Electrode Control Method and System for DC Submerged Arc Furnaces Based on Industrial Big Data

Abstract

This invention discloses an electrode control method and system for a DC submerged arc furnace based on industrial big data. The invention relates to the technical field of industrial big data. It determines the working content of the DC submerged arc furnace by identifying multiple data combinations, determines the spatial topological characteristics of the electrodes based on the detection of the DC submerged arc furnace, and determines a big data model of the electrodes based on the spatial topological characteristics, the working data of the electrodes, and the temperature data of the DC submerged arc furnace, thereby accurately controlling the data change curves of the electrodes. The electrode balance control content is determined based on the balance influencing factors, the current process of the DC submerged arc furnace, and the current state of the electrodes. Multiple balance control items are determined based on the identification of the electrode balance control content, and the corresponding balance control nodes are marked. The corresponding electrode control system is determined based on the item content of each balance control item, the corresponding balance control node, and the service life of the electrodes, ensuring the accuracy of the electrode control system.

Description

Technical Field

[0001] This invention relates to the technical field of industrial big data, and in particular to an electrode control method and system for a DC submerged arc furnace based on industrial big data. Background Technology

[0002] With the development of technology, DC submerged arc furnaces have emerged as industrial furnaces. A DC submerged arc furnace is a special electric furnace mainly used for smelting ferroalloys and industrial minerals. Its biggest difference from traditional AC submerged arc furnaces is that it uses DC power. Typically, it uses only one or more graphite electrodes as cathodes, while the furnace bottom or conductive lining acts as the anode. Current generates an electric arc between the electrodes and the molten pool, and the furnace charge is melted through resistance and arc heat.

[0003] This DC submerged arc furnace has multiple built-in sensors and collects corresponding data based on the sensors' readings. In existing technologies, the DC submerged arc furnace is detected, and its data is managed. During the management of multiple data, multiple industrial data of the DC submerged arc furnace are marked, and the corresponding working status is determined based on these industrial data. However, the management of the electrodes is neglected, making it impossible to perform multi-dimensional data management. This affects the accurate control of the electrode data change curves, resulting in low accuracy of the electrode control system. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides an electrode control method and system for a DC submerged arc furnace based on industrial big data.

[0005] This invention provides an electrode control method for a DC submerged arc furnace based on industrial big data, comprising: During the operation of the DC submerged arc furnace, multiple operational data of the DC submerged arc furnace are collected, and industrial big data of the DC submerged arc furnace is constructed based on multiple industrial data of the DC submerged arc furnace and corresponding work processes. The analysis of this industrial big data identifies multiple data combinations, and the identification of these multiple data combinations determines the working content of the DC submerged arc furnace, and the temperature data of the DC submerged arc furnace is marked. The spatial topological characteristics of the electrode are determined based on the detection of the DC submerged arc furnace. A big data model of the electrode is determined based on the spatial topological characteristics of the electrode, the working data of the electrode and the temperature data of the DC submerged arc furnace. The data change curve of the electrode is determined based on the big data model of the electrode and the working content of the DC submerged arc furnace. In the data change curve of the electrode, multiple imbalance characteristics of the electrode are marked. Based on the multiple imbalance characteristics and the working history of the electrode, the corresponding balance influencing factors are determined. Based on the balance influencing factors, the current process of the DC submerged arc furnace, and the current state of the electrode, the electrode balance control content is determined. Based on the identification of the electrode balance control content, multiple balance control items are determined and the corresponding balance control nodes are marked. According to the item content of each balance control item, the corresponding balance control node and the service life of the electrode, the corresponding electrode control system is determined and the DC submerged arc furnace is triggered to issue an early warning event to the electrode.

[0006] This invention provides an electrode control system for a DC submerged arc furnace based on industrial big data, which is applied to the aforementioned electrode control method for a DC submerged arc furnace based on industrial big data.

[0007] Compared with the prior art, the beneficial effects of the present invention are: (1) During the operation of the DC submerged arc furnace, multiple working data of the DC submerged arc furnace are collected. Based on the multiple industrial data of the DC submerged arc furnace and the corresponding working procedures, the industrial big data of the DC submerged arc furnace is constructed. Multiple data combinations are determined by analyzing the industrial big data. The working content of the DC submerged arc furnace is determined based on the identification of multiple data combinations, and the temperature data of the DC submerged arc furnace is marked. The spatial topological features of the electrode are determined based on the detection of the DC submerged arc furnace. The big data model of the electrode is determined based on the spatial topological features of the electrode, the working data of the electrode and the temperature data of the DC submerged arc furnace. The data change curve of the electrode is determined based on the big data model of the electrode and the working content of the DC submerged arc furnace. The spatial topological features of the electrode are introduced and the working data of the electrode and the temperature data of the DC submerged arc furnace are considered as a whole, which improves the accuracy of the big data model of the electrode and performs data control in multiple dimensions, thereby accurately controlling the data change curve of the electrode.

[0008] (2) In the data change curve of the electrode, mark multiple imbalance characteristics of the electrode, determine the corresponding balance influencing factors according to the multiple imbalance characteristics and the working history of the electrode, and determine the electrode balance control content according to the balance influencing factors, the current process of the DC submerged arc furnace, the electrode and the current state; determine multiple balance control items based on the identification of the electrode balance control content, and mark the corresponding balance control nodes, determine the corresponding electrode control system according to the item content of each balance control item, the corresponding balance control node and the service life of the electrode, control the balance dimension of the electrode, monitor the current process of the DC submerged arc furnace, the electrode and the current state in real time, and introduce the balance influencing factors to achieve accurate control of multiple balance control items, fully consider the item content of each balance control item, the corresponding balance control node and the service life of the electrode, ensure the accuracy of the electrode control system, and trigger the DC submerged arc furnace to issue an early warning event for the electrode. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating the electrode control method for a DC submerged arc furnace based on industrial big data in an embodiment of the present invention. Figure 2 This is a flowchart illustrating step S11 in the electrode control method for a DC submerged arc furnace based on industrial big data in an embodiment of the present invention. Figure 3 This is a flowchart illustrating step S12 in the electrode control method for a DC submerged arc furnace based on industrial big data in an embodiment of the present invention. Figure 4 This is a flowchart illustrating step S13 in the electrode control method for a DC submerged arc furnace based on industrial big data in an embodiment of the present invention. Figure 5 This is a flowchart illustrating step S14 in the electrode control method for a DC submerged arc furnace based on industrial big data in an embodiment of the present invention. Figure 6 This is a flowchart illustrating step S15 of the electrode control method for a DC submerged arc furnace based on industrial big data in an embodiment of the present invention. Figure 7 This is a schematic diagram of the structure of the electrode control system of a DC submerged arc furnace based on industrial big data in an embodiment of the present invention. Detailed Implementation

[0010] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0011] Please see Figures 1 to 7 An electrode control method for a DC submerged arc furnace based on industrial big data is proposed and applied to industrial big data scenarios. The electrode control method for a DC submerged arc furnace based on industrial big data includes: Step S11: During the operation of the DC submerged arc furnace, collect multiple working data of the DC submerged arc furnace, and construct industrial big data of the DC submerged arc furnace based on multiple industrial data of the DC submerged arc furnace and corresponding working procedures. Step S12: Analyze the industrial big data to determine multiple data combinations, determine the working content of the DC submerged arc furnace based on the identification of multiple data combinations, and mark the temperature data of the DC submerged arc furnace. Step S13: Determine the spatial topological characteristics of the electrode based on the detection of the DC submerged arc furnace, determine the big data model of the electrode based on the spatial topological characteristics of the electrode, the working data of the electrode and the temperature data of the DC submerged arc furnace, and determine the data change curve of the electrode based on the big data model of the electrode and the working content of the DC submerged arc furnace. Step S14: Mark multiple imbalance characteristics of the electrode in the data change curve of the electrode, determine the corresponding balance influencing factors based on the multiple imbalance characteristics and the working history of the electrode, and determine the electrode balance control content based on the balance influencing factors, the current process of the DC submerged arc furnace, and the current state of the electrode. Step S15: Based on the identification of the electrode balance control content, determine multiple balance control items and mark the corresponding balance control nodes. Determine the corresponding electrode control system according to the item content of each balance control item, the corresponding balance control node and the service life of the electrode, and trigger the DC submerged arc furnace to issue an early warning event to the electrode.

