A DC submerged arc furnace and its control system

Abstract

The present invention relates to the technical field of the power supply system of submerged arc furnaces. A DC submerged arc furnace and its control system provided by the present invention aim to solve the problem in the prior art that the dynamic adjustment of DC submerged arc furnaces does not match the power supply system, resulting in a reduction in operating efficiency. By constructing a control device including data processing, data analysis, data comparison, and data optimization units, precise control of the DC submerged arc furnace equipment and the power supply device is achieved. A dynamic regulation model is constructed using a random forest model, and dynamic adjustment parameters are generated according to the actual operating conditions. The adjustment state is detected through a production simulation model, and the adjustment parameters are optimized using a genetic algorithm to improve the operating efficiency. The system also has functions such as model retraining, dynamic threshold adjustment, and multi-objective optimization, improving the adaptability and accuracy of power supply control.

Description

Technical Field

[0001] The present invention relates to the technical field of submerged arc furnace power supply systems, and particularly to a DC submerged arc furnace and its control system. Background Art

[0002] A DC submerged arc furnace is an industrial electric furnace that uses direct current to generate high temperatures for ore smelting or chemical production. Its core principle is to introduce direct current into the conductive furnace charge (such as coke or metal particles) in the furnace chamber through electrodes, and use the arc heat and resistance heat (up to over 2000 °C) generated when the current passes through the furnace charge to melt and reduce ores (such as chromite, manganese ore) or raw materials (such as calcium carbonate), producing products such as ferroalloys and calcium carbide. Compared with traditional AC submerged arc furnaces, DC furnaces adopt a single electrode structure, have a stable magnetic field and low electrode consumption, and have advantages such as high electrical energy utilization rate, concentrated heat efficiency, precise process control, and low environmental pollution. They are particularly suitable for large-scale production of high-melting-point and high-resistivity materials and are key equipment for efficient and low-carbon smelting in modern metallurgy and chemical industries.

[0003] In the power supply control of a DC submerged arc furnace, the real-time changes in material composition, furnace charge resistivity, and arc state during the smelting process are highly non-linear and time-varying, and real-time regulation depends on a high-precision dynamic model. During the regulation of the DC submerged arc furnace, not only the parameter adjustment of the DC submerged arc furnace needs to be considered, but also the power supply system of the DC submerged arc furnace needs to be adjusted. The existing technology cannot adaptively adjust the power supply system according to the dynamic adjustment of the DC submerged arc furnace, resulting in a mismatch between the power supply power of the power supply system and the power required for the operation of the DC submerged arc furnace after parameter adjustment, leading to a reduction in the operating efficiency of the DC submerged arc furnace. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention provides a DC submerged arc furnace and its control system to solve the problem that the existing technology cannot adaptively adjust the power supply system according to the dynamic adjustment of the DC submerged arc furnace, resulting in a mismatch between the power supply power of the power supply system and the power required for the operation of the DC submerged arc furnace after parameter adjustment, leading to a reduction in the operating efficiency of the DC submerged arc furnace.

[0005] To solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0006] In the first aspect, a DC submerged arc furnace provided by the present invention includes: a DC submerged arc furnace main body, a DC submerged arc furnace control device, DC submerged arc furnace equipment, and a DC submerged arc furnace power supply device. The DC submerged arc furnace control device is used to control and adjust the operation of the DC submerged arc furnace equipment and the DC submerged arc furnace power supply device;

[0007] The DC submerged arc furnace control device includes:

[0008] A data processing unit acquires the operating condition data of a DC submerged arc furnace. The operating condition data of the DC submerged arc furnace includes the production efficiency data, control parameter data, and power supply data of the DC submerged arc furnace. The operating condition data of the DC submerged arc furnace is preprocessed to obtain the preprocessed operating condition data of the DC submerged arc furnace;

[0009] An analysis unit constructs a characteristic data set of the DC submerged arc furnace based on the preprocessed operating condition data of the DC submerged arc furnace, and uses the characteristic data set of the DC submerged arc furnace to train a random forest model to obtain a dynamic regulation model of the DC submerged arc furnace;

[0010] A data comparison unit generates dynamic adjustment parameters of the DC submerged arc furnace through the dynamic regulation model of the DC submerged arc furnace in response to a control instruction of the DC submerged arc furnace. The dynamic adjustment parameters of the DC submerged arc furnace include power supply adjustment parameters of the power supply equipment and equipment adjustment parameters of the DC submerged arc furnace;

[0011] A data optimization unit generates a simulated operating efficiency of the DC submerged arc furnace through a preset production simulation model of the DC submerged arc furnace, detects the dynamic adjustment state of the DC submerged arc furnace, optimizes the dynamic adjustment parameters of the DC submerged arc furnace according to the dynamic adjustment state of the DC submerged arc furnace, and uses the optimized dynamic adjustment parameters of the DC submerged arc furnace as execution parameters.

[0012] Furthermore, for the DC submerged arc furnace of the present invention, the analysis unit includes:

[0013] A data classification subunit is used to classify the preprocessed operating condition data of the DC submerged arc furnace according to preset dimensions. The preset dimensions include time dimension classification, equipment type classification, and process stage classification; A model training subunit constructs a characteristic data set of the DC submerged arc furnace based on the classified operating condition data of the DC submerged arc furnace. The characteristic data set of the DC submerged arc furnace contains the mapping relationship between equipment condition parameters and production efficiency parameters, and iteratively trains the random forest model through the mapping relationship; A model verification subunit uses the cross-validation method to verify the accuracy of the trained random forest model, and triggers a model retraining mechanism when the verification accuracy rate is lower than a preset threshold. The model retraining mechanism includes increasing the training sample size, adjusting the feature weight distribution, and optimizing the model hyperparameters.

[0014] Furthermore, for the DC submerged arc furnace of the present invention, the data comparison unit includes:

[0015] An instruction parsing unit for parsing the instruction types of DC submerged arc furnace control instructions, where the instruction types include real-time adjustment instructions and timed adjustment instructions; a dynamic parameter generation unit that calls a DC submerged arc furnace dynamic regulation model according to the instruction type to generate DC submerged arc furnace dynamic adjustment parameters matching the real-time working conditions; a parameter comparison unit that compares the generated DC submerged arc furnace dynamic adjustment parameters with a historical adjustment parameter database to calculate a parameter deviation value; an adjustment strategy generation unit that generates a parameter correction strategy when the parameter deviation value exceeds a preset safety range, and the parameter correction strategy includes:

[0016] In the case where the instruction type is a real-time adjustment instruction, a progressive parameter adjustment method is adopted;

[0017] In the case where the instruction type is a timed adjustment instruction, a batch parameter optimization method is adopted.

[0018] Further, for the DC submerged arc furnace of the present invention, the data optimization unit includes:

[0019] A data generation unit that collects the operation efficiency of the DC submerged arc furnace after executing the DC submerged arc furnace dynamic adjustment parameters, and substitutes the DC submerged arc furnace control instructions into a preset DC submerged arc furnace production simulation model to obtain the simulated operation efficiency of the DC submerged arc furnace;

[0020] A state detection unit that compares the simulated operation efficiency of the DC submerged arc furnace with the operation efficiency of the DC submerged arc furnace to obtain the operation efficiency error of the DC submerged arc furnace;

[0021] A parameter optimization unit that optimizes the DC submerged arc furnace dynamic adjustment parameters using a genetic algorithm based on the operation efficiency error of the DC submerged arc furnace to obtain the optimized DC submerged arc furnace dynamic adjustment parameters, and the optimized DC submerged arc furnace dynamic adjustment parameters include the optimized power supply adjustment parameters of the power supply equipment and the DC submerged arc furnace equipment adjustment parameters.

[0022] A data execution unit that sends the DC submerged arc furnace equipment adjustment parameters to the DC submerged arc furnace equipment for adjusting the operation parameters of the DC submerged arc furnace equipment, and sends the power supply adjustment parameters of the power supply equipment to the DC submerged arc furnace power supply device for adjusting the power supply parameters of the DC submerged arc furnace power supply device.

