Methods, systems, equipment and media for identifying superheat in aluminum electrolytic cells

Abstract

In the embodiment of the present disclosure, a method, system, device and medium for identifying the overheat of an aluminum reduction cell are provided, which belongs to the field of data processing technology, and specifically includes: selecting various types of production data as sample data sets; adding labels and performing isomorphic processing to obtain training sets and verification sets; constructing an initial recognition model based on a self-attention mechanism; pre-training the initial recognition model using an encoding-decoding method and a sample data set to complete initialization; training the initialized initial recognition model according to the weighted cross entropy loss of the training set and the verification set to obtain a target recognition model; collecting production data of the current period, performing isomorphic processing and inputting the target recognition model to obtain a recognition result. Through the scheme of the present disclosure, an initial recognition model based on a self-attention mechanism is constructed, and unsupervised pre-training and the use of a weighted cross entropy loss function are performed to reduce the bias of the classification model, thereby improving the adaptability, efficiency and accuracy of recognition.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of data processing technology, and in particular to a method, system, device and medium for identifying overheating of an aluminum reduction cell. Background Art

[0002] At present, in the aluminum electrolysis production process, superheat is one of the key indicators reflecting the current production efficiency and quality of aluminum electrolytic cells. Maintaining an appropriate superheat can improve current efficiency, stabilize the production process, and reduce production energy consumption. The existing superheat measurement method requires the electrolyte temperature and the primary crystal temperature to be measured separately. The electrolyte temperature is generally measured online using a thermocouple or an infrared thermometer, while the measurement of the primary crystal temperature needs to be carried out offline by sampling and testing. The measurement process is relatively complicated, the measurement cost is high, and real-time measurement cannot be performed. In the actual production process of the current aluminum electrolysis plant, the process personnel mainly observe the fire eye video to judge the superheat with their naked eyes. However, due to the high temperature, large magnetic field and strong corrosiveness of the aluminum electrolytic cell production environment, it will cause great interference to manual judgment. At the same time, the process of manual judgment of superheat is heavily dependent on a small number of experienced process personnel, and there are subjective factors in the judgment process. These two factors lead to unstable electrolysis state and difficulty in adjusting to the optimal state, resulting in low electrolysis efficiency and waste of resources.

[0003] In recent years, with the development of big data analysis technology, production data-driven methods have emerged, which train models based on production process data to replace process personnel in superheat identification. This processing method simplifies the time series information that reflects the changes in production data at different times during the aluminum electrolysis production process, and lacks a description of the dynamic changes in production data. At the same time, due to the different types of sensors at the aluminum electrolysis site, the sampling frequencies obtained are inconsistent, and the data presentation forms are also different. At the same time, due to occasional sensor failures in production and environmental interference during transmission, there are a small number of missing production data, both of which can interfere with the recognition performance of the model; since the superheat in the actual production process is decided by process personnel, it is costly to obtain superheat labels, the number of superheat labels is relatively small compared to production data, and the number of label categories is unbalanced, which may lead to deviations in the discrimination of subsequent training models.

[0004] It can be seen that the existing aluminum electrolytic cell superheat identification method has problems with adaptability, recognition efficiency and accuracy. Summary of the invention

[0005] In view of this, the embodiments of the present disclosure provide a method, system, device and medium for identifying the overheat of an aluminum electrolysis cell, which at least partially solve the problems of poor adaptability, identification efficiency and accuracy existing in the prior art.

[0006] In a first aspect, an embodiment of the present disclosure provides a method for identifying superheat of an aluminum electrolysis cell, comprising:

[0007] Select various types of production data from the aluminum electrolysis production data set as sample data sets;

[0008] Labeling the sample data set and performing isomorphism processing to obtain a training set and a validation set;

[0009] Construct an initial recognition model based on the self-attention mechanism;

[0010] Pre-training the initial recognition model using the encoding-decoding method and the sample data set to complete initialization;

[0011] The initial recognition model is initialized according to the weighted cross entropy loss training of the training set and the validation set to obtain a target recognition model;

[0012] The production data of the current period is collected, isomorphized and then input into the target recognition model to obtain the recognition result.

[0013] According to a specific implementation method of an embodiment of the present disclosure, the types of the sample data sets include monitoring attribute values ​​obtained online in real time during the production process, test attribute values ​​obtained by manual offline testing during the production process, the time value from the production operation in the production process to the current moment, and the 8-bit status value of the production status in the production process.