[0012] refer to Figure 2 In step S11, the specific steps are as follows: S111: Real-time monitoring of the operation of the DC submerged arc furnace. A distributed sensor array is installed at a preset position of the DC submerged arc furnace to collect multiple operating data of the DC submerged arc furnace, including current data, thermal field distribution data, gas concentration value, and electrode pressure displacement. S112: Based on the matching of multiple working data in various processes of the DC submerged arc furnace, the corresponding multi-source heterogeneous working data is determined, and the multi-source heterogeneous working data is preprocessed to output multiple key industrial data. Based on the multiple key industrial data and the corresponding spatial and temporal dimensions, the industrial big data of the DC submerged arc furnace is constructed.

[0013] In the embodiments of this application, the working process of the DC submerged arc furnace is monitored in real time. A distributed sensor combination is installed at a preset position of the DC submerged arc furnace to collect multiple working data of the DC submerged arc furnace. The multiple working data include current data, thermal field distribution data, gas concentration value, and electrode pressure displacement.

[0014] At this point, the system pre-sets distributed sensor arrays at key physical nodes of the DC submerged arc furnace (such as short grid busbars, electrode holders, furnace cover observation holes, pressure release hydraulic cylinders, and gas pipelines), uses sensor fusion technology to build a comprehensive monitoring network, and runs continuously at a high frequency sampling rate to ensure that it can capture millisecond-level electrical fluctuations and second-level process status changes, providing original, high-fidelity full data support for the subsequent big data pool.

[0015] At the specific data acquisition and execution level, the system focuses on four core types of working data: First, using DC Hall sensors installed in the DC-side main circuit, the real-time current values ​​of the three-phase electrodes are acquired at high frequency, serving as the most direct parameters for judging electrode load, arc stability, and grid energy input; second, using a furnace-mounted high-temperature infrared thermal imager with a dedicated observation hole on the furnace top, two-dimensional / three-dimensional thermal maps covering the arc zone, molten pool zone, and furnace wall are acquired in real time, visualizing the invisible furnace conditions; third, using an integrated gas composition analyzer to extract furnace gas for spectral analysis, mainly monitoring the concentrations of characteristic gases such as carbon monoxide (CO) and carbon dioxide (CO2) to determine the intensity of chemical reactions inside the furnace and the furnace's airtightness; fourth, using a high-precision displacement sensor installed on the electrode pressing mechanism, the pressing length and current absolute position of the electrodes are recorded in real time, and combined with current data, the electrode consumption rate and real-time insertion depth are calculated, providing crucial mechanical feedback signals for controlling the electrode position.

[0016] Specifically, for a 36MVA furnace in the "mid-reduction stage" of manganese-silicon alloy smelting, the system activated the distributed edge computing gateway and began polling all sensor nodes at preset locations to establish a data flow channel. DC Hall sensors installed on the A, B, and C phase short networks transmitted data in real time at a frequency of 1kHz. The system sensitively detected rapid fluctuations in the A-phase electrode current between 44kA and 47kA, indicating instability in the arc of that phase.

[0017] Simultaneously, a high-temperature infrared thermal imager on the furnace top scanned the furnace chamber. The generated thermal imaging data showed that the root region of the A-phase electrode exhibited a "localized high-temperature accumulation" characteristic, with the high-temperature area shifting towards the furnace wall. The pixel-level temperature peak reached 1850℃, significantly higher than the concurrent temperatures of the B and C phases (approximately 1700℃). A gas composition analyzer at the flue gas diaphragm showed a real-time jump in CO concentration from 22% to 26%, while CO2 concentration decreased, indicating a sudden intensification of the reduction reaction within the furnace, possibly accompanied by localized material collapse leading to an expansion of the reaction surface. Furthermore, the displacement sensor on the A-phase electrode pressing cylinder indicated that the A-phase electrode had been pressed down 15mm in the past 30 seconds, with the hydraulic pressure remaining high, suggesting that the control system was attempting to suppress current fluctuations by lowering the electrode.

[0018] Furthermore, based on the matching of multiple working data in various processes of the DC submerged arc furnace, corresponding multi-source heterogeneous working data is determined, and the multi-source heterogeneous working data is preprocessed to output multiple key industrial data. Based on the multiple key industrial data, the corresponding spatial and temporal dimensions, industrial big data of the DC submerged arc furnace is constructed, which takes into account the overall consideration of matching multiple working data in various processes of the DC submerged arc furnace and ensures the accuracy of the corresponding multi-source heterogeneous working data.

[0019] At this point, based on the phased characteristics of the DC submerged arc furnace smelting process, the system maps the real-time data stream to the current specific process through the process time sequence database, thereby filtering out a subset of effective data highly relevant to the current operating conditions and providing context for subsequent analysis. To address the differences in multi-source heterogeneous data, the system performs rigorous data preprocessing at edge computing nodes, using filtering algorithms to remove high-frequency electromagnetic noise from electrical data and correct outliers. Simultaneously, based on a unified PTP timestamp, data from different sources are strictly aligned on the time axis, and a mapping relationship is established between the thermal imager pixel coordinates and the three-dimensional physical space coordinates of the electrodes, ensuring a strict logical correspondence.

[0020] Based on this, the system constructs a multi-dimensional tensor data structure. In the time dimension, it integrates historical trend data and real-time status data to form the input sequence required by LSTM. In the spatial dimension, it integrates the electrical coupling relationship and physical position relationship between the three-phase electrodes to form the topology graph structure required by GCN. Finally, the cleaned, aligned and labeled key industrial data are packaged into standard industrial big data records according to the spatiotemporal dimensions.

[0021] Specifically, for submerged arc furnaces in the "reduction refining period", the system retrieves the process time sequence to confirm the characteristics of the current stage, and locks out the A-phase electrode current data, furnace CO concentration data and A-phase root thermal field distribution data that are strongly related to the "reduction reaction" from multi-source data streams, while ignoring irrelevant regional data.

[0022] In the preprocessing stage, edge node detection and a moving average algorithm were used to filter out high-frequency harmonic noise caused by the rectifier cabinet switch in the A-phase current signal, obtaining a smooth and true current value of 45.5kA. The system performed a spatiotemporal alignment operation. For the case where the thermal imager frame arrives with a 30ms lag, the thermal image data at time T+30ms was pushed forward based on the timestamp to be logically synchronized with the current data at time T. The two-dimensional image coordinates were mapped to the three-dimensional spatial coordinates of the A-phase electrode to accurately extract the root temperature pixels.

[0023] The system extracts a sequence vector containing 600 time points in the 10 minutes before time T to show the current change trend, constructs a topology map containing three phases A, B, and C and furnace wall nodes, marks the high weight of phase A nodes, generates a standardized data record that integrates electrical, thermal, chemical, and mechanical states and is highly consistent in time and space, and writes it into the big data pool.

[0024] refer to Figure 3 In step S12, the specific steps are as follows: S121: Perform data analysis on the industrial big data and select multiple data combinations during the analysis process; each data combination contains data in at least two dimensions, and label the corresponding data association features for each data in each data combination; S122: In multiple data combinations, the corresponding sub-work content is determined based on the data identification of each data combination. The work content of the DC electric arc furnace is determined according to multiple sub-work contents, the work history table of the DC electric arc furnace and the previous work content, so as to clarify the work details of the DC electric arc furnace in each process. S123: Perform temperature detection on the DC submerged arc furnace, determine multiple sub-temperature data during the detection process, mark the data detection position of each sub-temperature data, and determine the temperature data of the DC submerged arc furnace based on the data content of each sub-temperature data, the corresponding data detection position, and the thermal distribution diagram of the DC submerged arc furnace.

[0025] In the embodiments of this application, the industrial big data is parsed, and multiple data combinations are selected during the parsing process; each data combination contains data in at least two dimensions, and data association features corresponding to each data tag in each data combination are introduced.

[0026] At this point, based on the concept of "fully coupled multi-source data-driven modeling", the system does not view single-dimensional data in isolation, but uses multi-source information fusion technology to search for variables with physical or logical correlation in the full data pool, and forces each data combination to contain at least two different dimensions of data.

[0027] In terms of screening logic, the system pairs "current / voltage data" with "furnace thermal imaging data" to form an electro-thermal combination to describe the bidirectional effect of heat generated by current and the influence of temperature field on conductivity; pairs "electrode pressure displacement" with "gas concentration value" to form a mechanical-chemical combination to characterize the causal effect of electrode insertion depth on reaction zone temperature and reduction reaction rate; and pairs "furnace cover pressure data" with "three-phase electrode current" to form a pressure-current combination to identify the interference of furnace airtightness changes on electrical characteristics.

