Aluminum electrolysis cell temperature field prediction method and system based on improved CNN-LSTM

Abstract

The invention discloses an aluminum electrolysis cell temperature field prediction method and system based on improved CNN-LSTM, and relates to the technical field of aluminum electrolysis industry, and the method comprises the steps: building a multi-physical field coupling model of an aluminum electrolysis cell based on the operation performance parameters of the aluminum electrolysis cell; the multi-physics field coupling model comprises an electric field mathematical model, an electric contact mathematical model in an electric field and a temperature field mathematical model; calculating temperature field data of the aluminum electrolysis cell under different current data based on boundary conditions during operation of the aluminum electrolysis cell and a multi-physics field coupling model; carrying out dimension reduction processing on the temperature field data, and extracting an order reduction mode of the temperature field data; a CNN-LSTM model is constructed and trained based on different current data and the corresponding order reduction modes, and a trained CNN-LSTM model is obtained; and predicting the temperature field of the aluminum electrolysis cell based on the trained CNN-LSTM model. The technical problems of high calculation complexity and low accuracy in the prior art are relieved.

Description

Technical Field

[0001] The present invention relates to the technical field of aluminum electrolysis industry, and particularly to a method and system for predicting the temperature field of an aluminum electrolysis cell based on an improved CNN-LSTM. Background Art

[0002] The aluminum electrolysis cell plays a central role in the aluminum production process, and its operating efficiency and cost control are restricted by the heat management efficiency. Given the extreme complexity and harshness of the internal environment of the aluminum electrolysis cell, directly measuring its internal temperature distribution is not only extremely challenging technically and costly, but may also pose a threat to production safety. Traditional monitoring methods, such as manual measurement or relying on finite element analysis, all have limitations that cannot be ignored. Manual measurement is not only time-consuming and laborious, but also easily interfered by human factors, making it difficult to ensure the accuracy and consistency of data; while finite element analysis can simulate the temperature field distribution of the electrolysis cell, but its calculation process is cumbersome, has high requirements for computing resources, and is difficult to instantly capture the dynamic changes of the electrolysis cell, showing obvious timeliness deficiencies.

[0003] Chinese Patent Document CN114970320A discloses a method for dynamically monitoring the temperature field of an aluminum electrolysis cell. This method constructs a thermal field simulation model by collecting and cleaning the external thermal field data of the aluminum electrolysis cell in real time, combining structural, process parameters and the temperature of the melt in the cell, and at the same time, combines this model to reverse calculate the internal temperature field of the electrolysis cell, evaluate the thermal balance state, and conduct real-time monitoring. However, since this method requires collecting a large amount of data for cleaning and numerical simulation, it requires high computing resources and time costs, and focuses on real-time monitoring and reverse calculation, lacking the ability to predict the temperature field. Chinese Patent Document CN118429769A discloses a multi-modal contact performance detection method and system based on a weighted fusion and CNN-LSTM hybrid architecture. This method extracts key features and fuses them to form a comprehensive feature set by collecting and processing the visual and temperature data of the contact, and then uses the hybrid CNN-LSTM model for training and prediction, which can accurately predict the performance state of the contact. However, when CNN-LSTM faces high-dimensional data, not only will the amount of calculation increase, but the complexity of the model will also increase, which may lead to a problem of reduced prediction accuracy. Summary of the Invention

[0004] In order to solve the above technical problems existing in the prior art, the embodiments of the present invention provide a method and system for predicting the temperature field of an aluminum electrolysis cell based on an improved CNN-LSTM. The technical solutions are as follows:

[0005] On the one hand, a method for predicting the temperature field of an aluminum electrolysis cell based on an improved CNN-LSTM is provided. The method includes: constructing a multi-physical field coupling model of the aluminum electrolysis cell based on the operating performance parameters of the aluminum electrolysis cell; the multi-physical field coupling model includes an electric field mathematical model, an electric contact mathematical model in the electric field, and a temperature field mathematical model; calculating the temperature field data of the aluminum electrolysis cell under different current data based on the boundary conditions during the operation of the aluminum electrolysis cell and the multi-physical field coupling model; performing dimensionality reduction processing on the temperature field data to extract the reduced-order modes of the temperature field data; constructing and training a CNN-LSTM model based on different current data and the corresponding reduced-order modes to obtain a trained CNN-LSTM model; predicting the temperature field of the aluminum electrolysis cell based on the trained CNN-LSTM model.

