Submerged arc furnace electrode length measuring method and system

Abstract

The invention provides a submerged arc furnace electrode length measuring method and system, and belongs to the technical field of industrial automation. The method comprises the following steps: acquiring real-time production data; the real-time production data are preprocessed; inputting the preprocessed real-time production data into the trained electrode length prediction model to obtain an electrode length prediction result; the electrode length prediction model is obtained by training a bidirectional gating recurrent neural network based on an attention mechanism. A bidirectional gating recurrent neural network based on an attention mechanism is adopted to train an electrode length prediction model, the obtained electrode length prediction model considers a bidirectional propagation state, a deep learning algorithm is introduced to carry out learning and analysis according to production data, accurate and real-time prediction of the electrode length is realized, and the production efficiency is improved. The problems of low precision, complex operation, poor real-time performance and the like of a traditional measurement method are solved.

Description

Technical Field

[0001] The present application relates to the field of industrial automation technology, and in particular to a method for measuring the electrode length of an electric arc furnace, and a system for measuring the electrode length of an electric arc furnace. Background Art

[0002] Submerged arc furnace is an important smelting equipment for silicon, manganese, chromium, nickel iron and other ferroalloys, calcium carbide and yellow phosphorus. During the operation of the submerged arc furnace, the electrode length is a key factor affecting the stability of the furnace and production efficiency. However, due to the harsh working environment of the submerged arc furnace and the continuous consumption of electrodes, it has always been a technical problem to obtain the electrode length in real time during operation. The main measurement method currently used is to obtain the electrode length by manual measurement during the shutdown period, but the electrode length cannot be known during the operation of the submerged arc furnace, and therefore the electrode length cannot be accurately controlled, which affects the stability of the furnace and reduces production efficiency. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a method and system for measuring the electrode length of an electric arc furnace, so as to at least solve the above-mentioned problem that the electrode length cannot be measured accurately and in real time.

[0004] In order to achieve the above-mentioned object, the first aspect of the present application provides a method for measuring the length of an electrode of an electric arc furnace, and the method for measuring the length of an electric arc furnace comprises:

[0005] Get real-time production data;

[0006] Preprocessing of real-time production data;

[0007] The pre-processed real-time production data is input into the trained electrode length prediction model to obtain the electrode length prediction result;

[0008] The electrode length prediction model is obtained by training a bidirectional gated recurrent neural network based on an attention mechanism.

[0009] According to the above technical means, a bidirectional gated recurrent neural network based on the attention mechanism is used to train the electrode length prediction model. The obtained electrode length prediction model takes into account the bidirectional propagation state, and introduces a deep learning algorithm to learn and analyze according to the production data to achieve accurate and real-time prediction of the electrode length, overcoming the problems of low precision, complex operation and poor real-time performance of traditional measurement methods.

[0010] In an embodiment of the present application, preprocessing the real-time production data includes:

[0011] Abnormal data in real-time production data is eliminated, and missing data is supplemented by linear interpolation.

[0012] According to the above technical means, more accurate production data can be obtained to reduce the error of subsequent electrode length prediction and improve the prediction accuracy.

[0013] In an embodiment of the present application, the bidirectional gated recurrent neural network based on the attention mechanism includes: an input layer, an encoder layer, an attention layer and a prediction network layer; the encoder layer is used to map the input data sequence to the hidden state of the gated recurrent neural network; the attention layer is used to differentially fuse the input data sequence with the hidden state of the gated recurrent neural network to obtain the output of the attention layer; the prediction network layer is used to use the output of the attention layer as input to calculate the predicted value output.

[0014] In an embodiment of the present application, the encoder layer and the prediction network layer both include multiple bidirectional gated recurrent neural network units, which superimpose a first hidden state based on a historical data sequence of forward propagation and a second hidden state based on a future data sequence of backward propagation to obtain a hidden state of the current sequence.

[0015] According to the above technical means, the bidirectional gated recurrent neural network unit can simultaneously obtain the hidden state based on the historical data sequence and the hidden state based on the future data sequence, thereby improving the comprehensiveness of the output results of the network unit.