[0014] According to a specific implementation of the embodiment of the present disclosure, the step of adding labels to the sample data set and performing isomorphism processing to obtain a training set and a validation set includes:

[0015] Adding labels to the sample data sets, and converting different types of production data into time series data sets with the same representation form and sampling frequency;

[0016] The time series data set is divided into the training set and the validation set.

[0017] According to a specific implementation of an embodiment of the present disclosure, the step of constructing an initial recognition model based on a self-attention mechanism includes:

[0018] A convolutional neural network is used to form a residual convolution layer, and a multi-head self-attention mechanism is used to form a self-attention layer;

[0019] The residual convolution layer and the self-attention layer form the self-attention model;

[0020] Multiple stacked self-attention modules and a softmax classifier form the initial recognition model.

[0021] According to a specific implementation of the embodiment of the present disclosure, the step of pre-training the initial recognition model using the encoding-decoding method and the sample data set to complete the initialization includes:

[0022] Converting the multidimensional time series in the sample data set into high-dimensional deep features;

[0023] Decoding the high-dimensional deep features and converting the decoded features into approximate values ​​corresponding to the sample data set;

[0024] Calculate the mean square error loss function of the sample data set and its corresponding approximate value;

[0025] The initial recognition model is initialized according to the mean square error loss function.

[0026] According to a specific implementation of the embodiment of the present disclosure, the step of training the initial recognition model initialized according to the weighted cross entropy loss of the training set and the validation set to obtain the target recognition model includes:

[0027] Obtaining a weighted cross entropy loss function according to the training set and the validation set;

[0028] The initialized initial recognition model is trained according to the value of the weighted cross entropy loss function to obtain the target recognition model.

[0029] In a second aspect, the present disclosure provides an aluminum electrolysis cell superheat identification system, comprising:

[0030] A selection module is used to select various types of production data from the aluminum electrolysis production data set as a sample data set;

[0031] A homogeneity module, used to add labels to the sample data set and perform homogeneity processing to obtain a training set and a validation set;

[0032] Building module, used to build the initial recognition model based on the self-attention mechanism;

[0033] A pre-training module, used to pre-train the initial recognition model using an encoding-decoding method and the sample data set to complete initialization;

[0034] A training module, used for training the initialized initial recognition model according to the weighted cross entropy loss of the training set and the validation set to obtain a target recognition model;

[0035] The recognition module is used to collect the production data of the current period, perform isomorphism processing, and then input the data into the target recognition model to obtain the recognition result.

[0036] In a third aspect, an embodiment of the present disclosure further provides an electronic device, the electronic device comprising:

[0037] at least one processor; and,

[0038] a memory communicatively connected to the at least one processor; wherein,

[0039] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the aluminum electrolysis cell superheat identification method in the aforementioned first aspect or any implementation of the first aspect.

[0040] In a fourth aspect, an embodiment of the present disclosure further provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute the aluminum electrolysis cell superheat identification method in the aforementioned first aspect or any implementation of the first aspect.

[0041] In a fifth aspect, an embodiment of the present disclosure further provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the aluminum electrolysis cell superheat identification method in the aforementioned first aspect or any implementation of the first aspect.

[0042] The aluminum electrolytic cell overheat identification scheme in the disclosed embodiment includes: selecting multiple types of production data from the aluminum electrolysis production data set as sample data sets; adding labels to the sample data sets and performing isomorphism processing to obtain a training set and a verification set; constructing an initial recognition model based on a self-attention mechanism; using an encoding-decoding method and the sample data set to pre-train the initial recognition model to complete initialization; training the initialized initial recognition model according to the weighted cross entropy loss of the training set and the verification set to obtain a target recognition model; collecting the production data of the current period, performing isomorphism processing, and inputting it into the target recognition model to obtain a recognition result.