[0006] Further, the electric field mathematical model includes:

[0007]

[0008] J = -σ▽V = σE

[0009] In the formula, ▽ is the Hamiltonian operator; σ is the conductivity, σ x , σ y , σ z are the components of the conductivity in the x, y, and z directions respectively; J is the current density; V is the electric potential; E is the electric field strength; the electric contact mathematical model in the electric field includes:

[0010]

[0011] In the formula, J is the current density passing through the contact surface; σ c is the contact conductivity; and are the electric potentials of the contact surface; the temperature field mathematical model includes:

[0012]

[0013] In the formula, k x , k y , k z are the thermal conductivities in the x, y, and z directions respectively; T is the temperature; q s is the heat source intensity.

[0014] Further, the boundary conditions include electric field boundary conditions and thermal field boundary conditions; among them, the electric field boundary conditions include: setting a current to flow in at the inlet section of the column busbar, setting the electric potential at the outlet section of the cathode busbar to zero, and setting the four-side boundary of the molten aluminum to an insulating condition; the thermal field boundary conditions include: the ambient temperature of the aluminum electrolysis cell, the convective loss, radiative loss, emissivity of the outer surface of the cell shell, the convective heat transfer mode and the radiative heat transfer mode; the convective loss includes:

[0015] Q c =a c (T1-T0)S

[0016] In the formula, Q c is the convective loss, a c is the convective heat transfer coefficient; T1 is the temperature of the outer surface of the cell shell; T2 is the ambient temperature; S is the heat dissipation area; the radiative loss includes:

[0017]

[0018] In the formula, Q r is the radiative loss, ε is the emissivity of the heat dissipation surface of the cell shell; κ is the Stefan-Boltzmann constant; is the mutual radiation angle coefficient between the radiative surface and the adjacent surface.

[0019] Further, perform dimensionality reduction processing on the temperature field data and extract the reduced-order mode of the temperature field data, including: performing normalization processing on the temperature field data to obtain normalized temperature field data; based on the proper orthogonal decomposition method, performing dimensionality reduction processing on the normalized temperature field data to obtain the reduced-order mode of the temperature field data.

[0020] Further, based on the trained CNN-LSTM model, predict the temperature field of the aluminum electrolysis cell, including: using the current data to be predicted of the aluminum electrolysis cell as the input data of the trained CNN-LSTM model to obtain the target reduced-order mode corresponding to the current data to be predicted; reconstructing the temperature field data of the aluminum electrolysis cell based on the target reduced-order mode to obtain predicted temperature field data.

[0021] On the other hand, a temperature field prediction system for an aluminum electrolysis cell based on an improved CNN-LSTM is also provided, including: a first construction module, a simulation module, a dimensionality reduction module, a second construction module, and a prediction module; wherein, the first construction module is used to construct a multi-physical field coupling model of the aluminum electrolysis cell based on the operating performance parameters of the aluminum electrolysis cell; the multi-physical field coupling model includes an electric field mathematical model, an electric contact mathematical model in the electric field, and a temperature field mathematical model; the simulation module is used to calculate the temperature field data of the aluminum electrolysis cell under different current data based on the boundary conditions during the operation of the aluminum electrolysis cell and the multi-physical field coupling model; the dimensionality reduction module is used to perform dimensionality reduction processing on the temperature field data and extract the reduced-order mode of the temperature field data; the second construction module is used to construct and train a CNN-LSTM model based on different current data and the corresponding reduced-order mode to obtain a trained CNN-LSTM model; the prediction module is used to predict the temperature field of the aluminum electrolysis cell based on the trained CNN-LSTM model.