[0016] In an embodiment of the present application, the attention layer includes a similarity score calculation module, a normalization module and a weighted summation module. The similarity score calculation module is used to calculate the similarity between the input data sequence and the hidden state of the gated recurrent neural network; the normalization module is used to perform normalized calculation on the similarity; the weighted summation module is used to perform weighted summation of the result of the normalization calculation and the input data sequence to obtain the output of the attention layer.

[0017] In the embodiment of the present application, the electrode length prediction model is trained in the following manner:

[0018] Obtain historical production data and corresponding electrode length data;

[0019] Preprocess the historical production data and the corresponding electrode length data;

[0020] The pre-processed historical production data and the corresponding electrode length data are made into an electrode data set;

[0021] The electrode dataset was used to train a bidirectional gated recurrent neural network based on the attention mechanism to obtain the electrode length prediction model.

[0022] In an embodiment of the present application, the bidirectional gated recurrent neural network based on the attention mechanism includes an input layer, an encoder layer, an attention layer and a prediction network layer;

[0023] The electrode data set is used to train a bidirectional gated recurrent neural network based on an attention mechanism to obtain an electrode length prediction model, including:

[0024] Input the electrode dataset into the input layer;

[0025] Based on the encoder layer, the electrode dataset transmitted by the input layer is obtained, and the electrode dataset is mapped into a hidden state of the gated recurrent neural network;

[0026] Based on the attention layer, the electrode dataset is differentially fused with the hidden state of the gated recurrent neural network to obtain the output of the attention layer;

[0027] The prediction network layer calculates the predicted value output based on the output of the attention layer;

[0028] Adjusting the parameters of the attention-based bidirectional gated recurrent neural network according to the predicted value output and the corresponding electrode length;

[0029] The electrode length prediction model is obtained by iterating a preset number of times or the loss function reaches a preset condition.

[0030] In an embodiment of the present application, adjusting the parameters of the bidirectional gated recurrent neural network based on the attention mechanism according to the predicted value output and the corresponding electrode length includes:

[0031] Calculate the mean square error based on the predicted value output and the corresponding electrode length;

[0032] The parameters of the attention mechanism-based bidirectional gated recurrent neural network are adjusted according to the mean square error.

[0033] In an embodiment of the present application, both the real-time production data and the historical production data include at least: primary current, secondary current, primary voltage, active power, reactive power, apparent power, power factor, bottom ring temperature, and current-pressure ratio.

[0034] A second aspect of the present application provides a device for measuring the length of an electrode of an electric arc furnace, the device comprising:

[0035] A data acquisition unit, used to acquire real-time production data;

[0036] A data preprocessing unit, used for preprocessing real-time production data;

[0037] The electrode length measurement unit is used to input the pre-processed real-time production data into the trained electrode length prediction model to obtain the electrode length prediction result;

[0038] The electrode length prediction model is obtained by training a bidirectional gated recurrent neural network based on an attention mechanism.

[0039] According to the above-mentioned technical means, the device uses a bidirectional gated recurrent neural network based on the attention mechanism to train the electrode length prediction model. The obtained electrode length prediction model takes into account the bidirectional propagation state, and introduces a deep learning algorithm to learn and analyze according to the production data to achieve accurate and real-time prediction of the electrode length, overcoming the problems of low precision, complex operation and poor real-time performance of traditional measurement methods.

[0040] A third aspect of the present application provides a system for measuring the length of an electrode in a submerged arc furnace, the system comprising:

[0041] Data acquisition device, used to collect real-time production data of the ore-fired furnace;

[0042] A data acquisition and processing module is used to obtain real-time production data and perform preprocessing, and input the preprocessed real-time production data into a trained electrode length prediction model to obtain an electrode length prediction result. The electrode length prediction model is trained using a bidirectional gated recurrent neural network based on an attention mechanism;

[0043] The display module is used to display the electrode length prediction results.