[0023] Further, for the DC submerged arc furnace of the present invention, the data optimization unit further includes:

[0024] A dynamic threshold adjustment module that dynamically adjusts the convergence threshold of the optimization algorithm according to the deviation rate between the simulated operation efficiency and the measured operation efficiency of the DC submerged arc furnace, and the convergence threshold adjustment rule includes:

[0025] When the deviation rate > 10%, a first-level strict convergence criterion is adopted;

[0026] When 5% < deviation rate ≤ 10%, the second-level conventional convergence criterion is adopted;

[0027] When the deviation rate ≤ 5%, the third-level loose convergence criterion is adopted;

[0028] The multi-objective optimization module jointly optimizes the dynamic adjustment parameters of the DC submerged arc furnace through constructing a comprehensive evaluation function including equipment energy consumption, production efficiency and equipment loss rate; the parameter constraint module imposes at least one of the following constraint conditions during the optimization process:

[0029] The adjustment range of the power supply adjustment parameter of the power supply equipment does not exceed ±15% of the rated value;

[0030] The adjustment frequency of the adjustment parameter of the DC submerged arc furnace equipment is limited to ≤ 3 times per minute;

[0031] The time interval between two adjacent parameter adjustments ≥ 10 seconds.

[0032] In a second aspect, the present invention provides a DC submerged arc furnace control system, which is applied to the DC submerged arc furnace, including a DC submerged arc furnace main body, a DC submerged arc furnace control device, DC submerged arc furnace equipment and a DC submerged arc furnace power supply device. The DC submerged arc furnace equipment is installed inside the DC submerged arc furnace main body, and the DC submerged arc furnace equipment is respectively communicatively connected with the DC submerged arc furnace power supply device and the DC submerged arc furnace control device;

[0033] The DC submerged arc furnace control device includes:

[0034] A data processing unit, which acquires the operating condition data of the DC submerged arc furnace. The operating condition data of the DC submerged arc furnace includes DC submerged arc furnace production efficiency data, DC submerged arc furnace control parameter data and DC submerged arc furnace power supply data, and preprocesses the operating condition data of the DC submerged arc furnace to obtain the preprocessed operating condition data of the DC submerged arc furnace;

[0035] A data analysis unit, which constructs a DC submerged arc furnace feature data set based on the preprocessed operating condition data of the DC submerged arc furnace, and uses the DC submerged arc furnace feature data set to train a random forest model to obtain a DC submerged arc furnace dynamic regulation model;

[0036] A data comparison unit, in response to a DC submerged arc furnace control instruction, generates DC submerged arc furnace dynamic adjustment parameters through the DC submerged arc furnace dynamic regulation model. The DC submerged arc furnace dynamic adjustment parameters include power supply adjustment parameters of the power supply equipment and adjustment parameters of the DC submerged arc furnace equipment;

[0037] The data optimization unit generates the simulated operation efficiency of the DC submerged arc furnace through a preset production simulation model of the DC submerged arc furnace, detects the dynamic adjustment state of the DC submerged arc furnace, optimizes the dynamic adjustment parameters of the DC submerged arc furnace according to the dynamic adjustment state of the DC submerged arc furnace, and uses the optimized dynamic adjustment parameters of the DC submerged arc furnace as execution parameters.

[0038] Further, in the DC submerged arc furnace control system of the present invention, the data analysis unit includes:

[0039] The data classification sub-unit is used to classify the preprocessed operation condition data of the DC submerged arc furnace according to preset dimensions, and the preset dimensions include time dimension classification, equipment type classification, and process stage classification; the model training sub-unit constructs a feature data set of the DC submerged arc furnace based on the classified operation condition data of the DC submerged arc furnace. The feature data set of the DC submerged arc furnace contains the mapping relationship between equipment condition parameters and production efficiency parameters, and iteratively trains the random forest model through the mapping relationship; the model verification sub-unit uses the cross-validation method to verify the accuracy of the trained random forest model, and triggers the model retraining mechanism when the verification accuracy rate is lower than the preset threshold. The model retraining mechanism includes increasing the training sample size, adjusting the feature weight distribution, and optimizing the model hyperparameters.

[0040] Further, in the DC submerged arc furnace control system of the present invention, the data comparison unit includes:

[0041] The instruction parsing unit is used to parse the instruction type of the DC submerged arc furnace control instruction, and the instruction type includes real-time adjustment instructions and timed adjustment instructions; the dynamic parameter generation unit calls the dynamic regulation model of the DC submerged arc furnace according to the instruction type to generate the dynamic adjustment parameters of the DC submerged arc furnace matching the real-time working conditions; the parameter comparison unit compares the generated dynamic adjustment parameters of the DC submerged arc furnace with the historical adjustment parameter database and calculates the parameter deviation value; the adjustment strategy generation unit generates a parameter correction strategy when the parameter deviation value exceeds the preset safety range. The parameter correction strategy includes:

[0042] In the case where the instruction type is a real-time adjustment instruction, an incremental parameter adjustment method is adopted;

[0043] In the case where the instruction type is a timed adjustment instruction, a batch parameter optimization method is adopted.

[0044] Further, in the DC submerged arc furnace control system of the present invention, the data optimization unit includes:

[0045] The data generation unit collects the operation efficiency of the DC submerged arc furnace after executing the dynamic adjustment parameters of the DC submerged arc furnace, and substitutes the DC submerged arc furnace control instruction into the preset production simulation model of the DC submerged arc furnace to obtain the simulated operation efficiency of the DC submerged arc furnace;

[0046] A status detection unit compares the simulated operation efficiency of a DC submerged arc furnace with its actual operation efficiency to obtain the operation efficiency error of the DC submerged arc furnace.

[0047] A parameter optimization unit uses a genetic algorithm to optimize the dynamic adjustment parameters of the DC submerged arc furnace based on the operation efficiency error of the DC submerged arc furnace, obtaining optimized dynamic adjustment parameters for the DC submerged arc furnace. The optimized dynamic adjustment parameters of the DC submerged arc furnace include optimized power supply adjustment parameters of the power supply equipment and equipment adjustment parameters of the DC submerged arc furnace.

[0048] A data execution unit sends the equipment adjustment parameters of the DC submerged arc furnace to the DC submerged arc furnace equipment to adjust the operation parameters of the DC submerged arc furnace equipment, and sends the power supply adjustment parameters of the power supply equipment to the power supply device of the DC submerged arc furnace to adjust the power supply parameters of the power supply device of the DC submerged arc furnace.

[0049] Furthermore, for the DC submerged arc furnace control system of the present invention, the data optimization unit further includes:

[0050] A dynamic threshold adjustment module dynamically adjusts the convergence threshold of the optimization algorithm according to the deviation rate between the simulated operation efficiency and the measured operation efficiency of the DC submerged arc furnace. The convergence threshold adjustment rules include:

[0051] When the deviation rate > 10%, a first-level strict convergence criterion is adopted;

[0052] When 5% < deviation rate ≤ 10%, a second-level conventional convergence criterion is adopted;

[0053] When the deviation rate ≤ 5%, a third-level loose convergence criterion is adopted;

[0054] A multi-objective optimization module performs multi-objective joint optimization on the dynamic adjustment parameters of the DC submerged arc furnace by constructing a comprehensive evaluation function including equipment energy consumption, production efficiency, and equipment loss rate; a parameter constraint module imposes at least one of the following constraint conditions during the optimization process:

[0055] The adjustment range of the power supply adjustment parameters of the power supply equipment does not exceed ±15% of the rated value;

[0056] The adjustment frequency of the equipment adjustment parameters of the DC submerged arc furnace is limited to ≤ 3 times per minute;

[0057] The time interval between two adjacent parameter adjustments ≥ 10 seconds.

