A feature extraction model training method, device, and computer equipment

Abstract

The present application relates to the field of data processing technology, and specifically to a feature extraction model training method, device, and computer equipment; by training the feature extraction model, the key features in the electrolytic aluminum process are acquired and the operating status is characterized based on the key features. The specific process first uses the hidden layer of the denoising autoencoder as the input layer, and configures a new hidden layer and output layer after the input layer to form a three-layer BP artificial neural network, and then adds the loss function obtained from the process output layer to the loss function of the denoising autoencoder reconstruction layer; finally, the model structure is trained with the constructed distinguishing composite error function as the target criterion, which is expected to extract more robust feature information.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a feature extraction model training method, device, and computer equipment. Background Art

[0002] With the rapid development of artificial intelligence (AI) technology, AI technology has been widely used in various fields. For example, in the industrial field, AI technology can be used to extract key features in the industrial control process, and the operating status of the process can be obtained through the extraction of key features.

[0003] In the electrolytic aluminum equipment process system, there are many process parameters that are closely related to technical performance indicators (current efficiency and DC power consumption per ton of aluminum). However, the coupling and correlation between the various series of process parameters are too strong, and the operating effects of different parameter combinations are difficult to estimate. Therefore, it is not appropriate to use only the main process parameters to participate in the model analysis. In order to fully consider the synergistic effect between the various process parameters, the series of process parameters should be analyzed as a whole for comprehensive characteristics and then participate in precise modeling.

[0004] In order to solve the above problems in the prior art, a denoising autoencoder neural network can be constructed for processing. The denoising autoencoder neural network often sets the error function (loss function) as the target criterion, and trains the weights and thresholds of the network by gradient backpropagation. Its essence is a lossless and robust feature extraction of the input sample data. However, classification or regression information may only exist in certain components of the high-dimensional input sample. Therefore, in the above feature extraction process, optimizing only with the error function (loss function) as the target criterion is likely to cause the regression information to be unclear and ignore the feature information that occupies a dominant position. Summary of the invention

[0005] In order to solve the above technical problems, the present application provides a feature extraction model training method, device, and computer equipment to achieve accurate feature extraction by building a reasonable model and training the model.

[0006] In order to achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

[0007] In a first aspect, a feature extraction model training method is characterized in that it includes: obtaining sample data to be trained; inputting the sample data into the model to be trained to obtain reconstruction error function information and target prediction error function information respectively; fusing the reconstruction error function and the target prediction error function to distinguish a composite error function; training the model to be trained until the error meets a preset value, and extracting a feature extraction vector from the sample data, and inputting the feature extraction vector into a neural network for training until the feature extraction vector error meets a preset value.

[0008] In an exemplary embodiment of the present invention, the model structure includes distinguishing between a denoising autoencoder network and an ANN neural network.

[0009] In an exemplary embodiment of the present invention, the sample data is input into the model to be trained to obtain reconstruction error function information and target prediction error function information, respectively, including: inputting the sample data into the autoencoder network to obtain reconstruction error function information; inputting the sample data after encoding into the ANN neural network to obtain target prediction error function information.

[0010] In an exemplary embodiment of the present invention, training the model to be trained until the error meets a preset value includes: based on the distinguishable composite error function, using a gradient descent algorithm to train a distinguishable denoising autoencoder network until the obtained error meets a preset error value.

[0011] In an exemplary embodiment of the present invention, the discriminative denoising autoencoder network includes a hidden layer.

[0012] In an exemplary embodiment of the present invention, extracting a feature extraction vector from the sample data includes: when an error in a noise reduction autoencoder network satisfies a preset error value, obtaining a feature extraction vector of the input sample data from a hidden layer.

[0013] In an exemplary embodiment of the present invention, the feature extraction vector is input into a neural network for training until the feature extraction vector error meets a preset value, comprising: inputting the extracted feature extraction vector into the ANN neural network model for training until the feature extraction vector error meets a preset error value.

[0014] In a second aspect, the present invention also provides a feature extraction method, based on the feature extraction model obtained in any of the feature extraction model training methods mentioned above, specifically including the following method: inputting sample data into the feature extraction model; obtaining the corresponding feature expression through the hidden layer processing in the distinguishable autoencoder network.

[0015] In the third aspect, the present invention also provides a feature extraction model training device, including: an acquisition module, which acquires sample data to be used for training; a first function acquisition module, which is used to obtain reconstruction error function information and target prediction error function information; a fusion module, which is used to fuse the reconstruction error function and the target prediction error function to distinguish a composite error function; a training module, which is used to train the model to be trained until the error meets a preset value, and extract a feature extraction vector from the sample data, and input the feature extraction vector into a neural network for training until the feature extraction vector error meets a preset value.

[0016] In a fourth aspect, the present invention further provides an electronic device, comprising: a memory for storing executable instructions; and a processor for implementing any of the above-mentioned feature extraction model training methods when running the executable instructions stored in the memory.

[0017] In a fifth aspect, the present invention also provides a computer-readable storage medium storing executable instructions, which, when executed by a processor, implement any of the above-mentioned AI object feature extraction model training methods.