[0022] Further, the dimensionality reduction module is further used to: perform normalization processing on the temperature field data to obtain normalized temperature field data; perform dimensionality reduction processing on the normalized temperature field data based on the proper orthogonal decomposition method to obtain the reduced-order mode of the temperature field data.

[0023] Further, the prediction module is further used to: use the to-be-predicted current data of the aluminum electrolysis cell as the input data of the trained CNN-LSTM model to obtain the target reduced-order mode corresponding to the to-be-predicted current data; reconstruct the temperature field data of the aluminum electrolysis cell based on the target reduced-order mode to obtain predicted temperature field data.

[0024] On the other hand, an electronic device is also provided, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the method provided by the embodiments of the present invention is implemented.

[0025] On the other hand, a computer-readable storage medium is also provided, and program code is stored in the computer-readable storage medium, and the program code can be called by a processor to execute the method provided by the embodiments of the present invention.

[0026] The embodiments of the present invention provide a method and system for predicting the temperature field of an aluminum electrolysis cell based on an improved CNN-LSTM. By performing dimensionality reduction processing on the temperature field data, the problem that high computational resources and time costs are required in the prior art when using a large amount of data and in a high-dimensional situation can be effectively overcome. By constructing a CNN-LSTM model, the problem that there are large local prediction errors in the separate LSTM prediction method is overcome, and fast and accurate prediction of the temperature field of the aluminum electrolysis cell is realized. Brief Description of the Drawings

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0028] Figure 1 is a flowchart of a method for predicting the temperature field of an aluminum electrolysis cell based on an improved CNN-LSTM provided by an embodiment of the present invention;

[0029] Figure 2 is a flowchart of a rapid prediction of the temperature field provided by an embodiment of the present invention;

[0030] Figure 3 is a schematic diagram of a system for predicting the temperature field of an aluminum electrolysis cell based on an improved CNN-LSTM provided by an embodiment of the present invention. Detailed Embodiments

[0031] The following will describe the technical solutions in the present invention with reference to the drawings.

[0032] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or designs. Exactly, the use of the word "example" is intended to present concepts in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0033] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the drawings and specific embodiments.

[0034] Embodiment 1

[0035] Figure 1 is a flowchart of a method for predicting the temperature field of an aluminum electrolysis cell based on an improved CNN-LSTM provided by an embodiment of the present invention. As Figure 1 shown, the method specifically includes the following steps:

[0036] Step S102, based on the operating performance parameters of the aluminum electrolysis cell, construct a multi-physical field coupling model of the aluminum electrolysis cell; the multi-physical field coupling model includes an electric field mathematical model, an electric contact mathematical model in the electric field, and a temperature field mathematical model.

[0037] First, obtain the operation performance parameters of the aluminum electrolytic cell. Specifically, the sensor connected during the operation of the electrolytic cell is used to obtain parameter data such as current and temperature, and the conductivity, contact area, thermal conductivity, emissivity, convective heat transfer coefficient and other performance parameters of each structure of the electrolytic cell are obtained through the electrolytic cell design drawings and relevant literature.

[0038] Step S104: Based on the boundary conditions and the multi-physics field coupling model during the operation of the aluminum electrolytic cell, calculate the temperature field data of the aluminum electrolytic cell under different current data.

[0039] Step S106: Perform dimensionality reduction processing on the temperature field data and extract the reduced-order mode of the temperature field data.

[0040] Step S108: Based on different current data and the corresponding reduced-order modes, construct and train a CNN-LSTM model to obtain a trained CNN-LSTM model.

[0041] Step S110: Predict the temperature field of the aluminum electrolytic cell based on the trained CNN-LSTM model.