[0058] Advantages of the present invention:

[0059] The beneficial effects of the present invention are reflected in that by constructing a data-driven closed-loop control system, the problem of inaccurate power matching between the dynamic adjustment of the DC submerged arc furnace and the power supply system is systematically solved. The dynamic regulation algorithm based on the random forest model captures the complex non-linear relationship between the arc heat distribution, material resistivity, and power supply parameters, improving the generation accuracy of adjustment parameters under real-time working conditions; the multi-objective optimization module balances the competitive relationship among equipment energy consumption, production efficiency, and loss rate within the hard constraint boundary through the genetic algorithm, avoiding system imbalance caused by single-parameter optimization; the dynamic threshold adjustment mechanism adaptively converges the conditions according to the deviation rate between simulation and actual measurement efficiency, enhancing the response speed and optimization efficiency of the algorithm to process fluctuations; the parameter constraint module combines value range limitation, time series verification, and penalty functions to ensure that the adjustment range of the power supply voltage, the adjustment frequency, and the interval time of equipment parameters comply with engineering safety specifications. Through the closed-loop link of data collection, model prediction, deviation correction, and execution feedback, the dynamic synchronous matching of power supply and equipment requirements is realized, effectively reducing the production efficiency loss and abnormal equipment loss caused by model lag and rough adjustment in traditional technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] To more clearly illustrate the technical solutions of the present invention, the accompanying drawings required for use in the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the drawings.

[0061] Figure 1 It is a schematic diagram of a virtual device of a DC submerged arc furnace provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments and corresponding drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention will be described in detail below with reference to the drawings.

[0063] To better understand the objectives of the present invention, the present invention will be further described in detail below.

[0064] In a first aspect, please refer to Figure 1 , a DC submerged arc furnace provided by the present invention includes: a DC submerged arc furnace main body, a DC submerged arc furnace control device, DC submerged arc furnace equipment, and a DC submerged arc furnace power supply device. The DC submerged arc furnace control device is used to control and adjust the operation of the DC submerged arc furnace equipment and the DC submerged arc furnace power supply device;

[0065] The DC submerged arc furnace control device includes:

[0066] A data processing unit that acquires the operating condition data of the DC submerged arc furnace. The operating condition data of the DC submerged arc furnace includes the production efficiency data, control parameter data, and power supply data of the DC submerged arc furnace. The operating condition data of the DC submerged arc furnace is preprocessed to obtain the preprocessed operating condition data of the DC submerged arc furnace.

[0067] A data analysis unit that constructs a feature dataset of the DC submerged arc furnace based on the preprocessed operating condition data of the DC submerged arc furnace, and uses the feature dataset of the DC submerged arc furnace to train a random forest model to obtain a dynamic regulation model of the DC submerged arc furnace.

[0068] A data comparison unit that, in response to a control instruction for the DC submerged arc furnace, generates dynamic adjustment parameters for the DC submerged arc furnace through the dynamic regulation model of the DC submerged arc furnace. The dynamic adjustment parameters for the DC submerged arc furnace include power supply adjustment parameters for power supply equipment and equipment adjustment parameters for the DC submerged arc furnace.

[0069] A data optimization unit that generates a simulated operating efficiency of the DC submerged arc furnace through a preset production simulation model of the DC submerged arc furnace, detects the dynamic adjustment state of the DC submerged arc furnace, optimizes the dynamic adjustment parameters of the DC submerged arc furnace according to the dynamic adjustment state of the DC submerged arc furnace, and uses the optimized dynamic adjustment parameters of the DC submerged arc furnace as execution parameters.

[0070] A DC submerged arc furnace provided by the present invention has a control device that realizes dynamic regulation and optimization through the collaborative action of multiple technical modules. The control device first collects the operating condition data of the DC submerged arc furnace through the data processing unit, including production efficiency, control parameters, and power supply data, and performs data cleaning, normalization, and denoising processing on the data to eliminate the interference of outliers and dimensional differences on subsequent analysis, and generates a standardized operating condition dataset. This preprocessing process provides high-quality input for subsequent model training and parameter optimization.

[0071] In the data analysis unit, the standardized data is classified according to the time dimension, equipment type, and process stage, and a feature dataset including the mapping relationship between equipment condition parameters and production efficiency parameters is constructed. The feature data is iteratively trained through a random forest model, and the decision tree integration method is used to capture non-linear relationships to generate a dynamic regulation model. During the model training process, the cross-validation method is introduced to evaluate the accuracy of the model. When the validation accuracy rate is lower than the preset threshold, a retraining mechanism is triggered, and the generalization ability of the model is improved by increasing the training sample size, adjusting the feature weights, and optimizing the hyperparameters to ensure the adaptability of the dynamic regulation model to the real-time working conditions.

[0072] After the data comparison unit analyzes the control instruction type, it calls the dynamic regulation model to generate dynamic adjustment parameters matching the real-time working conditions. The generated parameters are compared with the historical adjustment parameter database, the deviation value is calculated, and a correction strategy is generated based on the preset safety range. Under the real-time adjustment instruction, a progressive parameter adjustment method is adopted to avoid system fluctuations through small-scale and multiple adjustments; under the timed adjustment instruction, a batch optimization method is adopted to conduct global optimization in combination with the historical optimal parameter set to balance the adjustment efficiency and stability.

[0073] The data optimization unit predicts the operating efficiency after executing the parameters through the production simulation model, and generates an efficiency error by comparing with the measured efficiency. Based on the error value, the genetic algorithm is used to perform multi-objective optimization on the adjustment parameters, and the optimal solution is iteratively searched through selection, crossover, and mutation operations. At the same time, constraint conditions such as the adjustment range of the power supply parameters, the adjustment frequency of the equipment, and the time interval are imposed to avoid over-adjustment or equipment loss. During the optimization process, the convergence threshold of the genetic algorithm is dynamically adjusted, and strict, conventional, or loose convergence criteria are set according to the deviation rate between the simulated efficiency and the measured efficiency to improve the optimization efficiency. The optimized parameters are synchronously sent to the power supply device and the equipment for execution to form a closed-loop control and achieve the dynamic matching of the power supply power and the equipment operating state.

[0074] The multi-objective optimization module further constructs a comprehensive evaluation function including equipment energy consumption, production efficiency, and equipment loss rate, and balances multi-objective conflicts through weighted summation or Pareto front methods to ensure that the control parameters reduce the comprehensive cost while meeting the process requirements. The parameter constraint module restricts the physical boundaries and operation frequencies of the adjustment parameters to ensure the safe operation of the system. Each technical module forms a collaborative mechanism through data flow and logical association, and finally realizes the synchronous improvement of the operating efficiency of the DC submerged arc furnace and the adaptability of the power supply system.

[0075] Specifically, for the DC submerged arc furnace described in the present invention, the data analysis unit includes:

[0076] The data classification sub-unit is used to classify the preprocessed operating condition data of the DC submerged arc furnace according to preset dimensions, and the preset dimensions include time dimension classification, equipment type classification, and process stage classification; the model training sub-unit constructs a characteristic data set of the DC submerged arc furnace based on the classified operating condition data of the DC submerged arc furnace. The characteristic data set of the DC submerged arc furnace contains the mapping relationship between equipment condition parameters and production efficiency parameters, and the random forest model is iteratively trained through the mapping relationship; the model verification sub-unit uses the cross-validation method to verify the accuracy of the trained random forest model, and triggers the model retraining mechanism when the verification accuracy rate is lower than the preset threshold. The model retraining mechanism includes increasing the training sample size, adjusting the feature weight distribution, and optimizing the model hyperparameters.

[0077] The data analysis unit realizes the construction and optimization of the dynamic regulation model through phased processing. The data classification subunit divides the preprocessed operating condition data into condition segments within continuous production cycles according to the time dimension, distinguishes the operating parameters of the electrode system, cooling device, and power supply module according to the equipment type, and separates the data sets of the charging, smelting, and discharging links according to the process stages. This classification method structurally reorganizes the multi-dimensional heterogeneous data by establishing a data label system, providing a context association basis for subsequent feature extraction.

[0078] The model training subunit extracts the mapping relationship between the equipment voltage, current, temperature condition parameters and the unit time output and energy consumption efficiency based on the classified data set, and constructs a multi-dimensional feature matrix including time series features, equipment status features, and process stage features. Using the parallel decision tree structure of the random forest model, it conducts distributed learning on the non-linear associations in the feature matrix, and gradually fits the optimal decision path of the equipment condition and production efficiency through multiple rounds of iteration. During the training process, the out-of-bag error estimation method is used to dynamically adjust the decision tree depth and node splitting criteria, enhancing the adaptability of the model to dynamic conditions.