[0044] The system introduces a deep learning algorithm to learn and analyze based on production data, achieving accurate and real-time prediction of electrode length, overcoming the problems of low precision, complex operation, and poor real-time performance of traditional measurement methods. The system can be applied to various types of submerged arc furnace equipment, including but not limited to AC submerged arc furnaces and DC submerged arc furnaces, and can be widely used in the production of submerged arc furnaces in the steel, chemical and other industries.

[0045] A fourth aspect of the present application provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the method for measuring the length of an electric arc furnace electrode.

[0046] Through the above technical solution, the provided method for measuring the electrode length of an electric arc furnace adopts a bidirectional gated recurrent neural network based on an attention mechanism to train an electrode length prediction model. The obtained electrode length prediction model takes into account the bidirectional propagation state, and introduces a deep learning algorithm to learn and analyze according to production data, so as to achieve accurate and real-time prediction of the electrode length, overcoming the problems of low precision, complex operation, and poor real-time performance of traditional measurement methods.

[0047] Through accurate electrode length prediction results, the real-time monitoring and adjustment of the furnace condition of the submerged arc furnace can be realized, thereby improving the stability of the furnace condition, reducing submerged arc furnace failures, further reducing production costs, and improving production efficiency. It can also reduce errors and accidents caused by manual operation in manual measurement and improve production safety.

[0048] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings:

[0050] Figure 1 This is a flow chart of a method for measuring the length of an electrode in a submerged arc furnace provided in one embodiment of the present application;

[0051] Figure 2 This is a schematic diagram of a bidirectional gated recurrent neural network structure based on an attention mechanism provided in one embodiment of the present application;

[0052] Figure 3 This is a schematic diagram of the structure of a bidirectional gated recurrent neural network unit provided in one embodiment of the present application;

[0053] Figure 4 This is a block diagram of a submerged arc furnace electrode length measuring device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0054] The specific implementation of the present application is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present application, and is not used to limit the present application.

[0055] Figure 1 This is a flow chart of a method for measuring the length of an electrode in an electric arc furnace provided in one embodiment of the present application. Figure 1 As shown in the figure, the method for measuring the electrode length of the electric arc furnace includes:

[0056] S1: Acquire real-time production data. In the embodiment of the present application, the real-time production data is collected by data acquisition devices such as sensors and data acquisition cards installed at corresponding positions on the ore-fired furnace and uploaded and stored in a database. The method of the present application can acquire real-time production data from the database. The real-time production data includes at least: primary current, secondary current, primary voltage, active power, reactive power, apparent power, power factor, bottom ring temperature, and flow-pressure ratio.

[0057] Among them, the primary current refers to the current on the primary side (high voltage side) of the transformer of the submerged arc furnace. It is the current connected to the transformer from the high voltage grid. The transformer converts the high voltage electric energy on the primary side into the low voltage and high current on the secondary side to provide energy for the smelting process of the submerged arc furnace. The primary current is usually small, and the size of the primary current is mainly determined by the equipment capacity and the load of the submerged arc furnace.

[0058] Secondary current refers to the current on the secondary side (low-voltage side) of the transformer of the submerged arc furnace. It is the output current of the low-voltage side of the transformer. The secondary current is transmitted to the charge through the electrode. As the voltage decreases, the current increases, providing strong heat for smelting. The secondary current is usually a large current of several thousand to tens of thousands of amperes, such as 20kA to 100kA. The secondary current is closely related to the conductivity of the charge, the electrode insertion depth and the electrode consumption.

[0059] Primary voltage is the input voltage of the electric arc furnace transformer connected to the power grid. It is usually determined by the level of the regional power grid, such as 6kV, 10kV, 35kV. The fluctuation of primary voltage will affect the output voltage of the transformer, thereby affecting the stability of the smelting process.

[0060] Active power is the part that is converted into heat energy in the furnace. It is the main energy input of the ore-heat furnace and the effective power part that is converted into heat and used to melt the charge in the ore-heat furnace. Active power determines the speed of the chemical reaction in the furnace and the efficiency of melting the charge. Too high or too low active power will lead to abnormal temperature in the furnace or excessive energy consumption.