[0028] For each selected data combination, the system extracts the associated features through edge computing or cloud analysis and explicitly labels them. These include spatiotemporal synchronization labels that add unified high-precision timestamps to different dimensions, spatial topological labels that map thermal image pixels to specific electrode physical coordinates, and causal coupling labels that label the strength of causal relationships between data based on feature patterns trained on historical data, thereby giving the original data a clear physical meaning.

[0029] Specifically, for a 36MVA furnace undergoing intense manganese-silicon reduction, the system scan data pool revealed abnormal fluctuations in the A-phase electrode current at time T, and thermal imaging of the A-phase region showed distortion in the local temperature field. To comprehensively analyze this phenomenon, the system selected key multi-dimensional data combinations: the first combination was [A-phase current time series data] + [A-phase electrode root thermal field distribution matrix], and the second combination, considering the influence of mechanical actions, selected [A-phase electrode pressure displacement] + [furnace inlet CO concentration data].

[0030] For the "electric-thermal combination", the system analysis found that each peak of the A-phase current caused a hysteresis response in temperature at a specific coordinate point of the thermal imaging. Therefore, the thermal imaging data point was marked as Thermal_Coupling_Node_A, and the associated feature Delay_Temp_Response: 2.5s (temperature response delay of 2.5 seconds) was marked on the current data, indicating that the current fluctuation caused heat accumulation on the electrode sidewall.

[0031] Regarding the "mechanical-chemical combination", the system analysis found that the displacement of the A phase electrode was 0 in the past minute and it was in a static state, but the CO concentration suddenly dropped from 20% to 15%. Therefore, the displacement data was marked as Static_State, and the associated feature Reaction_Suppression was marked on the CO concentration data, suggesting that although the electrode position did not change, the reaction environment in the furnace may have deteriorated due to reasons such as raw material collapse.

[0032] Furthermore, among multiple data combinations, the corresponding sub-work content is determined based on the data identification of each data combination. The work content of the DC electric arc furnace is determined according to multiple sub-work contents, the work history table of the DC electric arc furnace, and previous work content, so as to clarify the work details of the DC electric arc furnace in each process. This takes into account the overall consideration of multiple sub-work contents, the work history table of the DC electric arc furnace, and previous work content, ensuring the accuracy of the work content of the DC electric arc furnace.

[0033] At this point, the system uses a pre-trained feature pattern library to match and identify the various data combinations generated, translating physical phenomena into specific "sub-work content". For example, it identifies "current data and displacement data showing a positive correlation growth" as "electrode insertion arc ignition" and "the CO concentration in the gas composition drops sharply and the pressure fluctuates" as "the permeability inside the furnace deteriorates". At this point, the judgment is still a local preliminary conclusion based on a single data dimension.

[0034] To eliminate the ambiguity of a single feature, the system introduces a work history table that records all key operations, furnace lining status, and fault records over a period of time. The identified "sub-work content" is placed on the timeline for logical verification: if the history table shows that there was a recent feeding operation, the system will determine "abnormal current fluctuation" as "material layer fluctuation" rather than "electrical fault"; if the previous working condition was "end of melting" and "resistance rise" is currently occurring, the logical deduction is "refining period begins".

[0035] Through this cross-validation and elimination method using multi-source information and historical information, the system integrates multiple scattered "sub-work contents," eliminates contradictory false alarms, and finally determines a global and accurate work content. It further refines the work details under this process, clarifies the current allowable current adjustment range, the expected electrode consumption rate, and the target position of the center of the hot zone in the furnace. Finally, it outputs a detailed control context containing process constraint parameters, which is used as a constraint condition input into the big data model of S13.

[0036] Specifically, for a running 36MVA furnace, the system performs sub-task identification at time T. By analyzing data combination A [Phase A current] + [Phase A electrode displacement], the system found that the Phase A current had been continuously decreasing over the past 5 minutes while the electrode position remained unchanged. Based on feature mapping, the system initially identified the sub-task as "Phase A electrode arc lengthening and contact resistance increasing due to consumption." Simultaneously, analyzing data combination B [furnace gas CO concentration] + [furnace cover pressure], the system found that the CO concentration was stable but the furnace cover pressure exhibited periodic small fluctuations. Based on this, the system initially identified the sub-task as "boiling of the furnace reaction surface and active exhaust."

[0037] The system query of the work history table revealed that a "tap operation" was performed at T-30 minutes, which means that the molten pool level dropped, usually causing the electrode to be positioned deeper. The system performed a correlation analysis between "sub-work content A (arc length extension)" and "tap operation (molten pool level drop)" and determined that the actual reason for the deeper electrode position was due to the drop in molten pool level, rather than simple energy consumption. The decrease in current was due to the arc being compressed because the electrode was inserted too deeply.

[0038] Based on historical data from the "melting period" before tapping, the system ruled out the misjudgment of "excessive electrode consumption" and ultimately determined that the current task was the "preparation stage for the recovery of the molten pool after tapping." Based on this, the system clarified the specific details of the current process: the molten pool level is low, and the electrode insertion depth is too deep (over-insertion). The goal is to raise the electrode to restore a reasonable arc length and utilize the active venting characteristics to prevent material collapse. The constraint is that the three-phase power balance must be strictly controlled during the electrode raising process to avoid current surges.

[0039] Therefore, temperature detection is performed on the DC submerged arc furnace, and multiple sub-temperature data points are determined during the detection process. The data detection locations of each sub-temperature data point are marked. The temperature data of the DC submerged arc furnace is determined based on the data content of each sub-temperature data point, the corresponding data detection location, and the thermal distribution diagram of the DC submerged arc furnace. This comprehensive approach, which considers the data content of each sub-temperature data point, the corresponding data detection location, and the thermal distribution diagram of the DC submerged arc furnace, ensures the accuracy of the temperature data of the DC submerged arc furnace.

[0040] At this point, the system uses a pre-set high-temperature resistant infrared thermal imager to perform a non-contact scan of the furnace, and combines the data from the embedded thermocouples to discretize the continuous temperature field into multiple sub-temperature data through an image segmentation algorithm. These sub-data can be the temperature value of a specific point or the average temperature value of a specific area.

[0041] To give the data physical meaning, the system pre-established a mapping matrix between the camera pixel coordinate system and the three-dimensional physical coordinate system of the electric arc furnace. For each extracted sub-temperature data, the system performs data detection and location marking, including marking physical coordinates such as "A phase electrode, depth -1200mm", and topological node affiliation such as "belongs to electrode-furnace wall coupling edge", thereby "spatializing" the temperature data.

[0042] The system compares and merges the marked "sub-temperature data" with the dynamically updated thermal distribution map of the DC submerged arc furnace. For outliers with excessive deviations, it uses surrounding data for interpolation smoothing or marks them as anomalies. Finally, it generates a composite data structure that includes the current temperature value, spatial gradient information, isotherm distribution, and heat flow trend, which intuitively reflects the "thermal characteristics of electrode insertion depth" and provides high-precision input for big data models.

[0043] Specifically, for a 36MVA furnace in operation, the system performs temperature detection and sub-temperature data determination at time T. A high-temperature infrared thermal imager on the furnace top scans the furnace chamber and acquires the current frame thermal image. The system analyzes the image to extract the original temperature values ​​of key areas: the pixel value of the bright area slightly to the left of the image center is 1850℃, the pixel value of the furnace wall area at the bottom of the image is 320℃, and the pixel value of the lower part of the electrode column is 1250℃. Subsequently, data detection location marking is performed. The system calls calibration parameters to map the above sub-temperature data into the three-dimensional physical model of the A-type furnace: the 1850℃ data is marked as [Electrode A: Arc Contact Area, Depth: -1.5m], which is the core area for energy release; the 320℃ data is marked as [Furnace Wall: Southeast Side, Height: +0.5m]; and the 1250℃ data is marked as [Electrode A: Root Conductive Contact Surface, Depth: -1.0m].

[0044] The system retrieves the current thermal distribution model, which shows that under normal operating conditions, heat should be concentrated at the center of the arc and decrease uniformly outwards. A fusion analysis and comparison reveals that the temperature marked [Electrode A: Root] (1250℃) is significantly higher than the historical model value (typically it should be below 1000℃), and the spatial location of the arc zone (1850℃) has shifted downwards compared to the model. Based on these discrete points and the distribution map, the system generates the final structured temperature data: {The thermal center of phase A electrode has shifted downwards, severe heat accumulation at the root, and the furnace wall heat load is normal}. This set of temperature data, through the combination of numerical and spatial information, reveals to the control algorithm the key information that the excessive insertion of phase A electrode leads to the downward shift of the high-temperature zone, directly supporting the judgment of electrode status in subsequent steps.