[0042] Specifically, in the embodiment of the present invention, for the electric field problem of the aluminum electrolytic cell, the conductive busbar can be simplified and an equivalent model can be used, and then the Laplace equation and Ohm's law can be used to calculate the current and potential of the electrolytic cell. The electric field mathematical model includes:

[0043]

[0044] J = -σ▽V = σE

[0045] where ▽ is the Hamiltonian operator; σ is the conductivity, S·m -1 , σx, σy, and σz are the components of the conductivity in the x, y, and z directions respectively; J is the current density, A·m -2 ; V is the electric potential, V; E is the electric field strength, V·m -1 ;

[0046] The phenomenon of electrical contact is common in aluminum electrolytic cells. The contact voltage drop generated by electrical contact will affect the results of electrothermal field calculations. Therefore, electrical contact should also be considered in the electric field modeling analysis. Specifically, the electrical contact mathematical model in the electric field includes:

[0047]

[0048] where J is the current density passing through the contact surface, A·m -2 ; σ c is the contact conductivity, S·m -1 ; and are the electric potentials of the contact surface, V;

[0049] When calculating the temperature field, based on the continuous medium hypothesis and steady-state conditions, the heat conduction equation is introduced to quantify the heat exchange and temperature distribution between various parts. The mathematical model of the temperature field includes:

[0050]

[0051] In the formula, k x , k y , k z are the thermal conductivities in the x, y, and z directions, respectively, W / (m·K -1 ); T is the temperature, K; q s is the heat source intensity, W / m 3 .

[0052] Specifically, in the embodiments of the present invention, the boundary conditions include the electric field boundary conditions and the thermal field boundary conditions; among them,

[0053] The electric field boundary conditions include: setting a current to flow in at the inlet section of the column busbar, setting the electric potential at the outlet section of the cathode busbar to zero, and setting the boundaries around the molten aluminum to be insulated;

[0054] The thermal field boundary conditions include: the ambient temperature of the aluminum electrolysis cell, the convective loss, radiative loss, emissivity of the outer surface of the cell shell, the heat convection heat transfer mode, and the heat radiation heat transfer mode.

[0055] In an optional implementation manner provided by the embodiments of the present invention, the ambient temperature of the aluminum electrolysis cell is set to 40 °C.

[0056] The convective loss includes:

[0057] Q c = a c (T1 - T0)S

[0058] In the formula, Q c is the convective loss, a c is the convective heat transfer coefficient, J / (m 2 ·s·K); T1 is the temperature of the outer surface of the cell shell, K; T2 is the ambient temperature, K; S is the heat dissipation area, m 2 ;

[0059] The radiative loss includes:

[0060]

[0061] In the formula, Q r is the radiative loss, ε is the emissivity of the heat dissipation surface of the cell shell; κ is the Stefan-Boltzmann constant; is the mutual radiation angle coefficient between the radiation surface and the adjacent surface.

[0062] The cell shell of the aluminum electrolysis cell has a rough oxidized steel surface, with an emissivity of 0.8, the emissivity of the covering material is 0.4, and the emissivity of the aluminum busbar is 0.07. The view factor is 1.

[0063] The heat convection heat transfer mode is set as the heat flux model, and the heat radiation heat transfer mode is set as the surface-to-environment radiation model.

[0064] In an alternative embodiment provided by the embodiments of the present invention, after constructing the multi-physical field coupling model of the aluminum electrolysis cell, the method further includes: using the electrolysis cell temperature data obtained through actual measurement as a verification benchmark, and performing a detailed comparison and verification on the electrolysis cell temperature field results obtained through simulation calculation by the multi-physical field coupling model to ensure the accuracy and reliability of the multi-physical field coupling model. Through the comparison and analysis of the actual data and the simulation data, the present invention can evaluate the accuracy of the multi-physical field coupling model in calculating the temperature field of the aluminum electrolysis cell, and further optimize and adjust the multi-physical field coupling model according to the verification results to improve the prediction ability and practical application value of the simulation.

[0065] Specifically, step S106 further includes the following steps:

[0066] Step S1061, normalizing the temperature field data to obtain the normalized temperature field data.