[0061] Reactive power is the power generated by the inductance and capacitance in the ore-fired furnace. It cannot be used directly to do work, but it affects the operation of the power grid and equipment. Insufficient reactive power will cause voltage fluctuations.

[0062] Apparent power is the product of voltage and current in the circuit, and is the total power provided by the power supply to the ore-fired furnace.

[0063] The power factor is the ratio of active power to apparent power, reflecting the efficiency of electric energy utilization.

[0064] The bottom ring temperature refers to the temperature of the annular area at the bottom of the submerged arc furnace, which reflects the thermal state of the furnace bottom structure and cooling system and is used to monitor the safety of furnace bottom operation and heat distribution.

[0065] The current-to-voltage ratio is the ratio of current to voltage, reflecting the electrode insertion depth and arc status. High current-to-voltage ratio: may indicate a short circuit or the electrode is too deep. Low current-to-voltage ratio: may indicate that the electrode is too high or the charge resistance is too large.

[0066] S2: Preprocess the real-time production data.

[0067] In the embodiment of the present application, the real-time production data is preprocessed, including:

[0068] Abnormal data in the real-time production data is eliminated, and missing data is supplemented by linear interpolation. In some embodiments, the abnormal data in the real-time production data includes data with negative values.

[0069] By eliminating and completing abnormal data in real-time production data, more accurate and complete production data can be obtained to reduce the error of subsequent electrode length prediction and improve prediction accuracy.

[0070] S3: Input the preprocessed real-time production data into the trained electrode length prediction model to obtain an electrode length prediction result. The electrode length prediction model is trained using a bidirectional gated recurrent neural network based on an attention mechanism.

[0071] According to the above technical means, a bidirectional gated recurrent neural network based on the attention mechanism is used to train the electrode length prediction model. The obtained electrode length prediction model takes into account the bidirectional propagation state, and introduces a deep learning algorithm to learn and analyze according to the production data to achieve accurate and real-time prediction of the electrode length, overcoming the problems of low precision, complex operation and poor real-time performance of traditional measurement methods.

[0072] In the embodiments of the present application, Figure 2 As shown, the bidirectional gated recurrent neural network based on the attention mechanism includes: an input layer, an encoder layer, an attention layer and a prediction network layer; the encoder layer is used to map the input data sequence to the hidden state of the gated recurrent neural network; the attention layer is used to differentially fuse the input data sequence with the hidden state of the gated recurrent neural network to obtain the output of the attention layer; the prediction network layer is used to calculate the output of the attention layer as input to obtain the predicted value output. The input layer is used to input the original input time series.

[0073] In the embodiments of the present application, Figure 3 As shown, the encoder layer and the prediction network layer both include multiple bidirectional gated recurrent neural network units, which superimpose a first hidden state based on a historical data sequence in forward propagation and a second hidden state based on a future data sequence in backward propagation to obtain a hidden state of the current sequence.

[0074] According to the above technical means, the bidirectional gated recurrent neural network unit can simultaneously obtain the hidden state based on the historical data sequence and the hidden state based on the future data sequence, thereby improving the comprehensiveness of the output results of the network unit.

[0075] In the embodiments of the present application, Figure 2As shown, the attention layer includes a similarity score calculation module, a normalization module and a weighted summation module. The similarity score calculation module is used to calculate the similarity between the input data sequence and the hidden state of the gated recurrent neural network; the normalization module is used to perform normalized calculation on the similarity; the weighted summation module is used to perform weighted summation of the result of the normalized calculation and the input data sequence to obtain the output of the attention layer. In an embodiment of the present application, the similarity score calculation module uses the hyperbolic tangent function tanh to calculate the similarity between the input data sequence and the hidden state of the gated recurrent neural network; the normalization module uses the normalized exponential function Softmax for normalization calculation.

[0076] In the embodiment of the present application, the electrode length prediction model is trained in the following manner:

[0077] 1) Obtain historical production data and corresponding electrode length data. In the embodiment of the present application, historical production data is collected during the historical production process by data acquisition devices such as sensors and data acquisition cards installed at corresponding positions on the ore-fired furnace and uploaded and stored in the database, and the electrode length data manually measured multiple times during the historical production process are recorded. During the training of the electrode length prediction model, historical production data and corresponding electrode length data are obtained from the database.