[0043] The beneficial effects of the embodiments of the present disclosure are as follows: through the scheme of the present disclosure, model training samples are constructed and isomorphic processing is performed, and then an initial recognition model based on the self-attention mechanism is constructed, and unsupervised pre-training is performed and a weighted cross entropy loss function is used to reduce the bias of the classification model to obtain a target recognition model, thereby improving the adaptability, efficiency and accuracy of recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 A schematic flow chart of a method for identifying superheat of an aluminum electrolysis cell provided in an embodiment of the present disclosure;

[0046] Figure 2 A schematic diagram of a data processing flow involved in a method for identifying superheat of an aluminum electrolysis cell provided in an embodiment of the present disclosure;

[0047] Figure 3 A schematic diagram of the structure of an aluminum electrolysis cell superheat identification system provided in an embodiment of the present disclosure;

[0048] Figure 4 A schematic diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0049] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0050] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0051] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein may be used to implement this device and / or practice this method.

[0052] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0053] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0054] The superheat of the electrolyte in an aluminum electrolysis cell refers to the difference between the electrolyte temperature and the crystallization temperature. Superheat is one of the key indicators reflecting the current production efficiency and quality of the aluminum electrolysis cell. Maintaining an appropriate superheat can improve current efficiency, stabilize the production process, and reduce production energy consumption. The existing superheat measurement method requires measuring the electrolyte temperature and the crystallization temperature separately. The electrolyte temperature is generally measured online using a thermocouple or an infrared thermometer, while the measurement of the crystallization temperature needs to be carried out offline by sampling and testing. The measurement process is relatively complicated, the measurement cost is high, and real-time measurement is impossible. In the actual production process of the current aluminum electrolysis plant, the process personnel mainly observe the fire eye video to judge the superheat with their naked eyes. However, due to the high temperature, large magnetic field and strong corrosiveness of the aluminum electrolysis cell production environment, it will cause great interference to manual judgment. At the same time, the process of manual judgment of superheat relies heavily on a small number of experienced process personnel, and there are subjective factors in the judgment process. These two factors lead to unstable electrolysis state and difficulty in adjusting to the optimal state, resulting in low electrolysis efficiency and waste of resources.

[0055] In recent years, with the development of big data analysis technology, production data-driven methods have emerged, which use models trained based on production process data to replace process personnel in superheat identification. The superheat identification algorithm based on production process data mainly identifies the superheat in the aluminum electrolysis production process based on the monitoring values ​​of the aluminum electrolysis production process and the chemical analysis values ​​of the output raw aluminum. The monitoring values ​​of the aluminum electrolysis production process and the chemical analysis values ​​of the output raw aluminum reflect the overall overview of the production process and the production quality in the aluminum electrolysis production process, and indirectly reflect the superheat situation. At present, the superheat identification algorithm based on production process data mainly uses traditional machine learning algorithms such as rule books, random forests, extreme learning machines, etc. to model the above features and obtain the superheat in the aluminum electrolysis production process. However, these methods belong to static modeling methods. Assuming that the attribute variables in the aluminum electrolysis production process do not fluctuate much, the average, standard deviation and extreme value methods are used to unify the sampling frequency and construct a sample set for time series data with different sampling frequencies. This processing method simplifies the time series information reflecting the changes in production data at different times in the aluminum electrolysis production process, and lacks a description of the dynamic changes in production data.

[0056] Due to the different types of sensors at the aluminum electrolysis site, the sampling frequencies obtained are inconsistent, and there are also differences in the data presentation. At the same time, due to occasional sensor failures in production and environmental interference during transmission, there are small amounts of missing production data, both of which can interfere with the recognition performance of the model. Since the overheating degree is decided by process personnel in the actual production process, it is costly to obtain the overheating label, the number of overheating labels is relatively small compared to production data, and the number of label categories is unbalanced, which may lead to deviations in the model discrimination of subsequent training.

[0057] In general, the current aluminum electrolysis superheat identification method based on process data has the following problems and difficulties:

[0058] The data obtained from the aluminum electrolysis site have inconsistent sampling frequencies and a small amount of missing data;

[0059] The superheat label data obtained from the field is small and unevenly distributed, which may affect model training;

[0060] Existing superheat identification algorithms do not model the dynamic characteristics of production data, but only consider the static characteristics of aluminum electrolysis production data.

[0061] The embodiment of the present disclosure provides a method for identifying the superheat of an aluminum reduction cell, which can be applied to the superheat identification process of an aluminum reduction cell in an industrial production scenario.