[0045] refer to Figure 4 In step S13, the specific steps are as follows: S131: Mark the spatial position of the electrode in the DC submerged arc furnace. Based on the spatial position of the electrode, the corresponding overall shape and the shape of the furnace wall of the DC submerged arc furnace, determine the spatial topological characteristics of the electrode in the DC submerged arc furnace. The spatial topological characteristics characterize the spatial structure of the electromagnetic coupling relationship between the three-phase electrodes. S132: Data management and control of the electrodes, and marking of the electrode working data. The electrode spatial topology features, electrode working data and DC submerged arc furnace temperature data will be deeply integrated, and data processing will be performed at the multi-source information fusion level to construct a big data model of the electrodes. The big data model of the electrodes can dynamically map the physical state change law of the electrodes under different thermal field environments and electrical conditions. S133: Collect the working content of the DC submerged arc furnace, identify multiple working items based on the identification of the working content of the DC submerged arc furnace, trigger the working simulation of the electrode within a preset time period based on the big data model of the multiple working items and the electrode, and determine multiple data nodes of the electrode during the working simulation process, and determine the data change curve of the electrode based on the multiple data nodes and the corresponding data content.

[0046] In the embodiments of this application, the spatial position of the electrodes of the DC submerged arc furnace is marked, and the spatial topological features of the electrodes in the DC submerged arc furnace are determined based on the spatial position of the electrodes, the corresponding overall shape, and the shape of the furnace wall. The spatial topological features characterize the spatial structure of the electromagnetic coupling relationship between the three-phase electrodes, and take into account the overall consideration of the spatial position of the electrodes, the corresponding overall shape, and the shape of the furnace wall, so as to ensure the accuracy of the spatial topological features of the electrodes in the DC submerged arc furnace.

[0047] At this point, the system uses a high-precision displacement encoder, a furnace top laser rangefinder, and a furnace-in-furnace visual positioning system installed on the electrode holder to determine the absolute spatial position (x, y, z) of each electrode in the three-dimensional Cartesian coordinate system of the furnace in real time, and marks the electrode's attitude parameters, including tilt angle, irregular geometry of the end, and relative position deviation, thus transforming the continuous physical object into a set of spatial coordinate points that can be processed by a computer.

[0048] Based on this, the system constructs a topological graph describing the spatial relationships inside the furnace according to the spatial location and shape of the electrodes (such as length and taper) and the furnace wall shape of the DC submerged arc furnace (such as the inner wall outline after furnace lining erosion). The electrodes are defined as "master nodes", the key areas of the furnace wall are "constraint nodes", and the connecting lines are "edges".

[0049] The system calculates the spatial distance between nodes and identifies the intermediate medium. Then, based on the spatial distance and medium characteristics, it quantifies the mutual inductance coefficient and resistive coupling coefficient between electrodes, forming a coupling weight characterizing the strength of electromagnetic coupling. As the electrodes move and the furnace lining erodes, this spatial topology is dynamically updated. For example, when an electrode approaches the furnace wall, the topological features will mark "strong thermal / electrical coupling," providing a precise mathematical description for solving the problem of "three-phase strong coupling and nonlinearity."

[0050] Specifically, for a 36MVA furnace smelting manganese-silicon alloy with uneven erosion of the furnace lining, the system reads data in real time through a sensor array and marks the three-dimensional coordinates of the A, B, and C phase electrodes. During spatial positioning, the A phase electrode is marked at coordinates (xA, yA, zA). Its zA value indicates that the insertion depth has reached the lower limit of the reference value, and due to uneven pressure on the conductive busbar, the A phase is slightly tilted relative to the central axis. At the same time, the system, combined with historical thickness measurement data, marks that the thickness of the southeast side of the furnace wall (closer to the A phase) has eroded from 600mm to 350mm, and the furnace wall shape is concave inward.

[0051] The system constructs an asymmetric spatial topology map based on coordinates and morphology. The node relationship construction shows that the spatial distance between the A-phase electrode and the "southeast furnace wall node" is significantly shortened. In the coupling relationship characterization, because the A-phase is located deeper, its arc zone is separated from the B-phase in the vertical direction. The topological feature marks the electromagnetic coupling edge weight between the A and B phases as "medium". Due to the shortened distance, the system establishes a "strong coupling" edge between the A-phase and the furnace wall node, and marks the feature as "side wall short circuit overheating risk", which means that the A-phase current fluctuation will have a stronger impact on the furnace wall temperature in this area than usual. The B and C phases maintain normal symmetrical strong coupling characteristics.

[0052] Furthermore, data management and control of the electrodes are implemented, and the working data of the electrodes is marked. The spatial topological features of the electrodes are deeply integrated with the working data of the electrodes and the temperature data of the DC submerged arc furnace. Data processing is carried out at the multi-source information fusion level to construct a big data model of the electrodes. The big data model of the electrodes can dynamically map the physical state change law of the electrodes under different thermal field environments and electrical conditions.

[0053] At this point, the system performs strict data control on the collected raw electrode working data, uses edge computing technology to remove high-frequency noise, normalizes the data, and performs a second verification based on a unified timestamp to ensure that data from different sources are strictly aligned on the time axis. At the same time, the controlled data is semantically labeled, such as labeling current fluctuations as "electrical transient features", providing a semantic basis for multi-source fusion.

[0054] The system enters the multi-source heterogeneous data tensor quantization stage, which is crucial for breaking down data silos. This involves mapping the spatial topological features determined in S131, the electrode operating data, and the DC submerged arc furnace temperature data into the same high-dimensional feature space. At the fusion level, the system mines the implicit correlations between data. By performing feature cross-calculation between temperature data and electrical operating data, it calculates the "thermal impedance" correction value and combines the spatial topological features with the operating data to quantify the "adjacent phase interference coefficient," thereby generating a composite feature vector containing four dimensions of information: spatial, temporal, electrical, and thermal.

[0055] Based on the deeply fused feature vectors, the system constructs an electrode big data model using a GCN+LSTM architecture. GCN is used to capture the complex nonlinear spatial coupling relationship between electrodes, and LSTM is used to mine the long-term dependence of state changes. This model can dynamically map the physical state change law of electrodes under different thermal field environments and electrical conditions. By inputting the current composite state, the response trend of the electrode in the very short time in the future can be output.

[0056] Specifically, the system has determined the spatial topological characteristics of the A-phase electrode near the furnace wall through S131, and now executes S132 to construct the digital model of the A-phase. The system manages the real-time operating data of the A-phase electrode current (48kA) and voltage (320V), removes instantaneous spikes caused by power grid fluctuations, normalizes the data, and labels it with <Attribute: Electrical Quantity, Status: High Load, Quality: High Signal-to-Noise Ratio>.

[0057] The system combines three dimensions of data in the feature space: the topological feature of the short distance between phase A and the furnace wall, the operating data of the large current of phase A, and the temperature data of the uneven distribution of the temperature in the root region of phase A, which is as high as 1850℃. It calculates the "thermal-electric coupling feature", which is that the effective resistance decreases by about 5% due to the high temperature. It also calculates the "air-thermal coupling feature", which is that because phase A is close to the furnace wall, the cooling effect of the furnace wall causes the root thermal field to show a "one-sided strong gradient" distribution, which affects the stability of electric arc combustion. Finally, it generates a composite feature vector describing the "phase A electrode under high temperature, near wall, and high current conditions".

[0058] The system inputs this composite vector into a big data model of the A-phase electrode based on the GCN-LSTM architecture. This model uses historical data from thousands of hours of similar operating conditions to perform feature pattern inference, mapping the physical state changes of the A-phase electrode: under the current thermal and electrical conditions, without intervention, the arc of the A-phase electrode will exhibit an "unstable drift" trend, and due to the electromagnetic constraint of the furnace wall, the arc will tend to deflect towards the furnace wall ("arc deflection" phenomenon). The big data model successfully predicted the impending "arc deflection" risk in the A-phase, providing a quantitative basis for marking imbalance characteristics in the subsequent S14 step, demonstrating the model's powerful ability to map data perception to physical laws.

[0059] Therefore, the working content of the DC submerged arc furnace is collected, and multiple work items are identified based on the identification of the working content of the DC submerged arc furnace. Based on the big data model of the electrode and the multiple work items, the working simulation of the electrode is triggered within a preset time period. During the working simulation, multiple data nodes of the electrode are identified, and the data change curve of the electrode is determined based on the multiple data nodes and their corresponding data content. This comprehensive consideration of multiple data nodes and their corresponding data content ensures the accuracy of the electrode's data change curve. At the same time, the spatial topological features of the electrode are introduced, and the working data of the electrode and the temperature data of the DC submerged arc furnace are considered as a whole, which improves the accuracy of the big data model of the electrode and performs data management in multiple dimensions, thereby accurately controlling the data change curve of the electrode.