[0078] 2) Preprocessing the historical production data and the corresponding electrode length data. In the embodiment of the present application, the abnormal data in the historical production data is also first eliminated, and the missing data is supplemented by linear interpolation. In some embodiments, the abnormal data in the historical production data includes data with negative values, etc.

[0079] By removing and completing abnormal data in historical production data, we can obtain more accurate and complete historical production data, thereby improving the prediction accuracy of the trained model.

[0080] 3) The pre-processed historical production data and the corresponding electrode length data are made into an electrode data set. In some embodiments, different electrode data sets are made according to different electrodes. For example, for a submerged arc furnace using three-phase electrodes, the data corresponding to each electrode needs to be made into a three-phase electrode data set. In some embodiments, 80% of the data in the electrode data set is used as a training data set, 10% of the data is used as a validation set, and 10% of the data is used as a test set.

[0081] 4) The electrode dataset is used to train a bidirectional gated recurrent neural network based on the attention mechanism to obtain an electrode length prediction model.

[0082] First, a bidirectional gated recurrent neural network (Attention-GRU) model based on the attention mechanism is built, including: input layer, encoder layer, attention layer and prediction network layer.

[0083] Then, the electrode dataset is used to train a bidirectional gated recurrent neural network based on the attention mechanism to obtain an electrode length prediction model, which includes:

[0084] 4.1 Input the electrode dataset into the input layer.

[0085] The input layer is used to input the electrode dataset. The electrode dataset includes the original input time series. The original input time series is multi-attribute network data with n features. The input sample data is:

[0086] X=[x 1 ,x 2 ,…,x n ] T =[x1,x2,…,x T ]∈R n×T

[0087] Where T is the time step, and each training sample is an n×1 matrix.

[0088] 4.2 Based on the encoder layer, the electrode dataset transmitted by the input layer is obtained, and the electrode dataset is mapped to the hidden state of the gated recurrent neural network.

[0089] Encoder layer: The bidirectional gated recurrent neural network unit GRU is selected as the neural network processing unit of the encoder layer. The data training sample is obtained from the input layer and input into the encoder GRU. The input sample data sequence X is mapped to the hidden state h of the GRU through the GRU unit. t , the formula is as follows:

[0090] r t =σ(W r [h t-1 ,x t ]+b r )

[0091] z t =σ(W z [h t-1 ,x t ]+b z )

[0092]

[0093] In the formula, r t is the reset gate, z tis the update gate, which controls how to forget and save information respectively. W and b represent the weight matrix and bias value respectively. Indicates the dot product relationship. In addition, the output hidden state value h t Provide data support for encoding the attention layer.

[0094] 4.3 Based on the attention layer, the electrode dataset is differentially fused with the hidden state of the gated recurrent neural network to obtain the output of the attention layer.

[0095] Attention layer: The training sample sequence data of the input layer and the output of GRU are used as the input of the attention layer, and the training sample sequence data of each time period are differentially fused.

[0096] 4.4 Based on the prediction network layer, the predicted value output is calculated according to the output of the attention layer.

[0097] Prediction network layer: The output of the attention layer is used as the input of the prediction network layer to calculate the predicted value output.

[0098] 4.5 Adjusting the parameters of the bidirectional gated recurrent neural network based on the attention mechanism according to the predicted value output and the corresponding electrode length. In the embodiment of the present application, the mean square error is calculated according to the predicted value output and the corresponding electrode length.

[0099]

[0100] Where n is the number of samples, y i is the corresponding electrode length, Output is the predicted value.

[0101] Then, the parameters of the bidirectional gated recurrent neural network based on the attention mechanism are adjusted according to the mean square error. The smaller the mean square error value, the smaller the difference between the predicted value output and the corresponding electrode length, and the higher the accuracy of the model prediction.

[0102] 4.6 Iterate a preset number of times or the loss function reaches a preset condition to obtain an electrode length prediction model.