[0067] The data preprocessing of the temperature field of the aluminum electrolysis cell mainly involves data cleaning, that is, identifying and correcting errors, outliers, and missing information in the original data to ensure the integrity and accuracy of the data; at the same time, it also includes data standardization. In order to convert the temperature data at different positions or time points into a unified standard, thereby eliminating the dimension difference and improving the analysis accuracy; in addition, data dimensionality reduction techniques, such as principal component analysis, are also applied to reduce the data dimension to simplify the analysis process of complex data sets. Therefore, these data need to be preprocessed by normalization.

[0068] In an alternative embodiment provided by the embodiments of the present invention, z-score normalization is used to preprocess the aluminum electrolysis temperature field data. For a sample sequence of a certain variable x1, x2, x3 ···, x i , ···, x n , the z-score normalization formula is:

[0069]

[0070]

[0071] where x i is the original data; μ is the mean of the original data; σ is the standard deviation of the original data; z iis the data after z-score normalization; after the original data is z-score normalized, the data follows a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0072] Step S1062, based on the proper orthogonal decomposition method, perform dimensionality reduction on the normalized temperature field data to obtain the reduced-order modes of the temperature field data.

[0073] The proper orthogonal decomposition (POD) method is specifically used to extract characteristic information from discrete data. It can decompose a random variable into the sum of a series of basis functions defined by its inherent characteristics. The selection of these basis functions follows a principle: in each step of decomposition, ensure that the lowest-order mode contains the maximum energy. Using the proper orthogonal decomposition technique, effective dimensionality reduction and feature extraction are performed on the simulation data of the temperature field.

[0074] For the temperature field calculation results of m aluminum electrolysis cell shells under different current boundary conditions, assume T = [T1 T2 T3 T4 ··· T m . The results obtained under different boundary conditions are the temperature values at all coordinate points of the model, so the obtained T is full-order model data. The temperature field snapshot matrix containing the results of m different conditions can be expressed as:

[0075] K = [T(t1), T(t2), T(t3), T(t4), ···, T(t m )] ∈ R n×m

[0076] Singular value decomposition (SVD) is a commonly used mathematical tool for implementing POD proper orthogonal decomposition. It efficiently decomposes the temperature field snapshot matrix into a set of singular values and the corresponding left and right singular vectors. The expression is:

[0077] K = UΣV T

[0078] where U and V are orthogonal matrices containing the left and right singular vectors respectively, and Σ is a diagonal matrix. The elements s1, s2, s3, ···, s m on the diagonal are the singular values of the snapshot matrix. The left singular vectors represent the basis vectors of the new coordinate system in the space after dimensionality reduction of the original data. The right singular vectors represent the coefficients of each mode changing with different boundary conditions. The singular values in the diagonal matrix are arranged in descending order. The larger the singular value, the more dispersed the data distribution in the corresponding singular vector direction, and the more important the information in this direction. Each basis vector carries part of the information in the full-order model, and the square of the singular value represents the energy of the corresponding basis vector. Therefore, the purpose of order reduction can be achieved by selecting the principle that the first q-order basis functions carry most of the information in the full-order model. The proportion of the information energy of the first q-order basis vectors can be obtained by the following formula:

[0079]

[0080] Usually, E(q)≥99.99% is taken, that is, the original full-order data can be reconstructed by the first q-order singular values and the corresponding basis vectors. The temperature values of the full thermal field can be reconstructed by the first q POD order reduction modes and the corresponding coefficients, as shown in the following formula.

[0081]

[0082] In the embodiment of the present invention, a CNN-LSTM model is constructed and trained. CNN is widely used in many fields such as image processing and natural language processing due to its powerful feature extraction ability. The CNN module is used to process the input operating current data of the aluminum electrolysis cell. These data are regarded as a special one-dimensional time series data, which contains the operating state information of the electrolysis cell such as temperature. In the CNN module, multiple convolutional layers are designed to gradually extract the local features in the current data. These features represent the current change patterns of the electrolysis cell at different time periods, or the potential relationships between the current and other operating parameters of the electrolysis cell. At the same time, in order to enhance the generalization ability of the model and reduce the risk of overfitting, a Dropout layer is added after the convolutional layer, so that some neurons are randomly inactivated during the training process.