[0060] At this point, the system acquires the high-level working content of the current DC submerged arc furnace, such as "melting period" or "iron tapping preparation", and based on the preset process rule library, it parses these macroscopic "working contents" into specific, executable work items. For example, the "refining period" is decomposed into items such as "impedance fine-tuning", "constant power control" or "electrode depth correction". These items define the boundary conditions and objective functions in the simulation process.

[0061] The system utilizes a simulation triggering mechanism based on a big data model. It inputs the currently generated real-time state data (current, temperature, topology, etc.) as the initial state into the model and sets a preset time period for simulation prediction according to the control cycle requirements. Combined with the identified "work items," the system simulates the execution of these control actions within the model. Leveraging the time-series mining capabilities of LSTM, the model does not rely on real-time feedback but instead "fast-forwards" the changes in electrode state in digital space, corresponding to the "trend prediction" in the reference document, i.e., real-time assessment of molten pool impedance changes and arc length drift trends.

[0062] During the simulation, the system captures a series of discrete data nodes at a set sampling frequency. Each node represents a complete snapshot of the physical state at the simulation moment, including expected voltage, current, electrode positions, temperature field distribution, etc. All data nodes are then connected in time series and smoothed to generate data change curves for the electrodes. These curves, as predicted trajectories, visually demonstrate the evolution trend of key parameters over time, providing a dynamic benchmark for the imbalance feature marking in the subsequent S14 step.

[0063] Specifically, the system performs work content acquisition and work item identification. The system confirms that the current work content is "reduction refining period," and the process objective is to maintain stable high-power operation and optimize arc efficiency. Based on the low impedance of phase A (approximately 4.5mΩ, lower than the target value of 4.8mΩ), the system identifies the specific work items as [Item A: Phase A electrode lifting action] and [Item B: Power fluctuation suppression]. Then, the system triggers a work simulation, loading the phase A electrode big data model constructed by S132 with the current state: current 48kA, phase A root temperature 1800℃, spatial topology display showing proximity to the furnace wall, and setting the preset time period to the next 15 minutes.

[0064] The system triggered a model simulation to perform the operation of "lifting the A-phase electrode by 20mm". The model, based on the GCN-LSTM architecture, quickly deduced the physical processes inside the furnace during these 15 minutes, taking into account the thermal constraints of the furnace wall: 0-2 minutes for arc elongation, 2-10 minutes for thermal redistribution and temperature gradient changes, and 10-15 minutes for electrical parameters to tend towards a new steady state. Subsequently, data nodes and generated curves were determined, and the system extracted key nodes on the simulation time axis: node T+2min showed the impedance jumping to 5.0mΩ and the power dropping instantaneously; node T+8min showed the impedance returning to 4.85mΩ due to the disappearance of thermal inertia; and node T+15min showed the steady-state impedance stabilizing at 4.82mΩ and the furnace wall temperature dropping from 1850℃ to 1750℃.

[0065] Based on the above nodes, the system plots a predicted curve of the impedance change of phase A electrode. This curve exhibits a "damped oscillation" pattern, characterized by a rapid rise followed by a gradual decline and eventual stabilization. This predicted curve indicates to the control system that the system will experience a brief power fluctuation after the electrode is raised, but will reach the optimal process impedance value after 15 minutes. This allows the control system to anticipate the "arc length drift trend" in advance, thereby enabling it to formulate a confident balance control strategy in S14.

[0066] refer to Figure 5 In step S14, the specific steps are as follows: S141: Identify the data change curve of the electrode, determine multiple imbalance regions during the identification process, and mark the data set of each imbalance region. Based on the identification of the data set of the imbalance region, determine the corresponding imbalance feature to mark multiple imbalance features. S142: Retrieve the working process of the electrode and determine the working history of the electrode based on the working behavior of the motor in the working process. The working history includes past adjustment records, consumption rate and accident log; perform spatiotemporal correlation mining between the working history of the electrode and the current imbalance characteristics, and determine the corresponding balance influencing factors in the attribution analysis dimension. S143: Collect the current state of the electrode, and perform multi-objective optimization on the balance influencing factors, the current process of the DC submerged arc furnace, and the electrode and current state. In the optimization process, determine the targeted electrode balance control content, which specifies the specific adjustment content, adjustment direction and adjustment timing.

[0067] In the embodiments of this application, the data change curve of the electrode is identified, and multiple imbalance regions are determined during the identification process. The data set of each imbalance region is marked, and the corresponding imbalance feature is determined based on the identification of the data set of the imbalance region. Multiple imbalance features are marked, which takes into account the overall consideration of the identification of the data set of the imbalance region and ensures the accuracy of the corresponding imbalance feature.

[0068] At this point, the system compares the generated "electrode data change curve" with the preset ideal process reference envelope in real time, and calculates the deviation between the measured or predicted curve and the reference line. This comparison process requires extremely high accuracy; for example, the power deviation needs to be controlled within 0.5%. The identification process not only focuses on static numerical deviations but also captures potential "trend deterioration" signals by monitoring the dynamic slope changes of the curve.

[0069] When the deviation value exceeds the set threshold or the relative deviation between the three phases exceeds the safety limit, the system locks the corresponding time period on the time axis and defines these locked time periods and their corresponding data intervals as unbalanced regions, whether they are brief peak transients or continuous slow drifts.

[0070] For each identified imbalance region, the system performs slicing operations in the multidimensional data space, extracting the complete data set of related electrical quantities, thermal field data, and mechanical quantities within that time period, and assigning spatiotemporal index labels for logical isolation. The system then performs deep mining on the labeled data set, extracting imbalance features that describe the essence of the anomaly, including amplitude features, phase features, temporal features, and coupling features. The extracted features are quantified and labeled to form feature vectors, providing crucial evidence for attribution analysis in subsequent steps.

[0071] Specifically, when performing curve identification, the system loads the A-phase electrode impedance prediction curve and calls the current standard impedance reference curve for the "refining period" (target value 4.8mΩ, allowable fluctuation ±0.2mΩ). By comparing point by point, the system finds that the prediction curve shows a significant deviation trend between the 5th and 12th minutes, thus identifying the imbalance region.

[0072] Based on the comparison results, the system determined that the impedance value remained below the baseline lower limit (reaching a minimum of 4.3 mΩ) within the time window [T+5 min, T+12 min], and the impedance prediction curves of both phases B and C were within the normal range. Therefore, [region U1: Phase A impedance imbalance area] was identified, and this was determined to be a localized imbalance region. For region [U1], the system quickly sliced ​​and labeled the data set for this time period from the industrial big data pool. This set included the phase A current, voltage, corresponding phase A root thermal imaging temperature matrix, and hydraulic system pressure logs for the period from T+5 to T+12 minutes. All data were tagged with "correlated imbalance U1".

[0073] In determining the imbalance characteristics, the system performed feature extraction analysis on the [U1] dataset and found that the impedance decrease showed a linear and monotonically decreasing trend. Furthermore, the thermal field temperature data for phase A in the same dataset showed a continuous increase, while the displacement sensor indicated that the electrode position remained unchanged. Based on this, the system identified the imbalance characteristics and labeled them as: <Characteristic 1: Low impedance drift>, <Characteristic 2: Abnormal conductivity due to thermal accumulation>, and <Characteristic 3: Mechanical position lock-up>. These characteristics clearly indicate that the imbalance in phase A is not due to simple voltage fluctuations, but rather to "changes in electrical characteristics caused by thermal effects," providing crucial clues for the subsequent attribution of S142.

[0074] Furthermore, the working process of the electrode is retrieved, and the working history of the electrode is determined based on the working behavior of the motor in this process. This working history includes past adjustment records, consumption rates, and accident logs. The working history of the electrode is spatiotemporally correlated with the current imbalance characteristics, and the corresponding balance influencing factors are determined in the attribution analysis dimension. This takes into account the overall working behavior of the motor in this process, ensuring the accuracy of the working history of the electrode.