[0103] During the training process, the parameters of the network model are optimized and iterated according to the predicted values ​​and manual measurement values, and the prediction accuracy of the model is evaluated using the validation set. Finally, the electrode length prediction model is obtained and saved.

[0104] The second aspect of the present application provides a device for measuring the length of an electrode of an electric arc furnace. Figure 4 As shown, the electrode length measuring device of the ore-fired furnace comprises:

[0105] A data acquisition unit, used to acquire real-time production data;

[0106] A data preprocessing unit, used for preprocessing real-time production data;

[0107] The electrode length measurement unit is used to input the pre-processed real-time production data into the trained electrode length prediction model to obtain the electrode length prediction result;

[0108] The electrode length prediction model is obtained by training a bidirectional gated recurrent neural network based on an attention mechanism.

[0109] According to the above-mentioned technical means, the device uses a bidirectional gated recurrent neural network based on the attention mechanism to train the electrode length prediction model. The obtained electrode length prediction model takes into account the bidirectional propagation state, and introduces a deep learning algorithm to learn and analyze according to the production data to achieve accurate and real-time prediction of the electrode length, overcoming the problems of low precision, complex operation and poor real-time performance of traditional measurement methods.

[0110] A third aspect of the present application provides a system for measuring the length of an electrode in a submerged arc furnace, the system comprising:

[0111] The data acquisition device is used to collect real-time production data of the ore-fired furnace. In the embodiment of the present application, the data acquisition device can be a sensor, a data acquisition card, etc.

[0112] A data acquisition and processing module is used to obtain real-time production data and perform preprocessing, and input the preprocessed real-time production data into a trained electrode length prediction model to obtain an electrode length prediction result. The electrode length prediction model is trained using a bidirectional gated recurrent neural network based on an attention mechanism;

[0113] The display module is used to display the electrode length prediction results.

[0114] In some feasible embodiments, the submerged arc furnace electrode length measurement system further includes a safety protection module, which monitors whether the real-time production data of the submerged arc furnace is normal, thereby ensuring the safe operation of the system.

[0115] The system introduces a deep learning algorithm to learn and analyze production data, and realizes accurate and real-time prediction of electrode length, overcoming the problems of low precision, complex operation, and poor real-time performance of traditional measurement methods. The system can be applied to various types of submerged arc furnace equipment, including but not limited to AC submerged arc furnaces and DC submerged arc furnaces, and can be widely used in the production of submerged arc furnaces in the steel, chemical and other industries. The predicted electrode length is provided to the operators in the central control room to adjust the operating parameters of the submerged arc furnace, such as power factor, electrode lifting, etc., to optimize the furnace condition and improve production efficiency.

[0116] A fourth aspect of the present application provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the method for measuring the length of an electric arc furnace electrode.

[0117] Those skilled in the art will understand that all or part of the steps in the method for implementing the above-mentioned embodiments can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including several instructions for making a single-chip microcomputer, a chip or a processor (processor) execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0118] The optional implementation modes of the present application are described in detail above in conjunction with the accompanying drawings. However, the implementation modes of the present application are not limited to the specific details in the above implementation modes. Within the technical concept of the implementation modes of the present application, the technical scheme of the implementation modes of the present application can be subjected to a variety of simple modifications, and these simple modifications all belong to the protection scope of the implementation modes of the present application. It should also be noted that the various specific technical features described in the above specific implementation modes can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the implementation modes of the present application will not be described separately for various possible combinations.

[0119] In addition, the various implementation modes of the present application may be arbitrarily combined, and as long as they do not violate the concept of the implementation modes of the present application, they should also be regarded as the contents disclosed in the implementation modes of the present application.

Claims

1. A method for measuring the length of an electrode of an electric arc furnace, characterized in that: The method for measuring the length of an ore-fired furnace electrode comprises: Get real-time production data; Preprocessing of real-time production data; The pre-processed real-time production data is input into the trained electrode length prediction model to obtain the electrode length prediction result; The electrode length prediction model is obtained by training a bidirectional gated recurrent neural network based on an attention mechanism.