[0079] The model verification subunit uses the K-fold cross-validation method to divide the training set into multiple mutually exclusive subsets, and sequentially uses different subsets as verification data to evaluate the model prediction accuracy. When the verification accuracy is lower than the preset threshold, it triggers the model retraining mechanism: expand the training samples to cover extreme condition data such as high load and low resistivity, adjust the feature weights to enhance the sensitivity to power supply parameter fluctuations, and optimize hyperparameters such as the number of decision trees, maximum depth, and minimum number of samples in leaf nodes through grid search. The retrained model is reconnected to the verification process until the accuracy requirement is met, forming an adaptive model update closed-loop.

[0080] Each subunit is connected in series through data hierarchical processing and feedback mechanisms. Data classification provides structured condition information for model training, and model verification improves the generalization ability of the model through dynamic evaluation and parameter iteration. The definition of the classification dimension and feature mapping relationship directly affects the quality of the model input space, while the cross-validation and retraining mechanisms ensure the dynamic matching of the model output with the real-time conditions, jointly supporting the accurate generation of regulation parameters.

[0081] Specifically, for the DC submerged arc furnace described in the present invention, the data comparison unit includes:

[0082] Instruction parsing unit, which is used to parse the instruction types of the DC submerged arc furnace control instructions, and the instruction types include real-time adjustment instructions and timed adjustment instructions; dynamic parameter generation unit, which calls the DC submerged arc furnace dynamic regulation model according to the instruction type and generates DC submerged arc furnace dynamic adjustment parameters matching the real-time working conditions; parameter comparison unit, which compares the generated DC submerged arc furnace dynamic adjustment parameters with the historical adjustment parameter database and calculates the parameter deviation value; adjustment strategy generation unit, when the parameter deviation value exceeds the preset safety range, generates a parameter correction strategy, and the parameter correction strategy includes:

[0083] In the case that the instruction type is a real-time adjustment instruction, an incremental parameter adjustment method is adopted;

[0084] In the case that the instruction type is a timed adjustment instruction, a batch parameter optimization method is adopted.

[0085] The data comparison unit realizes the dynamic adaptation and correction of the control parameters through hierarchical logical processing. The instruction parsing unit distinguishes real-time adjustment instructions and timed adjustment instructions based on the metadata identifier of the control instruction: real-time instructions carry an instant timestamp and a priority mark, triggering a millisecond-level response mechanism; timed instructions contain preset cycle parameters, activating a periodic task scheduling module. During the parsing process, the process stage code and the target parameter threshold in the instruction are extracted to provide context constraint conditions for subsequent parameter generation.

[0086] The dynamic parameter generation unit selects the call mode of the dynamic regulation model according to the instruction type: under real-time instructions, the streaming calculation interface of the model is activated, and the power supply voltage and power adjustment parameters are generated based on the real-time data of the current electrode current, furnace temperature and material resistivity; under timed instructions, the batch prediction interface of the model is called, and multiple sets of equipment operation parameter combinations are generated in combination with the historical same-period working condition data and the production plan target. The generated dynamic adjustment parameters are attached with a timestamp and a version identifier, forming an associated mapping with the optimal parameter record in the historical database.

[0087] The parameter comparison unit matches the current working condition characteristics with the historical records through database indexing, and calculates the power supply voltage deviation rate, the power adjustment amplitude difference value and the equipment parameter offset. The deviation calculation adopts a weighted comprehensive evaluation method, where the voltage deviation weight reflects the stability requirements of the power supply system, and the equipment parameter offset weight is associated with the equipment loss risk. The comparison result generates a structured report including the deviation type, the deviation degree and the impact level, providing a quantitative basis for strategy generation.

[0088] The adjustment strategy generation unit selects a correction mode according to the safety level division in the deviation report: under the real-time adjustment instruction, progressive parameter adjustment is adopted, approaching the target value step by step with a preset gradient. The voltage fluctuation is limited to no more than 2% of the reference value within each adjustment cycle, and the single adjustment amount of the equipment parameters is controlled within 30% of the historical maximum variation range; under the timed adjustment instruction, batch parameter optimization is started. An effective parameter set under similar working conditions is screened from the historical database, and a new parameter group that meets multi-objective constraints is generated through genetic algorithm crossover and mutation. After simulation verification, the equipment control sequence is updated in batches. Progressive adjustment gives priority to ensuring the real-time stability of the system, while batch optimization focuses on improving the overall efficiency. The two modes achieve a balance between control accuracy and resource consumption through instruction type diversion.

[0089] Each sub-unit forms a closed-loop processing link driven by the instruction flow: instruction parsing determines the control mode boundary, parameter generation establishes a dynamic response baseline, deviation comparison quantifies the adjustment requirements, and strategy generation outputs the execution rules. The historical database serves as an empirical knowledge base, providing both a reference benchmark for parameter comparison and accumulating case data for strategy optimization, realizing the continuous self-optimization ability of the control system.

[0090] Specifically, for the DC submerged arc furnace described in the present invention, the data optimization unit includes:

[0091] A data generation unit, which collects the operation efficiency of the DC submerged arc furnace after executing the dynamic adjustment parameters of the DC submerged arc furnace, substitutes the DC submerged arc furnace control instruction into the preset DC submerged arc furnace production simulation model, and obtains the simulated operation efficiency of the DC submerged arc furnace;

[0092] A state detection unit, which compares the simulated operation efficiency of the DC submerged arc furnace with the operation efficiency of the DC submerged arc furnace to obtain the operation efficiency error of the DC submerged arc furnace;

[0093] A parameter optimization unit, which optimizes the dynamic adjustment parameters of the DC submerged arc furnace using a genetic algorithm based on the operation efficiency error of the DC submerged arc furnace, and obtains the optimized dynamic adjustment parameters of the DC submerged arc furnace. The optimized dynamic adjustment parameters of the DC submerged arc furnace include the optimized power supply adjustment parameters of the power supply equipment and the equipment adjustment parameters of the DC submerged arc furnace.

[0094] A data execution unit, which sends the equipment adjustment parameters of the DC submerged arc furnace to the DC submerged arc furnace equipment for adjusting the operation parameters of the DC submerged arc furnace equipment, and sends the power supply adjustment parameters of the power supply equipment to the power supply device of the DC submerged arc furnace for adjusting the power supply parameters of the power supply device of the DC submerged arc furnace.

[0095] The data optimization unit realizes the adaptive optimization and execution of control parameters through multi-level linkage technical means. The data generation unit collects the actual operation efficiency data of the DC submerged arc furnace after dynamically adjusting the parameter execution, including the output per unit time, energy consumption index, and equipment status parameters, and inputs the current control instruction into a preset production simulation model. This model is constructed based on historical operation data and process constraints, and simulates the coupling relationship of the arc heat distribution, material melting rate, and power supply power in the furnace through physical equations to generate the simulated operation efficiency under the corresponding instruction. The simulation results are attached with operating condition labels, forming a spatio-temporal aligned comparison benchmark with the actual operation data.

[0096] The state detection unit conducts multi-dimensional comparison of the simulated efficiency and the actual efficiency, calculates the output deviation rate, energy consumption difference value, and equipment status offset. During the comparison process, a sliding window mechanism is introduced to extract the efficiency fluctuation characteristics within a continuous time series, and the comprehensive error value is calculated by combining the process stage division with weighted calculation. The error analysis results are marked as levels such as normal fluctuation, local deviation, or system mismatch, providing classified input for subsequent optimization.

[0097] The parameter optimization unit starts the genetic algorithm optimization process based on the error level: the initial population consists of the historical optimal parameter set and the current dynamically adjusted parameters, and the fitness function is used to evaluate the correction ability of individual parameters to the comprehensive error. The selection operation retains the individuals with high fitness, the crossover operation performs segment exchange between the power supply parameters and the equipment parameters to explore the solution space, and the mutation operation performs small random perturbations on the parameter values to enhance diversity. During the optimization process, dynamic constraint conditions are embedded to limit the adjustment range of the power supply voltage not to exceed ±15% of the rated value, and the adjustment frequency of the equipment parameters is controlled within 3 times per minute. After iterative convergence, the non-dominated solution set on the Pareto front is output, and the optimal parameter combination that takes into account both efficiency improvement and equipment loss is selected by combining the multi-objective weights.