[0062] See also Figure 1 , is a schematic diagram of a flow chart of a method for identifying superheat of an aluminum electrolytic cell provided by an embodiment of the present disclosure. Figure 1 As shown, the method mainly comprises the following steps:

[0063] S101, selecting multiple types of production data from the aluminum electrolysis production data set as a sample data set;

[0064] For example, the historical production data of the aluminum reduction cell during the aluminum electrolysis production process may be collected first. Considering that the production data set contains multiple types of data, some data may interfere with the subsequent identification process, multiple types of production data may be selected from the production data set as the sample data set according to actual needs.

[0065] S102, adding labels to the sample data set and performing isomorphism processing to obtain a training set and a validation set;

[0066] In specific implementation, taking into account the differences in source sensors, the representation forms and sampling frequencies of different attributes of production data may not be the same. Therefore, labels can be added to the sample data sets and isomorphized to unify the data format. The unified data sets can then be divided into the training set and the validation set.

[0067] S103, constructing an initial recognition model based on the self-attention mechanism;

[0068] In the specific implementation, considering that the existing methods all simplify the time series information reflecting the changes in production data at different times in the aluminum electrolysis production process, and lack the description of the dynamic changes of production data, an initial recognition model based on the self-attention mechanism can be constructed to identify and extract the dynamic features in the sample data set through the self-attention mechanism.

[0069] S104, pre-training the initial recognition model using the encoding-decoding method and the sample data set to complete initialization;

[0070] In the specific implementation, considering that most of the data obtained in the electrolysis production process are unlabeled data, in order to make full use of the production data obtained in the aluminum electrolysis industry, based on the encoder-decoder idea, a large amount of unlabeled data is collected to pre-train the initial recognition model, and the common features in the unlabeled data are learned in an unsupervised manner. The encoder layer obtained through unsupervised training is used as the pre-training model of the model, and its weight is used as the initialization of superheat recognition training.

[0071] S105, training the initialized initial recognition model according to the weighted cross entropy loss of the training set and the validation set to obtain a target recognition model;

[0072] In specific implementation, taking into account the uneven distribution of data labels obtained in the aluminum electrolysis production process, which will lead to increased bias of the model, the initial recognition model after initialization can be trained according to the weighted cross entropy loss of the training set and the validation set to obtain a target recognition model, thereby reducing the bias of the model and making the recognition result more accurate.

[0073] S106, collecting production data of the current period, performing isomorphism processing, and inputting the data into the target recognition model to obtain a recognition result.

[0074] For example, after obtaining the target recognition model, the production data within the current 5 minutes can be collected, and the collected production data can be isomorphized and then input into the target recognition model to obtain the corresponding aluminum electrolysis superheat recognition result. Figure 2 shown.

[0075] The aluminum electrolytic cell superheat identification method provided in this embodiment constructs a model training sample and performs isomorphism processing, then builds an initial recognition model based on the self-attention mechanism, performs unsupervised pre-training, and uses a weighted cross entropy loss function to reduce the bias of the classification model to obtain a target recognition model, thereby improving the adaptability, efficiency and accuracy of recognition.

[0076] Based on the above embodiment, the types of sample data sets include monitoring attribute values ​​obtained online in real time during the production process, test attribute values ​​obtained by manual offline testing during the production process, the time value from the production operation to the current moment in the production process, and the 8-bit status value of the production status in the production process.

[0077] Furthermore, in step S102, adding labels to the sample data set and performing isomorphism processing to obtain a training set and a validation set includes:

[0078] Adding labels to the sample data sets, and converting different types of production data into time series data sets with the same representation form and sampling frequency;

[0079] The time series data set is divided into the training set and the validation set.

[0080] For example, let X = {X1, X2, X3, X4} be the production data of an aluminum electrolysis plant for one day, where represents the i-th type of production data, T i Indicates the number of data sampling points per day, D iIndicates the number of attributes, i.e., dimensions. Production data includes four types: data type X1 represents the monitoring attribute values ​​obtained online and in real time during the production process; data type X2 represents the test attribute values ​​obtained by manual offline testing during the production process; data type X3 represents the time value obtained online and in real time during the production process to represent a certain operation to date; data type X4 represents the 8-bit status value obtained online and in real time during the production process to represent certain production statuses. The steps of adding labels to the sample data set and performing isomorphism processing to obtain the training set and validation set may include:

[0081] 1) For data types Where T1 = 720. Perform one-dimensional convolution on the input data in the T1 dimension, where the convolution kernel size is 3, the step size is 1, and the expansion is 1. The process is expressed as:

[0082] X 1_emb =X1*W conv

[0083] Where (*) represents the convolution operation.