[0083] After being processed by the CNN module, a data representation containing rich local features is obtained. However, these features may still lack global time series correlation. In order to capture this global time dependence, the output of the CNN module is used as the input and passed to the LSTM module. LSTM is suitable for processing time series data. By introducing mechanisms such as input gates, output gates, and forget gates, it effectively solves the problems of gradient disappearance and gradient explosion in traditional recurrent neural networks. In the LSTM module, three layers of LSTM cells are used to gradually construct the time series representation of the current data. These representations not only contain local current features but also integrate global time dependence, thus forming more comprehensive and accurate data features.

[0084] In the CNN-LSTM neural network model constructed in the embodiments of the present invention, the CNN module is responsible for extracting local features, while the LSTM module is responsible for capturing global temporal dependencies. The data processed by the CNN is made to meet the input data dimensions of the LSTM, and together with connection and parameter setting, they jointly form a powerful feature extractor. During the model training process, the mean square error and the gradient descent method are used as the loss function and the optimization algorithm to minimize the prediction error, and the model performance is gradually optimized by iteratively updating the model parameters.

[0085] Figure 2 is a flowchart for rapid prediction of the temperature field provided according to the embodiments of the present invention. As Figure 2 shown, specifically, there is a physical information correlation between the measured current and the temperature field state during the operation of the electrolytic cell, and it can intuitively reflect the temperature distribution inside the electrolytic cell. Therefore, the current data and the modal coefficients obtained by POD decomposition of the temperature field are used as the data set of the neural network, that is, the busbar current data is used as the input of the model, and the POD modal coefficients are set as the output of the model. And this data set has been carefully preprocessed and normalized to ensure the accuracy and consistency of the data.

[0086] In order to make full use of the time series data and objectively and fairly evaluate the performance of the model, the data set is divided into a training set and a test set in a ratio of 8:2, and the neural network prediction model is trained with the current under different conditions of the aluminum electrolytic cell as the input and the modal coefficients obtained by POD dimensionality reduction decomposition as the output; in the collectively constructed CNN-LSTM neural network model, the CNN module is responsible for extracting local features, and the LSTM module is responsible for capturing global temporal dependencies. The data processed by the CNN is made to meet the input data dimensions of the LSTM, and together with connection and parameter setting, they jointly form a powerful feature extractor. During the model training process, the mean square error and the gradient descent method are used as the loss function and the optimization algorithm to minimize the prediction error, and the model performance is gradually optimized by iteratively updating the model parameters.

[0087] Specifically, step S110 further includes the following steps:

[0088] Step S1101, using the current data to be predicted of the aluminum electrolytic cell as the input data of the trained CNN-LSTM model to obtain the target reduced-order mode corresponding to the current data to be predicted.

[0089] Step S1102, reconstructing the temperature field data of the aluminum electrolytic cell based on the target reduced-order mode to obtain the predicted temperature field data.

[0090] Specifically, using the POD basis function selection formula, the modal coefficients that satisfy the energy condition are the first four orders. Therefore, the first four-order modal coefficients are predicted by using the constructed model. Then, three key indicators, namely the mean absolute error, root mean square error, and coefficient of determination, are adopted to comprehensively evaluate the prediction accuracy and effect of the proposed method on experimental data and real data, so as to ensure the accuracy and reliability of the model performance. The specific formulas are as follows:

[0091]

[0092]

[0093] There are great difficulties in directly measuring the internal temperature of an aluminum electrolysis cell. In actual operation, the temperature on the surface of the cell shell is often collected. This is because the electrolysis cell has the above temperature distribution characteristics, and by monitoring and analyzing the temperature on the outer surface of the cell shell, the overall operating state of the electrolysis cell can be effectively estimated. Therefore, the cell shell is selected as the analysis object. By combining the predicted modal coefficients with the modes obtained from the POD decomposition of the thermal field and using the reconstruction formula, the temperature field of the cell shell can be reconstructed.