[0098] The data execution unit classifies and distributes the optimized parameters according to the instruction type: the power supply adjustment parameters of the power supply equipment are transmitted to the rectifier control module through a real-time communication protocol to dynamically adjust the output current and voltage waveforms; the equipment adjustment parameters of the DC submerged arc furnace are written into the parameter queue of the equipment controller after verification, and the action of the execution mechanism such as the electrode lifting rate and the cooling water flow is updated according to the preset time stamp or trigger condition. During the distribution process, a redundant verification and rollback mechanism is adopted. If an abnormal alarm is triggered after the parameter execution, it will automatically switch to the previous valid parameter set to maintain the continuity of the system operation.

[0099] Each sub-unit forms an iterative optimization link through data closed-loop: data generation establishes the mapping relationship between simulation and reality, state detection quantifies the control effect, parameter optimization generates improvement plans, and data execution realizes the response of the physical system. The production simulation model serves as a virtual test environment to support the pre-verification of parameter optimization; the genetic algorithm approaches the global optimal solution through multiple generations of evolution; the parameter feedback mechanism of the execution unit provides incremental data input for the next round of optimization, ultimately achieving continuous improvement in the operation efficiency of the DC submerged arc furnace and the adaptability of the power supply system.

[0100] Specifically, for the DC submerged arc furnace described in the present invention, the data optimization unit further includes:

[0101] A dynamic threshold adjustment module that dynamically adjusts the convergence threshold of the optimization algorithm according to the deviation rate between the simulated operation efficiency and the measured operation efficiency of the DC submerged arc furnace. The convergence threshold adjustment rules include:

[0102] When the deviation rate > 10%, a first-level strict convergence criterion is adopted;

[0103] When 5% < deviation rate ≤ 10%, a second-level conventional convergence criterion is adopted;

[0104] When the deviation rate ≤ 5%, a third-level loose convergence criterion is adopted;

[0105] A multi-objective optimization module that jointly optimizes multiple objectives for the dynamically adjusted parameters of the DC submerged arc furnace by constructing a comprehensive evaluation function including equipment energy consumption, production efficiency, and equipment loss rate; a parameter constraint module that imposes at least one of the following constraint conditions during the optimization process:

[0106] The adjustment range of the power supply adjustment parameters of the power supply equipment does not exceed ±15% of the rated value;

[0107] The adjustment frequency of the adjustment parameters of the DC submerged arc furnace equipment is limited to ≤ 3 times per minute;

[0108] The time interval between two adjacent parameter adjustments ≥ 10 seconds.

[0109] The data optimization unit enhances the adaptability and security of parameter optimization through a hierarchical control mechanism. The dynamic threshold adjustment module monitors the deviation rate between the simulated operation efficiency and the measured efficiency in real time, and dynamically sets the convergence threshold of the genetic algorithm based on the degree of deviation: when the deviation rate exceeds 10%, the first-level strict convergence criterion is activated, the convergence condition is set as the fitness change rate of the optimal solution of the population being less than 0.5% for 20 consecutive generations, and the population size is forced to increase to 1.5 times the benchmark value to enhance the global search ability; when the deviation rate is between 5% and 10%, the second-level conventional convergence criterion is enabled, requiring the fitness change rate to be less than 1% for 15 consecutive generations, and maintaining the standard population size; when the deviation rate is less than 5%, it switches to the third-level loose convergence criterion, allowing the fitness change rate to be within 2% for 10 consecutive generations, and at the same time reducing the number of iterations to reduce the computational load. The threshold grading mechanism realizes the dynamic balance between optimization accuracy and computational efficiency.

[0110] The comprehensive evaluation function constructed by the multi-objective optimization module integrates the equipment energy consumption weight coefficient, the production efficiency gain coefficient, and the equipment loss rate penalty coefficient, and transforms the multi-objective into a single-objective optimization problem through linear weighting. The weight coefficients are dynamically configured according to the process priority: the production efficiency weight is increased during the smelting stage, and the energy consumption optimization weight is strengthened during the steady-state operation stage. In the optimization process, the non-dominated sorting genetic algorithm (NSGA-II) is used to generate the Pareto optimal solution set, and the diversity of the solution set is maintained through crowding comparison. Finally, the operator or the preset strategy selects the parameter combination with the highest comprehensive score. This design takes into account the competitive relationship between multiple objectives and avoids system imbalance caused by single-index optimization.

[0111] The parameter constraint module embeds hard boundary conditions in the optimization iteration stage: the adjustment range constraint of the power supply equipment is realized through the value range limit during parameter coding, and the power supply voltage adjustment value is clamped within the range of ±15% of the rated value during algorithm initialization; the equipment adjustment frequency constraint is transformed into a time series optimization problem, and a penalty term for violating the frequency is introduced into the fitness function of the genetic algorithm, and the fitness of individuals exceeding 3 adjustments per minute is attenuated proportionally; the adjacent adjustment interval constraint is implemented through a timestamp verification mechanism, and candidate solutions with an interval less than 10 seconds are filtered out during the parameter mutation operation. The multi-level embedding method of the constraint conditions not only ensures the safe operation of the physical system but also maintains the search freedom of the optimization process.

[0112] Each module forms a closed-loop optimization system through collaborative actions: the dynamic threshold adjustment module regulates the convergence characteristics of the algorithm according to real-time deviations, the multi-objective optimization module explores the optimal parameter combination within the constraint framework, and the parameter constraint module ensures the engineering feasibility of the output solution. The linkage between the deviation rate grading and the convergence threshold improves the response speed of the algorithm to operating condition fluctuations; the multi-objective weight configuration and the constraint penalty mechanism jointly guide the parameter optimization direction; the combination of hard boundaries and soft penalties achieves the unity of safety and optimization effect. The historical optimal solution set serves as an important source of the initial population, and the continuously accumulated case data optimizes the weight coefficient and the constraint threshold setting through the feedback mechanism, promoting the progressive improvement of the overall control performance of the system.

[0113] In a second aspect, the present invention provides a DC submerged arc furnace control system, which is applied to the DC submerged arc furnace. The DC submerged arc furnace control system includes a DC submerged arc furnace main body, a DC submerged arc furnace control device, DC submerged arc furnace equipment, and a DC submerged arc furnace power supply device. The DC submerged arc furnace equipment is installed inside the DC submerged arc furnace main body, and the DC submerged arc furnace equipment is respectively communicatively connected to the DC submerged arc furnace power supply device and the DC submerged arc furnace control device;

[0114] The DC submerged arc furnace control device includes:

[0115] a data processing unit, which acquires the operating condition data of the DC submerged arc furnace. The operating condition data of the DC submerged arc furnace includes the production efficiency data of the DC submerged arc furnace, the control parameter data of the DC submerged arc furnace, and the power supply data of the DC submerged arc furnace, and preprocesses the operating condition data of the DC submerged arc furnace to obtain the preprocessed operating condition data of the DC submerged arc furnace;

[0116] a data analysis unit, which constructs a characteristic data set of the DC submerged arc furnace based on the preprocessed operating condition data of the DC submerged arc furnace, and uses the characteristic data set of the DC submerged arc furnace to train a random forest model to obtain a dynamic regulation model of the DC submerged arc furnace;

[0117] a data comparison unit, which, in response to a control instruction of the DC submerged arc furnace, generates dynamic adjustment parameters of the DC submerged arc furnace through the dynamic regulation model of the DC submerged arc furnace. The dynamic adjustment parameters of the DC submerged arc furnace include power supply adjustment parameters of the power supply equipment and equipment adjustment parameters of the DC submerged arc furnace;

[0118] a data optimization unit, which generates a simulated operating efficiency of the DC submerged arc furnace through a preset production simulation model of the DC submerged arc furnace, detects the dynamic adjustment state of the DC submerged arc furnace, optimizes the dynamic adjustment parameters of the DC submerged arc furnace according to the dynamic adjustment state of the DC submerged arc furnace, and uses the optimized dynamic adjustment parameters of the DC submerged arc furnace as execution parameters.