[0075] At this point, the system retrieves the complete working process record of a specific electrode within a specific time window from a long-term stored historical database. By analyzing the working behavior of the motor and hydraulic actuator, such as the pressing and releasing frequency, lifting acceleration, and the proportion of static holding time, a dynamic working history archive containing physical states is constructed. The specific content of this archive includes detailed records of past adjustments for each control command execution and response, consumption rates calculated based on historical data from displacement sensors to determine whether the device is in an "abnormal wear" state, and accident logs with strong spatiotemporal markers.

[0076] Based on this, the system performs spatiotemporal correlation mining of electrode working history and imbalance characteristics. It uses a long short-term memory network (LSTM) to analyze whether the current imbalance characteristics are the lag effect of historical operations in the time dimension, and associates the historical consumption of the electrode with the current spatial topological features in the spatial dimension. It mines how changes in electrode shape change the spatial position of the arc, and finds event sequences that are highly overlapping in time and space through multidimensional feature matching.

[0077] The system identifies the factors affecting the balance based on attribution analysis, and locks down the root cause through logical elimination and causal inference. The factors leading to the imbalance are classified into equipment factors (such as changes in contact resistance caused by excessive electrode consumption), raw material factors (such as sudden changes in electrode embedment depth caused by furnace charge collapse), or control factors (such as excessive past adjustment ranges causing system overshoot). In this way, physical phenomena are transformed into specific and operable factors, pointing the way for subsequent multi-objective optimization.

[0078] Specifically, step S141 has identified an imbalance characteristic of "continuous monotonous decrease in impedance" in phase A electrode. The system now executes step S142 to investigate the cause. The system retrieves the working process data of phase A electrode over the past 6 hours, constructs a working history file, and analyzes it. It finds that the hydraulic system of phase A electrode performed 12 compression and release operations during these 6 hours, with a total compression and release amount 150mm greater than that of phases B and C. The file shows that phase A is in a fatigue working state of "high-frequency consumption and rapid insertion".

[0079] When analyzing the specific details of the work process, past adjustment records showed that in order to maintain the current of phase A, the system frequently issued "micro-increase" commands in the past 2 hours, but the current would drop rapidly after each increase, indicating that the adjustment effect was not good; the consumption rate calculation showed that phase A was 18mm / h, which was much higher than the process standard of 12mm / h, suggesting that the physical environment at the bottom of phase A was harsh; the accident log query found that a slight "furnace pressure fluctuation" was recorded at time T-3, which may correspond to a small-scale collapse of the furnace charge.

[0080] The system correlated the "continuous impedance decrease" with the "charge collapse at T-3 hours" in terms of time, finding that the A-phase current began to increase abnormally 30 minutes after the collapse. Spatially, combined with consumption rate analysis, due to the excessively rapid consumption of phase A, the actual working end was 200mm shorter than expected. The system correlated the "excessive consumption" with the "upward shift of the arc's spatial position," finding that the upward shift of the arc led to a non-linear decrease in arc resistance. The conclusion was that the current impedance decrease was caused by a chain reaction where "charge collapse" triggered "environmental changes," which in turn accelerated "electrode consumption," ultimately leading to "changes in electrical characteristics." The system ruled out the possibility of grid fluctuations and control algorithm errors, identifying the corresponding balance influencing factor as [the physical shortening of the arc length caused by abnormal consumption of the phase A electrode]. This means that subsequent control strategies cannot simply rely on reducing the current to restore impedance; instead, compensation for electrode length or adjustment of the pressure release strategy must be considered, otherwise the imbalance will recur.

[0081] Therefore, the current state of the electrode is collected, and multi-objective optimization is performed on the influencing factors of the balance, the current process of the DC submerged arc furnace, and the current state of the electrode. In the optimization process, targeted electrode balance control content is determined. This electrode balance control content specifies the specific adjustment content, adjustment direction, and adjustment timing.

[0082] At this time, the system collects the complete physical state vector of the electrode in real time, covering electrical state, mechanical state, thermal state and equipment health state, and maps these heterogeneous data into a unified state space to form optimized initial conditions. At the same time, the current deviation is converted into per-unit value and the temperature change is converted into thermal stress index, thus forming the boundary constraints of the optimization algorithm to ensure that no control command will exceed the safe physical limits of the equipment.

[0083] The system performs multi-objective optimization based on influencing factors and processes, constructing a multi-dimensional objective function that includes balance, energy efficiency, stability, and safety indicators. Balance influencing factors from S142 (such as "abnormal consumption"), the current process (such as "refining period"), and the current state are used as constraint inputs. Utilizing a model predictive control (MPC) algorithm, the system iteratively solves the optimization problem in the control time domain, predicts the system response under different control strategies, and selects an optimal trajectory that satisfies balance requirements, minimizes interference with other phases, and meets process requirements.

[0084] The system determines the targeted electrode balance control content based on the output trajectory optimized by MPC, generates a set of instructions containing logical constraints, defines specific adjustment content (such as electrode position lifting, current setpoint correction), clear adjustment direction (such as forward lifting, reverse current reduction), and uses "matrix switch peak shifting adjustment" technology to specify the precise timing of actions, such as specifying the specific time points of mechanical prediction and electrical microsecond-level peak shifting compensation, to ensure the dynamic stability of the high-power DC circuit during drastic adjustments.

[0085] Specifically, for a 36MVA furnace in the "refining period" with phase A experiencing "physical shortening of arc length due to abnormal consumption", the phase A impedance is too low (4.3mΩ) and the power is severely insufficient. The system collected data showing that the phase A electrode is located at a depth of Z=-1520mm, with a current as high as 52kA (overload) and a root temperature of 1900℃ (critical danger). Meanwhile, the normal operating current of phases B and C is 45kA, and the hydraulic system status shows that the phase A brake pressure is normal and has the ability to operate.

[0086] The system's objectives are as follows: The primary objective (safety) is to immediately reduce the A-phase current to below 50kA to prevent tripping. The secondary objective (balance) is to restore the A-phase impedance to 4.8mΩ, ensuring the three-phase power deviation is less than 0.5%. The third objective (process) is to maintain stable temperature during the refining period to avoid significant fluctuations. Regarding input constraints, considering the influencing factor of "physical arc length shortening," adjusting the current alone cannot fundamentally change the impedance; the electrode must be raised. Given that the current phase is "refining," the adjustment must be gradual, and rapid arc-breaking methods cannot be used. The MPC algorithm simulated two schemes: "current reduction only" and "electrode raising + current reduction." It was found that if only the current is reduced, the power of phases B and C will fluctuate due to magnetic coupling. The optimal solution is to raise the A-phase electrode by 20mm to physically lengthen the arc, while simultaneously fine-tuning the current setting.

[0087] The system generates combined instructions encompassing both mechanical and electrical adjustments: In terms of adjustment content, it generates instructions for the operation of the A-phase hydraulic pressure release system and the adjustment of the A-phase rectifier cabinet thyristor firing angle; in terms of adjustment direction, it sets the A-phase electrode to lift upwards (+Z direction) by 20mm, and the current setting to be finely adjusted downwards by -2kA; in terms of adjustment timing, the mechanical timing is set to immediate (T+0s) with a medium-speed start, while the electrical timing employs a peak-shaving adjustment strategy, adjusting the firing angle within a microsecond-level window near the grid voltage zero-crossing point, with the action lagging 1.5 seconds after the mechanical lifting to ensure a smooth transition of electrical characteristics and avoid shocks. This control content is directly transmitted to the execution layer, driving the DC submerged arc furnace to transition from "imbalance" to "perfect balance."

[0088] refer to Figure 6In step S15, the specific steps are as follows: S151: Dynamically identify the electrode balance control content, mark multiple balance control markers during the identification process, determine the corresponding balance control items based on the traceability of each balance control marker, determine multiple balance control items, and determine the corresponding balance control nodes according to the item content of each balance control item, the working procedures of the DC submerged arc furnace and the corresponding working content. S152: Collect the service life of the electrodes, comprehensively consider the specific execution parameters of each balance control project, the timeliness requirements of key control nodes, and the service life of the electrodes, and construct a corresponding electrode control system based on multiple constraints; the electrode control system monitors the difference between the operating parameters and the safety threshold in real time. If the difference exceeds the preset difference, the electrode early warning event of the DC submerged arc furnace is triggered, and the corresponding protection mode is executed.

[0089] In the embodiments of this application, the electrode balance control content is dynamically identified, and multiple balance control markers are marked during the identification process. The corresponding balance control items are determined based on the tracing of each balance control marker, thereby identifying multiple balance control items. The corresponding balance control nodes are determined according to the item content of each balance control item, the working process of the DC submerged arc furnace, and the corresponding working content. This approach takes into account the overall consideration of the item content of each balance control item, the working process of the DC submerged arc furnace, and the corresponding working content, ensuring the accuracy of the corresponding balance control nodes.