[0084] 2) For data types Where T2 = 1. Map it to another vector space through linear transformation and copy it along the dimension of time sampling points T emb times. The process is expressed as:

[0085]

[0086] Broadcast represents the copy operation along the time dimension.

[0087] 3) For data types Where T3 = 720. For each attribute dimension, the time value vector Among them, d∈[1,D3] uses fixed position embedding to map discrete time values ​​to a fixed vector space.

[0088]

[0089] Adding the position embeddings of all attribute dimensions gives:

[0090]

[0091] 4) For data types. Where T4 = 720. For each attribute dimension, the 8-bit state value vector d∈[1,D4], maps each 8-bit state value to a real-valued vector space. The process is expressed as:

[0092]

[0093] Combined with the above formula, different types of production data in the aluminum electrolysis production process are homogenized to obtain a unified time series representation. for:

[0094]

[0095] The production data is homogenized to obtain a time series production data set with the same representation and sampling frequency. The obtained production data set is divided to obtain training set samples and validation set samples.

[0096] On the basis of the above embodiment, step S103, constructing an initial recognition model based on the self-attention mechanism, includes:

[0097] A convolutional neural network is used to form a residual convolution layer, and a multi-head self-attention mechanism is used to form a self-attention layer;

[0098] The residual convolution layer and the self-attention layer form the self-attention model;

[0099] Multiple stacked self-attention modules and a softmax classifier form the initial recognition model.

[0100] For example, in the specific recognition process, let the input of the self-attention layer be Where L is the number of time sampling points of the input time series data, D k is the number of attributes or dimensions of the input time series data. First, three different linear transformations are used to project the input X into three different feature spaces. Then, the matrix Q, matrix K, and matrix V are divided into blocks, and the attribute quantity value dimension is divided into h parts, and the result is Where h is the number of channels of multi-head attention, indicating that h self-attention outputs are calculated in parallel, and l∈[1,h] represents the lth self-attention output. Respectively represent matrices The i-th row of the matrix is ​​the data at the i-th sampling time point. The l-th self-attention output is calculated for each With all key vectors Attention distribution obtained by similarity The corresponding value vector The weighted sum of .

[0101] The query vector at the i-th time point For the key vector at the jth time point The attention distribution is:

[0102]

[0103] remember is the lth self-attention output, then:

[0104]

[0105] Based on the above formula, all self-attention outputs are calculated and concatenated, and then the multi-head attention output O is obtained through linear transformation:

[0106]

[0107] Add the multi-head attention output O to the input X, and then perform layer normalization to get the final output of the self-attention layer. for:

[0108] Y=LN(O+X)

[0109] The residual convolution layer consists of two 1D convolution layers. The input Y passes through two convolution layers to obtain its output, and the convolution output is added to the original input Y to obtain the final output Z.

[0110] Z = Conv(ReLU(Conv(Y))) + Y

[0111] Residual connection can transfer the input information of a certain layer directly to the output, avoiding the information loss caused by the convolution layer and ensuring the model's ability to express complex data.

[0112] On the basis of the above embodiment, the step S104 described in which the initial recognition model is pre-trained using the encoding-decoding method and the sample data set to complete initialization includes:

[0113] Converting the multidimensional time series in the sample data set into high-dimensional deep features;

[0114] Decoding the high-dimensional deep features and converting the decoded features into approximate values ​​corresponding to the sample data set;

[0115] Calculate the mean square error loss function of the sample data set and its corresponding approximate value;

[0116] The initial recognition model is initialized according to the mean square error loss function.

[0117] In specific implementation, the encoder layer is composed of multiple stacked self-attention modules consistent with the aluminum electrolysis initial recognition model described above. The encoder layer is responsible for converting the isomorphized multidimensional time series into high-dimensional deep features; the decoder layer is divided into two parts: the feature decoding part and the reconstruction output part. The feature decoding part is composed of multiple stacked self-attention modules, which is responsible for gradually decoding the high-dimensional deep features. The reconstruction output part is responsible for converting the decoded features into an approximation of the original input production data of the encoder layer.

[0118] Unsupervised pre-training uses the mean squared error loss (MSELoss) between the original input production data and its corresponding approximation as the loss function:

[0119]

[0120] After completing the unsupervised pre-training, the weight parameters of the Encoder layer are taken to initialize the initial recognition model.