[0119] Specifically, for the DC submerged arc furnace control system of the present invention, the data analysis unit includes:

[0120] A data classification subunit for classifying the preprocessed DC submerged arc furnace operation condition data according to preset dimensions, where the preset dimensions include time dimension classification, equipment type classification, and process stage classification; a model training subunit for constructing a DC submerged arc furnace feature data set based on the classified DC submerged arc furnace operation condition data, where the DC submerged arc furnace feature data set contains the mapping relationship between equipment condition parameters and production efficiency parameters, and iteratively training a random forest model through the mapping relationship; a model verification subunit for verifying the accuracy of the trained random forest model using the cross-validation method, and triggering a model retraining mechanism when the verification accuracy rate is lower than a preset threshold, where the model retraining mechanism includes increasing the training sample size, adjusting the feature weight distribution, and optimizing the model hyperparameters.

[0121] Specifically, for the DC submerged arc furnace control system of the present invention, the data comparison unit includes:

[0122] An instruction parsing unit for parsing the instruction type of the DC submerged arc furnace control instruction, where the instruction type includes a real-time adjustment instruction and a timing adjustment instruction; a dynamic parameter generation unit for calling a DC submerged arc furnace dynamic regulation model according to the instruction type to generate DC submerged arc furnace dynamic adjustment parameters matching the real-time working conditions; a parameter comparison unit for comparing the generated DC submerged arc furnace dynamic adjustment parameters with a historical adjustment parameter database and calculating a parameter deviation value; an adjustment strategy generation unit for generating a parameter correction strategy when the parameter deviation value exceeds a preset safety range, and the parameter correction strategy includes:

[0123] In the case where the instruction type is a real-time adjustment instruction, an incremental parameter adjustment method is adopted;

[0124] In the case where the instruction type is a timing adjustment instruction, a batch parameter optimization method is adopted.

[0125] Specifically, for the DC submerged arc furnace control system of the present invention, the data optimization unit includes:

[0126] A data generation unit for collecting the operation efficiency of the DC submerged arc furnace after executing the DC submerged arc furnace dynamic adjustment parameters, and substituting the DC submerged arc furnace control instruction into a preset DC submerged arc furnace production simulation model to obtain the simulated operation efficiency of the DC submerged arc furnace;

[0127] A state detection unit for comparing the simulated operation efficiency of the DC submerged arc furnace with the operation efficiency of the DC submerged arc furnace to obtain the operation efficiency error of the DC submerged arc furnace;

[0128] The parameter optimization unit optimizes the dynamic adjustment parameters of the DC submerged arc furnace using a genetic algorithm based on the operation efficiency error of the DC submerged arc furnace, and obtains the optimized dynamic adjustment parameters of the DC submerged arc furnace. The optimized dynamic adjustment parameters of the DC submerged arc furnace include the power supply adjustment parameters of the optimized power supply equipment and the equipment adjustment parameters of the DC submerged arc furnace.

[0129] The data execution unit sends the equipment adjustment parameters of the DC submerged arc furnace to the DC submerged arc furnace equipment for adjusting the operation parameters of the DC submerged arc furnace equipment, and sends the power supply adjustment parameters of the power supply equipment to the power supply device of the DC submerged arc furnace for adjusting the power supply parameters of the power supply device of the DC submerged arc furnace.

[0130] Specifically, for the DC submerged arc furnace control system described in the present invention, the data optimization unit further includes:

[0131] The dynamic threshold adjustment module dynamically adjusts the convergence threshold of the optimization algorithm according to the deviation rate between the simulated operation efficiency and the measured operation efficiency of the DC submerged arc furnace. The convergence threshold adjustment rules include:

[0132] When the deviation rate > 10%, a first-level strict convergence criterion is adopted;

[0133] When 5% < deviation rate ≤ 10%, a second-level conventional convergence criterion is adopted;

[0134] When the deviation rate ≤ 5%, a third-level loose convergence criterion is adopted;

[0135] The multi-objective optimization module jointly optimizes the dynamic adjustment parameters of the DC submerged arc furnace through constructing a comprehensive evaluation function including equipment energy consumption, production efficiency, and equipment loss rate; the parameter constraint module imposes at least one of the following constraint conditions during the optimization process:

[0136] The adjustment range of the power supply adjustment parameters of the power supply equipment does not exceed ±15% of the rated value;

[0137] The adjustment frequency of the equipment adjustment parameters of the DC submerged arc furnace is limited to ≤ 3 times per minute;

[0138] The time interval between two adjacent parameter adjustments ≥ 10 seconds.

[0139] The explanations of the technical feature terms in the technical solution of the present invention are as follows:

[0140] The main body of the DC submerged arc furnace: refers to the core equipment for ore smelting, which forms arc heat and resistance heat between the electrode and the conductive charge through direct current to achieve high-temperature smelting. Its single-electrode structure has the characteristics of high magnetic field stability and low electrode loss compared with the AC furnace, and is suitable for large-scale production of high-resistivity materials.

[0141] DC submerged arc furnace control device: An intelligent control module integrating data processing, analysis, comparison, and optimization functions. By collecting real-time operating conditions data such as voltage, current, and temperature, it dynamically coordinates the operating parameters of the power supply system and equipment to solve the power matching problem. The core lies in constructing a closed-loop control link to achieve dynamic adaptation of process parameters.

[0142] Data processing unit: A preprocessing module responsible for cleaning, normalizing, and time-series alignment of raw operating conditions data (such as production efficiency, control parameters, power supply data). By eliminating noise interference and dimensional differences, it generates a standardized data set to provide a high signal-to-noise ratio input for subsequent modeling.

[0143] Data analysis unit: A model training module based on the random forest algorithm. Through classification and regression tree ensemble learning, it maps the relationship between equipment operating conditions parameters (such as arc length, material resistivity) and production efficiency. The cross-validation and retraining mechanisms enable the model to adapt to time-varying operating conditions such as fluctuations in material composition and changes in furnace charge resistivity, improving the dynamic regulation accuracy.

[0144] Data comparison unit: A logic processing module including instruction parsing, parameter generation, and deviation analysis. It real-time parses the type of control instruction (real-time / timing), calls the dynamic regulation model to generate adjustment parameters, and performs weighted deviation calculation with the optimal parameters in the historical database to identify the matching degree difference between the power supply power and equipment requirements.

[0145] Data optimization unit: A multi-objective optimization module using the genetic algorithm. With equipment energy consumption, production efficiency, and loss rate as optimization objectives, it iteratively searches for the Pareto optimal solution within the constraint conditions (such as voltage adjustment range ±15%, adjustment frequency ≤ 3 times / minute). The dynamic threshold adjustment module grades and tightens or relaxes the convergence criteria according to the deviation rate between simulation and measured efficiency, balancing the optimization speed and accuracy.

[0146] Dynamic threshold adjustment module: An adaptive module that dynamically sets the genetic algorithm convergence conditions according to the deviation rate. When the deviation rate > 10%, a strict convergence criterion is adopted (such as the fitness change < 0.5% for 20 consecutive generations), and when the deviation rate ≤ 5%, a loose criterion is enabled (such as the change < 2% for 10 consecutive generations) to achieve the matching of algorithm resource allocation and operating condition fluctuations.

[0147] Multi-objective optimization module: Constructs a comprehensive evaluation function, and dynamically allocates the optimization weights of equipment energy consumption, efficiency, and loss through weighted coefficients. It focuses on production efficiency during the smelting stage and strengthens energy consumption optimization during the steady state stage. It uses non-dominated sorting (NSGA-II) to generate a solution set and selects a parameter combination that takes into account multiple objectives by combining crowding degree comparison.

[0148] Parameter Constraint Module: An optimization constraint module that embeds hard boundaries and soft penalties. The adjustment range of power supply parameters is restricted by value range clamping. The device adjustment frequency filters out illegal solutions through time series verification. The adjacent adjustment intervals are constrained by mutation operations to prevent equipment overload or arc instability.

[0149] Model Retraining Mechanism: A self-update process triggered when the model validation accuracy is lower than the threshold. By expanding samples of extreme working conditions, adjusting feature weights (such as enhancing voltage fluctuation sensitivity), and optimizing hyperparameters (number and depth of decision trees), the generalization ability of the model to new working conditions is improved.

[0150] The specific implementation of the present invention is based on the demand for dynamic matching of the power supply system and equipment operating status during the smelting of high-resistivity materials in a DC submerged arc furnace, and realizes precise control of process parameters through hierarchical control and closed-loop optimization. In the data acquisition stage, sensors deployed on the furnace electrodes, cooling system, and power supply module collect voltage, current, temperature, and material resistivity data in real time. After denoising, normalization, and time alignment processing by the data preprocessing unit, a multi-dimensional data set containing the smelting stage, equipment status, and power supply parameters is generated. The preprocessed data is divided into a charging and preheating period, an arc melting period, and a discharging and stabilizing period according to the process stage. Each stage corresponds to a different baseline of power supply power demand, providing a structured input for subsequent analysis.