[0094] As can be seen from the above description, the embodiment of the present invention provides a method for predicting the temperature field of an aluminum electrolysis cell based on an improved CNN-LSTM. By performing dimensionality reduction processing on the temperature field data, the problems in the prior art that require high computing resources and time costs when using a large amount of data and in the high-dimensional situation can be effectively overcome. By constructing a CNN-LSTM model, the problem of large local prediction errors existing in the separate LSTM prediction method is overcome, and rapid and accurate prediction of the temperature field of the aluminum electrolysis cell is realized.

[0095] Embodiment Two

[0096] Figure 3 is a schematic diagram of a system for predicting the temperature field of an aluminum electrolysis cell based on an improved CNN-LSTM provided by an embodiment of the present invention. As Figure 3 shown, the system includes: a first construction module 10, a simulation module 20, a dimensionality reduction module 30, a second construction module 40, and a prediction module 50.

[0097] Specifically, the first construction module 10 is used to construct a multi-physical field coupling model of the aluminum electrolysis cell based on the operating performance parameters of the aluminum electrolysis cell; the multi-physical field coupling model includes an electric field mathematical model, an electrical contact mathematical model in the electric field, and a temperature field mathematical model;

[0098] The simulation module 20 is used to calculate the temperature field data of the aluminum electrolysis cell under different current data based on the boundary conditions during the operation of the aluminum electrolysis cell and the multi-physical field coupling model;

[0099] The dimensionality reduction module 30 is used to perform dimensionality reduction on the temperature field data and extract the reduced-order modes of the temperature field data;

[0100] The second construction module 40 is used to construct and train a CNN-LSTM model based on different current data and the corresponding reduced-order modes to obtain a trained CNN-LSTM model;

[0101] The prediction module 50 is used to predict the temperature field of the aluminum electrolysis cell based on the trained CNN-LSTM model.

[0102] Specifically, the dimensionality reduction module 30 is further used for:

[0103] Normalize the temperature field data to obtain normalized temperature field data;

[0104] Based on the proper orthogonal decomposition method, perform dimensionality reduction on the normalized temperature field data to obtain the reduced-order modes of the temperature field data.

[0105] Specifically, the prediction module 50 is further used for:

[0106] Use the current data to be predicted of the aluminum electrolysis cell as the input data of the trained CNN-LSTM model to obtain the target reduced-order mode corresponding to the current data to be predicted;

[0107] Reconstruct the temperature field data of the aluminum electrolysis cell based on the target reduced-order mode to obtain predicted temperature field data.

[0108] The present invention also provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method provided in the embodiment of the present invention is implemented.

[0109] The present invention also provides a computer-readable storage medium. Program codes are stored in the computer-readable storage medium, and the program codes can be called by the processor to execute the method provided in the embodiment of the present invention.

[0110] It should be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DR RAM).

[0111] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0112] It should be understood that in various embodiments of the present invention, the order numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0113] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0114] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0115] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections between each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0116] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0117] In addition, in each embodiment of the present invention, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0118] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0119] As described above, the above are only specific implementation manners of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for predicting the temperature field of an aluminum electrolytic cell based on an improved CNN-LSTM, characterized in that: The method comprises: Based on the operating performance parameters of the aluminum reduction cell, a multi-physical field coupling model of the aluminum reduction cell is constructed; the multi-physical field coupling model includes an electric field mathematical model, an electric contact mathematical model in the electric field, and a temperature field mathematical model; Based on the boundary conditions of the aluminum electrolysis cell during operation and the multi-physics field coupling model, calculating the temperature field data of the aluminum electrolysis cell under different current data; Performing dimensionality reduction processing on the temperature field data to extract reduced-order modes of the temperature field data; Based on different current data and corresponding reduced-order modes, a CNN-LSTM model is constructed and trained to obtain a trained CNN-LSTM model; The temperature field of the aluminum electrolysis cell is predicted based on the trained CNN-LSTM model.