[0090] At this point, the system receives the electrode balance control content, uses natural language processing or a rule engine to parse it into multiple indivisible atomic operations, and generates a balance control tag for each operation containing core metadata such as action type, target object, and adjustment attributes, thereby understanding the physical properties of the control command.

[0091] Based on these tags, the system performs traceability matching in the underlying "device capability library", mapping semantic action types to specific physical device control codes, thereby identifying executable balance control items containing specific execution parameters (such as displacement distance and current setpoint).

[0092] The system analyzes the logical relationships between various control items and calculates the optimal action sequence based on the current work procedures and specific work content of the DC submerged arc furnace. A series of balanced control nodes, including trigger times and execution content, are defined on the time axis. Based on the "matrix switch peak-shifting adjustment" technology in the reference document, the node planning deliberately arranges adjustment actions of different phases to be staggered at the microsecond level, or prioritizes mechanical actions over electrical actions, thereby forming an execution plan accurate to the millisecond level and effectively avoiding system oscillations.

[0093] Specifically, the decision-making layer in S143 issued the balance control content: "Lift the A-phase electrode by 20 mm, and at the same time lower the set value of the A-phase current by 2 kA to restore three-phase balance." The system analyzed the control content and identified two core actions: Tag 1 <Tag_1: Action type = Lift, Object = A-phase electrode, Amplitude = 20 mm>, Tag 2 <Tag_2: Action type = Current adjustment, Object = A-phase electrode, Amplitude = -2 kA>. At the same time, based on the current operating condition that the A-phase is in the high-temperature area, the system added a safety tag <safety: 防止过热, 优先级="高">.

[0094] based on<Tag_1> The system traces the equipment database to determine control item P1: [A-phase hydraulic brake release + pressure release cylinder performs 20mm upward stroke]; based on<Tag_2> The system traces the equipment library and identifies control item P2: [Modify the current setpoint I_ref of the A-phase DC controller, reducing the target value by 2kA]. At this point, the system has two specific execution items: mechanical action and electrical regulation.

[0095] Considering the current "refining period" requirement of "minimizing thermal disturbance," executing P1 and P2 simultaneously could cause the arc to break instantly, resulting in drastic power fluctuations. Therefore, the system implemented node planning: Node N1 (T+0s) triggers control item P1, instructing the hydraulic system to start lifting the A-phase electrode at a medium speed, while the electrical parameters remain unchanged; Node N2 (T+3s) triggers control item P2. The system predicts that after 3 seconds of mechanical lifting, the arc length has lengthened and the impedance has increased. At this point, the current reduction operation is executed to offset the power fluctuations caused by the increase in impedance, allowing the power to smoothly transition to the target value.

[0096] Furthermore, the lifespan of the collected electrodes is comprehensively considered, taking into account the specific execution parameters of each balance control item, the timeliness requirements of key control nodes, and the lifespan of the electrodes. Based on multiple constraints, a corresponding electrode control system is constructed. This electrode control system monitors the difference between the operating parameters and the safety threshold in real time. If the difference exceeds the preset difference, an electrode early warning event is triggered in the DC submerged arc furnace, and corresponding protection methods are implemented. Corresponding protection methods are introduced, and the balance influencing factors are also introduced, achieving precise control of multiple balance control items. The system fully considers the project content of each balance control item, the corresponding balance control nodes, and the lifespan of the electrodes, ensuring the accuracy of the electrode control system and triggering an early warning event for the electrodes in the DC submerged arc furnace.

[0097] At this point, the system first calculates the remaining service life index of the electrode in real time based on historical consumption rate, number of thermal shocks, and physical wear, and integrates multiple dimensions such as physical execution constraints (e.g., maximum thrust of hydraulic cylinders), timeliness constraints (e.g., critical node time windows), and equipment life constraints (e.g., dynamically adjusting the adjustment intensity based on the remaining length). Based on this, the system constructs an electrode control system, establishes dynamic safety boundaries, and generates control commands containing restrictive strategies according to the electrode state. For example, it automatically switches older electrodes nearing the end of their lifespan to a flexible adjustment mode of "low speed, small amplitude, high frequency," and packages the control commands, safety thresholds, and lifespan parameters into a closed-loop monitoring safety control context.

[0098] During execution, the system monitors operating parameters in real time at millisecond-level frequency, calculating the difference between real-time values ​​and safety thresholds. This difference measures not only whether the boundary has been exceeded but also the distance from the boundary. Once the difference exceeds the preset safety margin, the system determines it as an abnormal operating condition and triggers an electrode warning event. Based on the risk level, it executes multi-level protection methods, including automatically locking the current command, forcibly cutting off the power supply or raising the electrode, and triggering remote locking, to implement an "active defense" mechanism.

[0099] Specifically, for a 36MVA furnace undergoing the S151 balancing control project (raising the A-phase electrode by 20mm), step S152 is responsible for monitoring the safety of this process. System evaluation revealed that the bottom of the A-phase electrode had been reduced to 150mm, approaching the design limit of 200mm, and had experienced multiple thermal shocks in the past 24 hours, resulting in a high mechanical fatigue index. Based on this, the system limited the standard lifting speed from 10mm / s to 3mm / s and the acceleration to 30% of the standard value, thus creating multiple constraints for this adjustment.

[0100] Based on the above constraints, the system constructed a flexible lifting control system for the A-phase electrode. The adjustment was set to be completed within 10 seconds using a low-speed crawling method, with dynamic safety thresholds set: vibration acceleration < 2.0g and temperature rise rate < 5℃ / s. During real-time monitoring of the difference, when the A-phase electrode was slowly lifted to the 3rd second, the system detected increased airflow disturbance within the furnace. The lateral vibration frequency of the A-phase electrode jumped from 5Hz to 15Hz, and the real-time vibration acceleration reached 2.5g. The difference from the safety threshold of 2.0g was +0.5g, exceeding the ±0.2g safety margin. In the triggering warning and execution protection phase, the system determined that the A-phase electrode was on the verge of "resonance fracture" and immediately triggered a [Warning Event: Abnormal Electrode Mechanical Structure], classified as high-risk.

[0101] The system immediately implemented a proactive defense strategy: it forcibly cut off the hydraulic pressure relief valve signal to stop the A-phase electrode from continuing to rise, temporarily locked the A-phase rectifier cabinet current to reduce electromagnetic pull, and switched the A-phase control mode to "parameter hold" state, waiting for the vibration to decay before reassessing whether to continue. Through this mechanism, the system successfully avoided a major shutdown accident caused by the resonance fracture of an old electrode during the adjustment process.

[0102] Please see Figure 7 , Figure 7 This is a schematic diagram of the structure of the electrode control system of the DC submerged arc furnace based on industrial big data in an embodiment of the present invention; the electrode control system of the DC submerged arc furnace based on industrial big data includes: The industrial big data module 21 is used to collect multiple working data of the DC submerged arc furnace during its operation, and to construct industrial big data of the DC submerged arc furnace based on the multiple industrial data of the DC submerged arc furnace and the corresponding working procedures. The work content module 22 is used to analyze the industrial big data to determine multiple data combinations, determine the work content of the DC submerged arc furnace based on the identification of multiple data combinations, and mark the temperature data of the DC submerged arc furnace. The data change curve module 23 is used to determine the spatial topological characteristics of the electrode based on the detection of the DC submerged arc furnace. Based on the spatial topological characteristics of the electrode, the working data of the electrode and the temperature data of the DC submerged arc furnace, a big data model of the electrode is determined. Based on the big data model of the electrode and the working content of the DC submerged arc furnace, the data change curve of the electrode is determined. Electrode balance control content module 24 is used to mark multiple imbalance characteristics of the electrode in the data change curve of the electrode, determine the corresponding balance influencing factors based on the multiple imbalance characteristics and the working history of the electrode, and determine the electrode balance control content based on the balance influencing factors, the current process of the DC submerged arc furnace, and the current state of the electrode. The electrode control system module 25 is used to identify multiple balance control items based on the identification of the electrode balance control content, mark the corresponding balance control nodes, determine the corresponding electrode control system according to the item content of each balance control item, the corresponding balance control node and the service life of the electrode, and trigger the DC submerged arc furnace to issue an early warning event to the electrode.

