution optimization strategy for the electric furnace according to the optimal power curve; 3. The method of claim 1, wherein, obtaining real-time carbon emission data in the electric furnace; performing carbon balance calculation on the real-time carbon emission data according to the knowledge graph to determine a target input amount of reducing agent required to achieve carbon balance in the electric furnace, and generating an optimization strategy of the input amount of reducing agent according to the target input amount. The method further comprises:

4. The method of claim 3, wherein, if it is detected that there is production abnormal data in the at least one process data, querying a historical case associated with the production abnormal data according to the knowledge graph, reasoning the cause of the production abnormality based on the historical case associated with the production abnormal data, and determining an emergency plan corresponding to the cause of the production abnormality. The method further comprises:

5. The method of claim 1, wherein, if it is detected that the difference between two adjacent in-furnace temperature values in a plurality of consecutive in-furnace temperature values is greater than a temperature difference threshold value, querying a historical case related to temperature drop according to the knowledge graph, reasoning that the cause of the temperature drop is electrode fracture based on the historical case related to temperature drop, and determining that the corresponding emergency plan is load reduction maintenance. The method further comprises: extracting an ontology from the data source related to industrial silicon production, and constructing a schema layer of the knowledge graph based on the ontology. From the data source related to the production of industrial silicon, the entities related to the raw material ratio, the carbon balance in the furnace, the power curve of the electric furnace and the output index of the electric furnace are extracted, the relationship between the entities in the production process of industrial silicon is extracted, and the attributes of the entities in the production process of industrial silicon are extracted; the entities are associated according to the relationship and the attribute, the data layer of the knowledge graph is constructed, and the entities, the relationship and the attribute are matched with the ontology of the mode layer, so that the entities, the relationship and the attribute conform to the entity concept and entity type defined by the ontology, entity and relationship modeling, and constraints and rules; if there are entities, relationships and attributes that cannot be matched with the ontology, the mode layer is updated according to the entities, relationships and attributes that cannot be matched with the ontology.

6. The method of claim 5, wherein, The data source related to the production of industrial silicon includes at least one of the following: furnace temperature, current and voltage of the electric furnace; purity of silica in raw materials, carbon content of reducing agent; expert experience related to the production of industrial silicon, raw material ratio formula; electric furnace structure diagram for producing industrial silicon; historical optimization strategy of industrial silicon production.

7. The method of claim 1, wherein, The method further includes: According to the update cycle of the knowledge graph, the data related to the production of industrial silicon in the recent cycle is obtained; From the data related to the production of industrial silicon in the recent cycle, the entities, relationships and attributes are re-extracted; the knowledge graph is updated according to the re-extracted entities, relationships and attributes.

8. An optimization device (400) of an industrial silicon production process, characterized by, The device includes: A graph construction module (401) is configured to construct a knowledge graph corresponding to the production process of industrial silicon based on a data source related to the production of industrial silicon; A strategy generation module (402) is configured to generate an optimization strategy of an optimization target for indicating a target reached after optimizing the production process of industrial silicon according to the knowledge graph and data related to the optimization target if at least one process data of the production of industrial silicon meets a trigger condition of production process optimization. A process optimization module (403) is configured to adjust the process parameters of the production of industrial silicon according to the optimization strategy.

9. An electronic device (500), characterized by The electronic device (500) includes a processor (502), a communication interface (504), a memory (506) and a communication bus (508), the processor (502), the communication interface (504) and the memory (506) complete communication with each other through the communication bus (508); the memory (506) is used to store at least one executable instruction, and the executable instruction makes the processor (502) execute the operation corresponding to the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and the computer program is executed by a processor to implement the method of any one of claims 1-7. A computer program is stored thereon, and the computer program is executed by a processor to implement the method of any one of claims 1-7.

Images