[0018] In the technical solution provided by the embodiment of the present application, the key features in the electrolytic aluminum process are acquired by training the feature extraction model and the characterization of the operating status based on the key features is realized. The specific process first uses the hidden layer of the denoising autoencoder as the input layer, and configures a new hidden layer and output layer after the input layer to form a three-layer BP artificial neural network, and then adds the loss function obtained from the process output layer to the loss function of the denoising autoencoder reconstruction layer; finally, the model structure is trained with the constructed distinguishing composite error function as the target criterion, which is expected to extract more robust feature information. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] The methods, systems and / or programs in the accompanying drawings will be further described according to exemplary embodiments. These exemplary embodiments will be described in detail with reference to the drawings. These exemplary embodiments are non-limiting exemplary embodiments, wherein example numbers represent similar mechanisms in the various views of the accompanying drawings.

[0021] Figure 1 It is a schematic diagram of a usage scenario of the feature extraction model training method provided in an embodiment of the present invention.

[0022] Figure 2 It is a schematic diagram of the composition structure of the feature extraction model training device provided in an embodiment of the present invention.

[0023] Figure 3 It is a flowchart of the feature extraction model training method provided by an embodiment of the present invention.

[0024] Figure 4 It is a flowchart of the feature extraction method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0025] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present invention will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present invention. However, those skilled in the art will appreciate that the technical solutions of the present invention may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present invention.

[0026] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0027] Before further describing the embodiments of the present invention in detail, the nouns and terms involved in the embodiments of the present invention are described. The nouns and terms involved in the embodiments of the present invention are subject to the following interpretations.

[0028] (1) In response, it is used to indicate the conditions or states on which the executed operations depend. When the dependent conditions or states are met, one or more operations executed may be in real time or with a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations executed are executed.

[0029] (2) Based on is used to indicate the conditions or states on which the executed operations depend. When the dependent conditions or states are met, one or more operations executed may be in real time or have a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations executed are executed.

[0030] (3) Model training, multi-classification learning of image datasets. The model can be built using deep learning frameworks such as Tensor Flow and torch, and a multi-classification model is composed of multiple layers of neural network layers such as ANN.

[0031] (4) Neural Network (NN): Artificial Neural Network (ANN), also known as neural network or quasi-neural network, is a mathematical model or computational model that imitates the structure and function of biological neural networks (the central nervous system of animals, especially the brain) in the field of machine learning and cognitive science, and is used to estimate or approximate functions.

[0032] The method provided in the present invention can be applied to an aluminum electrolytic production line to obtain characteristics of the aluminum electrolytic process and to characterize and describe the aluminum electrolytic process. The following embodiments are illustrated by using applications in games.

[0033] Figure 1 A schematic diagram of a usage scenario of the feature extraction model training method provided in an embodiment of the present invention, see Figure 1 In the application of the feature extraction model training method provided in the embodiment of the present application, the terminal includes a terminal 110 and a terminal 120, wherein the terminal 120 is located on the developer side to control the training and use of the feature extraction feature model, and the terminal 120 is located on the user side to request the execution of the feature extraction strategy. Accordingly, the server obtains the process parameters sent from the electrolytic aluminum process system, and uses the deployed trained feature extraction model to extract the features of the process parameters generated in the process in the real-time stage; the terminal is connected to the server 200 through the network 300, and the network 300 can be a wide area network or a local area network, or a combination of the two, and uses wireless or wired links to realize data transmission.

[0034] As an example, the server 200 is used to deploy the feature extraction model training device to implement the feature extraction model training method provided by the present invention, which can deploy a trained feature extraction model to extract the characteristics of key parameters in the electrolytic aluminum process. The specific process includes: obtaining sample data to be trained; inputting the sample data into the model to be trained to obtain reconstruction error function information and target prediction error function information respectively; fusing the reconstruction error function and the target prediction error function to distinguish a composite error function; training the model to be trained until the error meets a preset value, and extracting a feature extraction vector from the sample data, and inputting the feature extraction vector into a neural network for training until the feature extraction vector error meets a preset value.

[0035] The structure of the feature extraction model training device of the embodiment of the present invention is described in detail below. The feature extraction model training device can be implemented in various forms, such as a dedicated terminal with a feature extraction model training device processing function, or a server provided with a feature extraction model training device processing function, such as the preamble Figure 1 Server 200 in. Figure 2 The schematic diagram of the structure of the feature extraction model training device provided in the embodiment of the present invention can be understood as follows: Figure 2 Only an exemplary structure of the feature extraction model training device is shown, not the entire structure, and can be implemented as needed. Figure 2 Partial or complete structure shown.

[0036] The feature extraction model training device provided in the embodiment of the present invention includes: at least one processor 201, a memory 202, a user interface 203 and at least one network interface 204. The various components in the feature extraction model training device are coupled together through a bus system 205. It can be understood that the bus system 205 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 205 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, in Figure 2 Various buses are labeled as bus system 205 .

[0037] The user interface 203 may include a display, a keyboard, a mouse, a trackball, a click wheel, keys, buttons, a touch pad or a touch screen.