2. The method for measuring the electrode length of a submerged arc furnace according to claim 1, characterized in that: Preprocessing of real-time production data, including: Abnormal data in real-time production data is eliminated, and missing data is supplemented by linear interpolation.

3. The method for measuring the electrode length of a submerged arc furnace according to claim 1, characterized in that: The bidirectional gated recurrent neural network based on the attention mechanism includes: an input layer, an encoder layer, an attention layer and a prediction network layer; the encoder layer is used to map the input data sequence to the hidden state of the gated recurrent neural network; the attention layer is used to differentially fuse the input data sequence with the hidden state of the gated recurrent neural network to obtain the output of the attention layer; the prediction network layer is used to use the output of the attention layer as input to calculate the predicted value output.

4. The method for measuring the electrode length of a submerged arc furnace according to claim 3, characterized in that: The encoder layer and the prediction network layer both include a plurality of bidirectional gated recurrent neural network units, which superimpose a first hidden state based on a historical data sequence of forward propagation and a second hidden state based on a future data sequence of backward propagation to obtain a hidden state of a current sequence.

5. The method for measuring the electrode length of a submerged arc furnace according to claim 3, characterized in that: The attention layer includes a similarity score calculation module, a normalization module and a weighted summation module, wherein the similarity score calculation module is used to calculate the similarity between the input data sequence and the hidden state of the gated recurrent neural network; The normalization module is used to perform normalization calculation on the similarity; the weighted summation module is used to perform weighted summation of the result of the normalization calculation and the input data sequence to obtain the output of the attention layer.

6. The method for measuring the electrode length of a submerged arc furnace according to claim 1, characterized in that: The electrode length prediction model is trained in the following way: Obtain historical production data and corresponding electrode length data; Preprocess the historical production data and the corresponding electrode length data; The pre-processed historical production data and the corresponding electrode length data are made into an electrode data set; The electrode dataset was used to train a bidirectional gated recurrent neural network based on the attention mechanism to obtain the electrode length prediction model.

7. The method for measuring the electrode length of a submerged arc furnace according to claim 6, characterized in that: The bidirectional gated recurrent neural network based on the attention mechanism includes an input layer, an encoder layer, an attention layer and a prediction network layer; The electrode data set is used to train a bidirectional gated recurrent neural network based on an attention mechanism to obtain an electrode length prediction model, including: Input the electrode dataset into the input layer; Based on the encoder layer, the electrode dataset transmitted by the input layer is obtained, and the electrode dataset is mapped into a hidden state of the gated recurrent neural network; Based on the attention layer, the electrode dataset is differentially fused with the hidden state of the gated recurrent neural network to obtain the output of the attention layer; The prediction network layer calculates the predicted value output based on the output of the attention layer; Adjusting the parameters of the bidirectional gated recurrent neural network based on the attention mechanism according to the predicted value output and the corresponding electrode length; The electrode length prediction model is obtained by iterating a preset number of times or the loss function reaches a preset condition.

8. The method for measuring the length of an electric arc furnace electrode according to claim 7, characterized in that: Adjusting the parameters of the bidirectional gated recurrent neural network based on the attention mechanism according to the predicted value output and the corresponding electrode length includes: Calculate the mean square error based on the predicted value output and the corresponding electrode length; The parameters of the attention mechanism-based bidirectional gated recurrent neural network are adjusted according to the mean square error.

9. The method for measuring the electrode length of a submerged arc furnace according to claim 1, characterized in that: Both real-time production data and historical production data include at least: primary current, secondary current, primary voltage, active power, reactive power, apparent power, power factor, bottom ring temperature and current-pressure ratio.

10. A submerged arc furnace electrode length measurement system, characterized in that: The electrode length measurement system of the ore-fired furnace comprises: Data acquisition device, used to collect real-time production data of the ore-fired furnace; A data acquisition and processing module is used to obtain real-time production data and perform preprocessing, and input the preprocessed real-time production data into a trained electrode length prediction model to obtain an electrode length prediction result. The electrode length prediction model is trained using a bidirectional gated recurrent neural network based on an attention mechanism; The display module is used to display the electrode length prediction results.

Images