[0103] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.< / safety:>

Claims

1. An electrode control method for a DC submerged arc furnace based on industrial big data, characterized in that, include: During the operation of the DC submerged arc furnace, multiple operational data of the DC submerged arc furnace are collected, and industrial big data of the DC submerged arc furnace is constructed based on multiple industrial data of the DC submerged arc furnace and corresponding work processes. The analysis of this industrial big data identifies multiple data combinations, and the identification of these multiple data combinations determines the working content of the DC submerged arc furnace, and the temperature data of the DC submerged arc furnace is marked. The spatial topological characteristics of the electrode are determined based on the detection of the DC submerged arc furnace. A big data model of the electrode is determined based on the spatial topological characteristics of the electrode, the working data of the electrode and the temperature data of the DC submerged arc furnace. The data change curve of the electrode is determined based on the big data model of the electrode and the working content of the DC submerged arc furnace. In the data change curve of the electrode, multiple imbalance characteristics of the electrode are marked. Based on the multiple imbalance characteristics and the working history of the electrode, the corresponding balance influencing factors are determined. Based on the balance influencing factors, the current process of the DC submerged arc furnace, and the current state of the electrode, the electrode balance control content is determined. Based on the identification of the electrode balance control content, multiple balance control items are determined and the corresponding balance control nodes are marked. According to the item content of each balance control item, the corresponding balance control node and the service life of the electrode, the corresponding electrode control system is determined and the DC submerged arc furnace is triggered to issue an early warning event to the electrode.

2. The electrode control method for a DC submerged arc furnace based on industrial big data according to claim 1, characterized in that, During the operation of the DC-DC submerged arc furnace, multiple operational data points are collected. Based on these data points and corresponding operational procedures, industrial big data for the DC-DC submerged arc furnace is constructed, including: The DC submerged arc furnace is monitored in real time. A distributed sensor array is installed at a preset position of the DC submerged arc furnace to collect multiple working data of the DC submerged arc furnace, including current data, thermal field distribution data, gas concentration value, and electrode pressure displacement. Based on the matching of multiple working data in various processes of the DC submerged arc furnace, corresponding multi-source heterogeneous working data is determined, and the multi-source heterogeneous working data is preprocessed to output multiple key industrial data. Based on the multiple key industrial data and the corresponding spatial and temporal dimensions, industrial big data of the DC submerged arc furnace is constructed.

3. The electrode control method for a DC submerged arc furnace based on industrial big data according to claim 1, characterized in that, The process involves analyzing the industrial big data to determine multiple data combinations, identifying the operating content of the DC submerged arc furnace based on the identification of these data combinations, and marking the temperature data of the DC submerged arc furnace, including: The industrial big data was analyzed, and multiple data combinations were selected during the analysis process. Each data combination contained data in at least two dimensions, and the corresponding data association features were marked for each data in each data combination. Among multiple data combinations, the corresponding sub-work content is determined based on the data identification of each data combination. The work content of the DC electric arc furnace is determined according to multiple sub-work contents, the work history table of the DC electric arc furnace and the previous work content, so as to clarify the work details of the DC electric arc furnace in each process. Temperature detection is performed on the DC submerged arc furnace, and multiple sub-temperature data points are determined during the detection process. The data detection locations of each sub-temperature data point are marked. The temperature data of the DC submerged arc furnace is determined based on the data content of each sub-temperature data point, the corresponding data detection location, and the thermal distribution diagram of the DC submerged arc furnace.

4. The electrode control method for a DC submerged arc furnace based on industrial big data according to claim 1, characterized in that, The spatial topological characteristics of the electrode are determined based on the detection of the DC submerged arc furnace. A large data model of the electrode is then determined based on these characteristics, the electrode's operating data, and the DC submerged arc furnace's temperature data. Finally, a data change curve of the electrode is determined based on this large data model and the DC submerged arc furnace's operating conditions. This includes: The spatial position of the electrode in the DC submerged arc furnace is marked. Based on the spatial position of the electrode, its corresponding overall shape, and the shape of the furnace wall, the spatial topological features of the electrode in the DC submerged arc furnace are determined. These spatial topological features characterize the spatial structure of the electromagnetic coupling relationship between the three-phase electrodes. Data management and control of the electrodes are implemented, and the working data of the electrodes are marked. The spatial topological features of the electrodes are deeply integrated with the working data of the electrodes and the temperature data of the DC submerged arc furnace. Data processing is carried out at the multi-source information fusion level to construct a big data model of the electrodes. The big data model of the electrodes can dynamically map the physical state change law of the electrodes under different thermal field environments and electrical conditions.

5. The electrode control method for a DC submerged arc furnace based on industrial big data according to claim 4, characterized in that, The method of determining the spatial topological characteristics of the electrode based on the detection of the DC submerged arc furnace, determining a large data model of the electrode based on the spatial topological characteristics of the electrode, the electrode's working data, and the temperature data of the DC submerged arc furnace, and determining the data change curve of the electrode based on the large data model of the electrode and the working content of the DC submerged arc furnace, further includes: The system collects the working content of the DC submerged arc furnace, identifies multiple work items based on the identification of the working content of the DC submerged arc furnace, triggers the working simulation of the electrode within a preset time period based on the big data model of the multiple work items and the electrode, and determines multiple data nodes of the electrode during the working simulation process. Based on the multiple data nodes and the corresponding data content, the system determines the data change curve of the electrode.

6. The electrode control method for a DC submerged arc furnace based on industrial big data according to claim 1, characterized in that, In the data change curve of the electrode, multiple imbalance characteristics of the electrode are marked. Based on these imbalance characteristics and the electrode's working history, corresponding balance influencing factors are determined. Based on these balance influencing factors, the current process of the DC submerged arc furnace, and the current state of the electrode, the electrode balance control content is determined, including: The data change curves of the electrodes are identified, and multiple imbalance regions are determined during the identification process. The data sets of each imbalance region are marked, and the corresponding imbalance features are determined based on the identification of the data sets of the imbalance regions, so as to mark multiple imbalance features. The working process of the electrode is retrieved, and the working history of the electrode is determined based on the working behavior of the motor during the working process. This working history includes past adjustment records, consumption rates, and accident logs. The working history of the electrode is spatiotemporally correlated with the current imbalance characteristics, and the corresponding balance influencing factors are determined in the attribution analysis dimension.

7. The electrode control method for a DC submerged arc furnace based on industrial big data according to claim 6, characterized in that, The process involves marking multiple imbalance characteristics of the electrode in the data change curve, determining corresponding balance influencing factors based on these characteristics and the electrode's operational history, and determining electrode balance control content based on these influencing factors, the current process of the DC submerged arc furnace, and the current state of the electrode. This also includes: The current state of the electrodes is collected, and multi-objective optimization is performed on the influencing factors of the balance, the current process of the DC submerged arc furnace, and the current state of the electrodes. During the optimization process, targeted electrode balance control content is determined, which specifies the specific adjustment content, adjustment direction, and adjustment timing.

8. The electrode control method for a DC submerged arc furnace based on industrial big data according to claim 1, characterized in that, The process involves identifying multiple balance control items based on the electrode balance control content, marking corresponding balance control nodes, determining the corresponding electrode control system based on the item content of each balance control item, the corresponding balance control node, and the electrode's lifespan, and triggering early warning events for the electrodes in the DC submerged arc furnace, including: The electrode balance control content is dynamically identified, and multiple balance control markers are marked during the identification process. The corresponding balance control items are determined based on the traceability of each balance control marker, thereby identifying multiple balance control items. The corresponding balance control nodes are determined according to the item content of each balance control item, the working procedures of the DC submerged arc furnace, and the corresponding work content.

9. The electrode control method for a DC submerged arc furnace based on industrial big data according to claim 8, characterized in that, The process of identifying multiple balance control items based on the electrode balance control content, marking corresponding balance control nodes, determining the corresponding electrode control system based on the item content of each balance control item, the corresponding balance control node, and the electrode's lifespan, and triggering an early warning event for the electrode in the DC submerged arc furnace, also includes: The lifespan of the collected electrodes is comprehensively considered, taking into account the specific execution parameters of each balance control project, the timeliness requirements of key control nodes, and the lifespan of the electrodes. Based on multiple constraints, a corresponding electrode control system is constructed. This electrode control system monitors the difference between the operating parameters and the safety threshold in real time. If the difference exceeds the preset difference, an electrode early warning event of the DC submerged arc furnace is triggered, and the corresponding protection mode is executed.

10. An electrode control system for a DC submerged arc furnace based on industrial big data, characterized in that, The electrode control system of the DC submerged arc furnace based on industrial big data is applied to the electrode control method of the DC submerged arc furnace based on industrial big data as described in any one of claims 1-9.

Images