[0151] The construction of the dynamic regulation model takes the random forest algorithm as the core and is trained using the mapping relationship between equipment condition parameters and production efficiency in historical data. During the training process, the model selects arc length, material bulk density, and cooling water flow rate as key input variables through feature importance analysis, and dynamically adjusts the depth and number of decision trees based on the cross-validation results. When the model detects that the prediction error under new working conditions exceeds 5% during the validation stage, the retraining process is triggered: expanding the training set to include recent extreme working condition data, optimizing the feature weight distribution to enhance the sensitivity to voltage fluctuations, and finally generating a dynamic regulation model adapted to the current smelting process. The model responds to the type of control instruction, generates power supply power adjustment suggestions and electrode lifting rate parameters in real time, and conducts deviation analysis with the historical optimal solution set through the parameter comparison unit.

[0152] In the parameter optimization and execution stage, the genetic algorithm takes the power supply power deviation rate, the increment of equipment loss, and the production efficiency improvement rate as the optimization objectives, and iteratively searches for the Pareto optimal solution within the constraint conditions. The adjustment range of the supply voltage is limited to ±15% of the rated value to avoid unstable arcs caused by over-adjustment; the adjustment frequency and interval time of the equipment parameters are constrained by the time series verification module to prevent the actuator from being overloaded. After the optimized parameters are verified by the simulation model, they are issued by the execution unit according to priorities: real-time instructions trigger the millisecond-level response of the power supply module, and timed instructions drive the batch parameter update of the equipment controller. The actual operation efficiency data is fed back to the simulation model to form a closed loop, and the dynamic threshold adjustment module tightens or relaxes the convergence conditions according to the deviation rate level, so that the system can quickly converge to a stable state during process fluctuations. Through the above implementation method, the output power of the power supply system is dynamically matched with the furnace body demand, effectively solving the problem of operation efficiency loss caused by parameter lag and multi-objective conflict in the traditional technology.

[0153] Aiming at the problem of insufficient adaptability between the dynamic adjustment of DC submerged arc furnaces and the power supply system in the prior art, the present invention proposes a technical solution of hierarchical control and multi-objective collaborative optimization. By constructing a dynamic control model based on random forest, multi-dimensional feature extraction and non-linear relationship modeling are performed on the operation condition data of DC submerged arc furnaces, and power supply power and equipment parameter adjustment instructions matching the real-time arc state and the change of material resistivity are generated. During the model training process, a cross-validation and re-training mechanism is introduced to enhance the adaptability to time-varying working conditions, solve the defect that traditional static models cannot capture dynamic process characteristics, and avoid inaccurate power supply power prediction caused by model deviation.

[0154] The data comparison unit identifies the matching degree deviation between the output power of the power supply system and the actual demand of the furnace body through the difference analysis of the historical parameter database and the real-time generated parameters. When it is detected that the deviation value exceeds the preset safety threshold, an incremental or batch optimization strategy is selected according to the instruction type: under the real-time adjustment instruction, a small step size is used to iteratively approach the target value, and under the timed adjustment instruction, the historical optimal parameter set is integrated for global optimization. A sliding window mechanism is embedded in the parameter correction process, and the correction amplitude is dynamically adjusted in combination with the characteristics of the process stage, overcoming the limitation of the traditional single correction mode's lag response to sudden working conditions, and realizing the dynamic synchronization of the power supply power and the equipment operation state.

[0155] The multi-objective optimization module uses a genetic algorithm to jointly optimize equipment energy consumption, production efficiency, and loss rate, generating a Pareto-optimal solution set within the hard boundaries of power supply adjustment amplitude, frequency, and interval set by the parameter constraint module. The dynamic threshold adjustment module sets the algorithm convergence criteria based on the deviation rate between simulated and measured efficiency, enhancing global search capabilities when the deviation rate is high and focusing on localized fine-grained optimization when the deviation rate is low. The power supply device and equipment actuators synchronize adjustments based on the optimized parameters, forming a closed-loop control chain. This solves the operational efficiency loss caused by delayed power supply system adjustment and multi-objective conflicts in traditional technologies.

Claims

1. A DC submerged arc furnace, characterized in that, Including: The DC submerged arc furnace main body, the DC submerged arc furnace control device, the DC submerged arc furnace equipment, and the DC submerged arc furnace power supply device. The DC submerged arc furnace control device is used to control and adjust the operation of the DC submerged arc furnace equipment and the DC submerged arc furnace power supply device; The DC submerged arc furnace control device includes: A data processing unit that acquires the operating condition data of the DC submerged arc furnace. The operating condition data of the DC submerged arc furnace includes the production efficiency data of the DC submerged arc furnace, the control parameter data of the DC submerged arc furnace, and the power supply data of the DC submerged arc furnace. The data processing unit preprocesses the operating condition data of the DC submerged arc furnace to obtain the preprocessed operating condition data of the DC submerged arc furnace; A data analysis unit that constructs a feature dataset of the DC submerged arc furnace based on the preprocessed operating condition data of the DC submerged arc furnace, and uses the feature dataset of the DC submerged arc furnace to train a random forest model to obtain a dynamic regulation model of the DC submerged arc furnace; A data comparison unit that, in response to a control instruction for the DC submerged arc furnace, generates dynamic adjustment parameters for the DC submerged arc furnace through the dynamic regulation model of the DC submerged arc furnace. The dynamic adjustment parameters for the DC submerged arc furnace include power supply adjustment parameters for the power supply equipment and adjustment parameters for the DC submerged arc furnace equipment; A data optimization unit that generates a simulated operating efficiency of the DC submerged arc furnace through a preset production simulation model of the DC submerged arc furnace, detects the dynamic adjustment state of the DC submerged arc furnace, optimizes the dynamic adjustment parameters of the DC submerged arc furnace according to the dynamic adjustment state of the DC submerged arc furnace, and uses the optimized dynamic adjustment parameters of the DC submerged arc furnace as execution parameters; The data analysis unit includes: A data classification subunit that classifies the preprocessed operating condition data of the DC submerged arc furnace according to a preset dimension. The preset dimension includes time dimension classification, equipment type classification, and process stage classification; A model training subunit that constructs a feature dataset of the DC submerged arc furnace based on the classified operating condition data of the DC submerged arc furnace. The feature dataset of the DC submerged arc furnace contains the mapping relationship between equipment condition parameters and production efficiency parameters, and iteratively trains the random forest model through the mapping relationship; A model verification subunit that uses the cross-validation method to verify the accuracy of the trained random forest model, and triggers a model retraining mechanism when the verification accuracy rate is lower than a preset threshold. The model retraining mechanism includes increasing the training sample size, adjusting the feature weight distribution, and optimizing the model hyperparameters; The data optimization unit includes: A data generation unit that collects the operating efficiency of the DC submerged arc furnace after executing the dynamic adjustment parameters of the DC submerged arc furnace, and substitutes the control instruction of the DC submerged arc furnace into the preset production simulation model of the DC submerged arc furnace to obtain the simulated operating efficiency of the DC submerged arc furnace; A state detection unit that compares the simulated operating efficiency of the DC submerged arc furnace with the operating efficiency of the DC submerged arc furnace to obtain an operating efficiency error of the DC submerged arc furnace; A parameter optimization unit that optimizes the dynamic adjustment parameters of the DC submerged arc furnace using a genetic algorithm based on the operating efficiency error of the DC submerged arc furnace to obtain the optimized dynamic adjustment parameters of the DC submerged arc furnace. The optimized dynamic adjustment parameters of the DC submerged arc furnace include the optimized power supply adjustment parameters for the power supply equipment and the adjustment parameters for the DC submerged arc furnace equipment. The data execution unit sends the adjustment parameters of the DC submerged arc furnace equipment to the DC submerged arc furnace equipment for adjusting the operating parameters of the DC submerged arc furnace equipment, and sends the power supply adjustment parameters of the power supply equipment to the power supply device of the DC submerged arc furnace for adjusting the power supply parameters of the power supply device of the DC submerged arc furnace.