[0121] Furthermore, in step S105, the initial recognition model initialized according to the weighted cross entropy loss training of the training set and the validation set is obtained to obtain the target recognition model, including:

[0122] Obtaining a weighted cross entropy loss function according to the training set and the validation set;

[0123] The initialized initial recognition model is trained according to the value of the weighted cross entropy loss function to obtain the target recognition model.

[0124] In specific implementation, a weighted cross entropy loss function may be obtained according to the training set and the validation set. For example, the weighted cross entropy loss function is as follows:

[0125]

[0126] Where C = 3 is the number of superheat categories, w j is the weight of the j-th category loss, w j Satisfy w j ∝1 / N j And ∑w j =C,N j is the number of samples in the jth category. Then, the initialized initial recognition model is trained according to the value of the weighted cross entropy loss function to obtain the target recognition model.

[0127] Corresponding to the above method embodiment, see Figure 3 The embodiment of the present disclosure further provides an aluminum electrolysis cell superheat identification system 30, comprising:

[0128] A selection module 301 is used to select multiple types of production data from the aluminum electrolysis production data set as a sample data set;

[0129] A homogeneity module 302 is used to add labels to the sample data set and perform homogeneity processing to obtain a training set and a validation set;

[0130] A construction module 303 is used to construct an initial recognition model based on a self-attention mechanism;

[0131] A pre-training module 304 is used to pre-train the initial recognition model using the encoding-decoding method and the sample data set to complete initialization;

[0132] A training module 305 is used to train the initialized initial recognition model according to the weighted cross entropy loss of the training set and the validation set to obtain a target recognition model;

[0133] The recognition module 306 is used to collect the production data of the current period, perform isomorphism processing, and then input the data into the target recognition model to obtain the recognition result.

[0134] Figure 3 The system shown can correspondingly execute the contents of the above method embodiment. For the parts not described in detail in this embodiment, refer to the contents recorded in the above method embodiment and will not be repeated here.

[0135] See also Figure 4 The embodiment of the present disclosure further provides an electronic device 40, which includes: at least one processor and a memory connected to the at least one processor. The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the aluminum electrolysis tank superheat identification method in the above method embodiment.

[0136] The embodiment of the present disclosure also provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute the aluminum electrolysis cell superheat identification method in the aforementioned method embodiment.

[0137] The embodiments of the present disclosure also provide a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the aluminum electrolysis cell superheat identification method in the aforementioned method embodiment.

[0138] Reference below Figure 4, which shows a schematic diagram of the structure of an electronic device 40 suitable for implementing the embodiment of the present disclosure. The electronic device in the embodiment of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0139] like Figure 4 As shown, the electronic device 40 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage device 408 to a random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation of the electronic device 40 are also stored. The processing device 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0140] Typically, the following devices may be connected to the I / O interface 405: input devices 406 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 407 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 408 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 409. The communication device 409 may allow the electronic device 40 to communicate wirelessly or wired with other devices to exchange data. Although the electronic device 40 with various devices is shown in the figure, it should be understood that it is not required to implement or have all the devices shown. More or fewer devices may be implemented or have instead.

[0141] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 409, or installed from the storage device 408, or installed from the ROM 402. When the computer program is executed by the processing device 401, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0142] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0143] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0144] The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device can execute the relevant steps of the method embodiment.

[0145] Alternatively, the computer-readable medium carries one or more programs, and when the one or more programs are executed by the electronic device, the electronic device can perform the relevant steps of the method embodiment.

[0146] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0147] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0148] The units involved in the embodiments described in the present disclosure may be implemented by software or by hardware.

[0149] It should be understood that various parts of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.