2. The method according to claim 1, characterized in that: The electric field mathematical model includes: In the formula, is the Hamiltonian operator; σ is the conductivity, σ x , σ y , σ z are the components of conductivity in the x, y, and z directions respectively; J is the current density; V is the electric potential; E is the electric field strength; The mathematical model of electrical contact in the electric field includes: Where J is the current density passing through the contact surface; σ c is the contact conductivity; and is the contact surface potential; The temperature field mathematical model includes: In the formula, k x , k y , k z are the thermal conductivity in the x, y, and z directions respectively; T is the temperature; q s is the intensity of the heat source.

3. The method according to claim 1, characterized in that The boundary conditions include electric field boundary conditions and thermal field boundary conditions; wherein, The electric field boundary conditions include: setting a current flow into the column busbar inlet section, setting the potential at the cathode busbar outlet section to zero, and setting the boundaries around the aluminum liquid to insulation conditions; The thermal field boundary conditions include: the ambient temperature of the aluminum electrolysis cell, the convection loss, radiation loss, blackness of the outer surface of the cell shell, the convection heat transfer mode and the radiation heat transfer mode; The convection losses include: Q c =a c (T1-T0)S In the formula, Q c is the convection loss, a c is the convective heat transfer coefficient; T1 is the outer surface temperature of the tank shell; T2 is the ambient temperature; S is the heat dissipation area; The radiation losses include: In the formula, Q r is the radiation loss, ε is the blackness of the heat dissipation surface of the tank shell; κ is the Stefan-Boltzmann constant; It is the mutual radiation angle coefficient between the radiating surface and the adjacent surface.

4. The method according to claim 1, characterized in that: Performing dimensionality reduction processing on the temperature field data to extract reduced-order modes of the temperature field data includes: Normalizing the temperature field data to obtain normalized temperature field data; Based on the intrinsic orthogonal decomposition method, the normalized temperature field data is subjected to dimensionality reduction processing to obtain the reduced-order modes of the temperature field data.

5. The method according to claim 1, characterized in that The temperature field of the aluminum reduction cell is predicted based on the trained CNN-LSTM model, including: Taking the current data to be predicted of the aluminum electrolysis cell as input data of the trained CNN-LSTM model, obtaining a target reduced-order mode corresponding to the current data to be predicted; The temperature field data of the aluminum electrolysis cell is reconstructed based on the target reduced-order mode to obtain predicted temperature field data.

6. An aluminum electrolysis cell temperature field prediction system based on improved CNN-LSTM, characterized in that: include: A first building module, a simulation module, a dimensionality reduction module, a second building module and a prediction module; wherein, The first construction module is used to construct a multi-physical field coupling model of the aluminum electrolysis cell based on the operating performance parameters of the aluminum electrolysis cell; the multi-physical field coupling model includes an electric field mathematical model, an electric contact mathematical model in the electric field, and a temperature field mathematical model; The simulation module is used to calculate the temperature field data of the aluminum electrolysis cell under different current data based on the boundary conditions of the aluminum electrolysis cell during operation and the multi-physics field coupling model; The dimension reduction module is used to perform dimension reduction processing on the temperature field data and extract the reduced-order modes of the temperature field data; The second building module is used to build and train a CNN-LSTM model based on different current data and corresponding reduced-order modes to obtain a trained CNN-LSTM model; The prediction module is used to predict the temperature field of the aluminum electrolysis cell based on the trained CNN-LSTM model.

7. The system according to claim 6, characterized in that The dimension reduction module is also used for: Normalizing the temperature field data to obtain normalized temperature field data; Based on the intrinsic orthogonal decomposition method, the normalized temperature field data is subjected to dimensionality reduction processing to obtain the reduced-order modes of the temperature field data.

8. The system according to claim 6, characterized in that The prediction module is further used for: Taking the current data to be predicted of the aluminum electrolysis cell as input data of the trained CNN-LSTM model, obtaining a target reduced-order mode corresponding to the current data to be predicted; The temperature field data of the aluminum electrolysis cell is reconstructed based on the target reduced-order mode to obtain predicted temperature field data.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 5 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program codes, which can be called by a processor to execute the method according to any one of claims 1 to 5.

Images