2. The DC submerged arc furnace according to claim 1, wherein The data comparison unit includes: An instruction parsing unit for parsing the instruction type of the DC submerged arc furnace control instruction, where the instruction type includes a real-time adjustment instruction and a timing adjustment instruction; a dynamic parameter generation unit that calls the DC submerged arc furnace dynamic regulation model according to the instruction type to generate DC submerged arc furnace dynamic adjustment parameters matching the real-time working conditions; a parameter comparison unit that compares the generated DC submerged arc furnace dynamic adjustment parameters with the historical adjustment parameter database to calculate the parameter deviation value; an adjustment strategy generation unit that generates a parameter correction strategy when the parameter deviation value exceeds the preset safety range, and the parameter correction strategy includes: In the case where the instruction type is a real-time adjustment instruction, an incremental parameter adjustment method is adopted; In the case where the instruction type is a timing adjustment instruction, a batch parameter optimization method is adopted.

3. The DC submerged arc furnace according to claim 1, wherein The data optimization unit further includes: A dynamic threshold adjustment module that dynamically adjusts the convergence threshold of the optimization algorithm according to the deviation rate between the simulated operation efficiency and the measured operation efficiency of the DC submerged arc furnace, and the convergence threshold adjustment rule includes: When the deviation rate > 10%, a first-level strict convergence criterion is adopted; When 5% < deviation rate ≤ 10%, a second-level conventional convergence criterion is adopted; When the deviation rate ≤ 5%, a third-level loose convergence criterion is adopted; A multi-objective optimization module that jointly optimizes the DC submerged arc furnace dynamic adjustment parameters through constructing a comprehensive evaluation function including equipment energy consumption, production efficiency, and equipment loss rate; a parameter constraint module that imposes at least one of the following constraint conditions during the optimization process: The adjustment range of the power supply adjustment parameters of the power supply equipment does not exceed ±15% of the rated value; The adjustment frequency of the adjustment parameters of the DC submerged arc furnace equipment is limited to ≤ 3 times per minute; The time interval between two adjacent parameter adjustments ≥ 10 seconds.

4. A DC submerged arc furnace control system is applied to the DC submerged arc furnace according to any one of claims 1 to 3, characterized in that, It includes: The DC submerged arc furnace main body, the DC submerged arc furnace control device, the DC submerged arc furnace equipment, and the DC submerged arc furnace power supply device. The DC submerged arc furnace equipment is installed inside the DC submerged arc furnace main body, and the DC submerged arc furnace equipment is respectively communicatively connected to the DC submerged arc furnace power supply device and the DC submerged arc furnace control device; The DC submerged arc furnace control device includes: A data processing unit that acquires the operating condition data of the DC submerged arc furnace, where the operating condition data of the DC submerged arc furnace includes the production efficiency data of the DC submerged arc furnace, the control parameter data of the DC submerged arc furnace, and the power supply data of the DC submerged arc furnace, and preprocesses the operating condition data of the DC submerged arc furnace to obtain the preprocessed operating condition data of the DC submerged arc furnace; An analysis data unit that constructs a DC submerged arc furnace feature data set based on the preprocessed operating condition data of the DC submerged arc furnace, and uses the DC submerged arc furnace feature data set to train a random forest model to obtain a DC submerged arc furnace dynamic regulation model; A data comparison unit, in response to a DC submerged arc furnace control instruction, generates DC submerged arc furnace dynamic adjustment parameters through a DC submerged arc furnace dynamic regulation model. The DC submerged arc furnace dynamic adjustment parameters include power supply adjustment parameters of the power supply equipment and equipment adjustment parameters of the DC submerged arc furnace. A data optimization unit generates the simulated operation efficiency of the DC submerged arc furnace through a preset DC submerged arc furnace production simulation model, detects the dynamic adjustment state of the DC submerged arc furnace, optimizes the DC submerged arc furnace dynamic adjustment parameters according to the dynamic adjustment state of the DC submerged arc furnace, and uses the optimized DC submerged arc furnace dynamic adjustment parameters as execution parameters. The data analysis unit includes: A data classification sub-unit for classifying the preprocessed DC submerged arc furnace operation condition data according to preset dimensions. The preset dimensions include time dimension classification, equipment type classification, and process stage classification. A model training sub-unit constructs a DC submerged arc furnace feature data set based on the classified DC submerged arc furnace operation condition data. The DC submerged arc furnace feature data set contains the mapping relationship between equipment condition parameters and production efficiency parameters, and iteratively trains a random forest model through the mapping relationship. A model verification sub-unit uses the cross-validation method to verify the accuracy of the trained random forest model, and triggers a model retraining mechanism when the verification accuracy rate is lower than a preset threshold. The model retraining mechanism includes increasing the training sample size, adjusting the feature weight distribution, and optimizing the model hyperparameters. The data optimization unit includes: A data generation unit collects the operation efficiency of the DC submerged arc furnace after executing the DC submerged arc furnace dynamic adjustment parameters, and substitutes the DC submerged arc furnace control instruction into a preset DC submerged arc furnace production simulation model to obtain the simulated operation efficiency of the DC submerged arc furnace. A state detection unit compares the simulated operation efficiency of the DC submerged arc furnace with the operation efficiency of the DC submerged arc furnace to obtain the operation efficiency error of the DC submerged arc furnace. A parameter optimization unit optimizes the DC submerged arc furnace dynamic adjustment parameters using a genetic algorithm based on the operation efficiency error of the DC submerged arc furnace to obtain the optimized DC submerged arc furnace dynamic adjustment parameters. The optimized DC submerged arc furnace dynamic adjustment parameters include optimized power supply adjustment parameters of the power supply equipment and equipment adjustment parameters of the DC submerged arc furnace. A data execution unit sends the equipment adjustment parameters of the DC submerged arc furnace to the DC submerged arc furnace equipment to adjust the operation parameters of the DC submerged arc furnace equipment, and sends the power supply adjustment parameters of the power supply equipment to the power supply device of the DC submerged arc furnace to adjust the power supply parameters of the power supply device of the DC submerged arc furnace.

5. The DC submerged arc furnace control system according to claim 4, characterized in that, The data comparison unit includes: An instruction parsing unit for parsing the instruction type of the DC submerged arc furnace control instruction. The instruction type includes a real-time adjustment instruction and a timing adjustment instruction. A dynamic parameter generation unit calls a DC submerged arc furnace dynamic regulation model according to the instruction type to generate DC submerged arc furnace dynamic adjustment parameters matching the real-time working conditions. A parameter comparison unit compares the generated DC submerged arc furnace dynamic adjustment parameters with the historical adjustment parameter database and calculates the parameter deviation value. An adjustment strategy generation unit generates a parameter correction strategy when the parameter deviation value exceeds a preset safety range. The parameter correction strategy includes: In the case where the instruction type is a real-time adjustment instruction, a progressive parameter adjustment method is adopted; In the case where the instruction type is a timed adjustment instruction, a batch parameter optimization method is adopted.

6. The DC submerged arc furnace control system according to claim 4, wherein, The data optimization unit further includes: A dynamic threshold adjustment module that dynamically adjusts the convergence threshold of the optimization algorithm according to the deviation rate between the simulated operation efficiency and the measured operation efficiency of the DC submerged arc furnace. The convergence threshold adjustment rules include: When the deviation rate > 10%, a first-level strict convergence criterion is adopted; When 5% < deviation rate ≤ 10%, a second-level conventional convergence criterion is adopted; When the deviation rate ≤ 5%, a third-level loose convergence criterion is adopted; A multi-objective optimization module that jointly optimizes the dynamic adjustment parameters of the DC submerged arc furnace through constructing a comprehensive evaluation function including equipment energy consumption, production efficiency, and equipment loss rate; a parameter constraint module that imposes at least one of the following constraint conditions during the optimization process: The adjustment range of the power supply adjustment parameter of the power supply equipment does not exceed ±15% of the rated value; The adjustment frequency of the adjustment parameter of the DC submerged arc furnace equipment is limited to ≤ 3 times per minute; The time interval between two adjacent parameter adjustments ≥ 10 seconds.

Images