[0150] The above is only a specific implementation of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present disclosure should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for identifying superheat of an aluminum electrolytic cell, characterized in that: include: Select multiple types of production data from the aluminum electrolysis production data set as sample data sets, where the types of the sample data sets include monitoring attribute values ​​obtained online and in real time during the production process, test attribute values ​​obtained by manual offline testing during the production process, time values ​​from the production operation to the current moment during the production process, and 8-bit status values ​​of the production status during the production process; Labeling the sample data set and performing isomorphism processing to obtain a training set and a validation set; The step of adding labels to the sample data set and performing isomorphism processing to obtain a training set and a validation set includes: Adding labels to the sample data sets, and converting different types of production data into time series data sets with the same representation form and sampling frequency; The step of adding labels to the sample data set and converting different types of production data into a time series data set with the same representation form and sampling frequency includes: For data types in T 1 ＝720 ,right T Perform a one-dimensional convolution on the input data in 1 dimension, where the convolution kernel size is 3, the step size is 1, and the expansion is 1. The process is expressed as: X 1_emb =X1*W conv Where (*) represents the convolution operation; For data types in T 2 ＝1 , map it to another vector space through a linear transformation and replicate it along the dimension of the number of time sampling points T em b times, the process is expressed as: Among them, Broadcast represents the copy operation along the time dimension; For data types in T 3 ＝720 , for each attribute dimension, the time value vector in d∈[1,D 3 ] ,Fixed position embedding is used to map discrete time values ​​into a fixed vector space; Adding the position embeddings of all attribute dimensions gives: For data types in T 4 ＝720 , for each attribute dimension, the 8-bit state value vector Mapping each 8-bit state value to a real-valued vector space is expressed as: Combined with the above formula, different types of production data in the aluminum electrolysis production process are homogenized to obtain a unified time series representation. for: Dividing the time series data set into the training set and the validation set; Construct an initial recognition model based on the self-attention mechanism; The step of constructing an initial recognition model based on the self-attention mechanism includes: A convolutional neural network is used to form a residual convolution layer, and a multi-head self-attention mechanism is used to form a self-attention layer; The residual convolution layer and the self-attention layer form the self-attention model; Multiple stacked self-attention modules and a softmax classifier form the initial recognition model; Pre-training the initial recognition model using the encoding-decoding method and the sample data set to complete initialization; The initial recognition model is initialized according to the weighted cross entropy loss training of the training set and the validation set to obtain a target recognition model; The production data of the current period is collected, isomorphized and then input into the target recognition model to obtain the recognition result.

2. The method according to claim 1, characterized in that The step of pre-training the initial recognition model using the encoding-decoding method and the sample data set to complete the initialization includes: Converting the multidimensional time series in the sample data set into high-dimensional deep features; Decoding the high-dimensional deep features and converting the decoded features into approximate values ​​corresponding to the sample data set; Calculate the mean square error loss function of the sample data set and its corresponding approximate value; The initial recognition model is initialized according to the mean square error loss function.

3. The method according to claim 1, characterized in that The step of training the initial recognition model initialized according to the weighted cross entropy loss of the training set and the validation set to obtain the target recognition model includes: Obtaining a weighted cross entropy loss function according to the training set and the validation set; The initialized initial recognition model is trained according to the value of the weighted cross entropy loss function to obtain the target recognition model.

4. A superheat identification system for an aluminum electrolytic cell, characterized in that: include: A selection module is used to select multiple types of production data from the aluminum electrolysis production data set as a sample data set, wherein the types of the sample data set include monitoring attribute values ​​obtained online in real time during the production process, test attribute values ​​obtained by manual offline testing during the production process, time values ​​from the production operation to the current moment during the production process, and 8-bit status values ​​of the production status during the production process; A homogeneity module, used to add labels to the sample data set and perform homogeneity processing to obtain a training set and a validation set; The isomorphism module is also used to: Adding labels to the sample data sets, and converting different types of production data into time series data sets with the same representation form and sampling frequency; Dividing the time series data set into the training set and the validation set; Building module, used to build the initial recognition model based on the self-attention mechanism; The building blocks are also used to: The step of constructing an initial recognition model based on the self-attention mechanism includes: A convolutional neural network is used to form a residual convolution layer, and a multi-head self-attention mechanism is used to form a self-attention layer; The residual convolution layer and the self-attention layer form the self-attention model; Multiple stacked self-attention modules and a softmax classifier form the initial recognition model; A pre-training module, used to pre-train the initial recognition model using an encoding-decoding method and the sample data set to complete initialization; A training module, used for training the initialized initial recognition model according to the weighted cross entropy loss of the training set and the validation set to obtain a target recognition model; The recognition module is used to collect the production data of the current period, perform isomorphism processing, and then input the data into the target recognition model to obtain the recognition result.

5. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the aluminum electrolysis cell superheat identification method according to any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the aluminum electrolysis cell superheat identification method according to any one of claims 1 to 3.

Images