[0038] It is understood that the memory 202 can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. The memory 202 in the embodiment of the present invention can store data to support the operation of the terminal (such as 10-1). Examples of these data include: any computer program for operating on the terminal (such as 10-1), such as an operating system and an application. Among them, the operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., which are used to implement various basic services and process hardware-based tasks. The application can include various applications.

[0039] In some embodiments, the feature extraction model training device provided by the embodiment of the present invention can be implemented in a combination of software and hardware. As an example, the feature extraction model training device provided by the embodiment of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the feature extraction training method provided by the embodiment of the present invention. For example, the processor in the form of a hardware decoding processor can adopt one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs), or other electronic components.

[0040] As an example of a feature extraction model training device provided in an embodiment of the present invention being implemented by a combination of software and hardware, the feature extraction model training device provided in an embodiment of the present invention can be directly embodied as a combination of software modules executed by a processor 201, and the software module can be located in a storage medium, and the storage medium is located in a memory 202. The processor 201 reads the executable instructions included in the software module in the memory 202, and completes the feature extraction training method provided in an embodiment of the present invention in combination with necessary hardware (for example, including a processor 201 and other components connected to a bus 205).

[0041] As an example, the processor 201 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., where the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0042] As an example of hardware implementation of the feature extraction model training device provided in an embodiment of the present invention, the device provided in an embodiment of the present invention can be directly executed by a processor 201 in the form of a hardware decoding processor, for example, one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs) or other electronic components to implement the feature extraction model training method provided in an embodiment of the present invention.

[0043] The memory 202 in the embodiment of the present invention is used to store various types of data to support the operation of the feature extraction model training device. Examples of such data include: any executable instructions for operating on the feature extraction model training device, such as executable instructions, and the program for implementing the feature extraction model training method of the embodiment of the present invention can be included in the executable instructions.

[0044] In other embodiments, the feature extraction model training device provided by the embodiment of the present invention can be implemented in software. Figure 2 The feature extraction model training device stored in the memory 202 is shown, which may be software in the form of a program and a plug-in, and includes a series of modules. As an example of a program stored in the memory 202, a feature extraction model training device may be included, and the feature extraction model training device includes the following software modules:

[0045] The acquisition module 2021 acquires sample data to be used for training.

[0046] The first function acquisition module 2022 is used to obtain reconstruction error function information and target prediction error function information.

[0047] The fusion module 2023 is used to fuse the reconstruction error function and the target prediction error function into a distinguishable composite error function.

[0048] The training module 2024 is used to train the model to be trained until the error meets the preset value, and extract the feature extraction vector in the sample data, and input the feature extraction vector into the neural network for training until the feature extraction vector error meets the preset value.

[0049] In some embodiments, server 200 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks (CDN, Content Delivery Network), and big data and artificial intelligence platforms. The terminal (such as terminal 10-1) can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected by wired or wireless communication, which is not limited in the embodiments of the present application.

[0050] according to Figure 2 The electronic device shown, in one aspect of the present application, the present application also provides a computer program product or a computer program, the computer program product or the computer program includes a computer instruction, and the computer instruction is stored in a computer-readable storage medium. The processor of the computer device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the different embodiments and combinations of embodiments provided in various optional implementations of the above-mentioned point game model training method.

[0051] Before introducing the feature extraction model training method deployed in the server, the background of the application of this model and the defects of the technical solutions adopted for this background in the prior art are explained.

[0052] In the electrolytic aluminum equipment process system, there are many process parameters that are closely related to technical performance indicators (such as current efficiency and DC power consumption per ton of aluminum). However, the coupling and correlation between the various series of process parameters are too strong, and the operating effects of different parameter combinations are difficult to estimate. Therefore, it is not appropriate to use only the main process parameters to participate in the model analysis. In order to fully consider the synergistic effect between the various process parameters, the series of process parameters should be analyzed as a whole for comprehensive characteristics and then participate in precise modeling.

[0053] To solve this problem, we will continue to introduce the training process of the feature extraction model deployed in the server, including Figure 3 , Figure 3 An optional flow chart of the feature extraction model training method provided in an embodiment of the present invention is provided. It can be understood that: Figure 3 The steps shown can be performed by various electronic devices running the feature extraction model training device, wherein the dedicated terminal with the feature extraction model training device can be the preamble Figure 2 The embodiment shown is an electronic device with a feature extraction model training device.

[0054] The following is for Figure 3 The steps shown are described in detail.

[0055] Step S301: Obtain sample data to be used for training.

[0056] In this embodiment, because the sample data types included in the electrolytic aluminum process equipment are huge, it is necessary to obtain key data from the huge sample data. In this embodiment, the acquisition of key data is based on historical sample data, specifically: historical sample data is obtained, high-frequency data in the historical sample data is marked as initial key data, and two key data are obtained in combination with the electrolytic aluminum manufacturing process, namely: DC power consumption per ton of aluminum (kW·h / t-Al) y1 and current efficiency y2.

[0057] Set the ideal power consumption-current correspondence model in the process, and the power consumption-current correspondence model obtains process variables: series current, tank working voltage, AE times, NB times, RC times, AE waiting time, NB interval, AE interval, molecular ratio, aluminum output, aluminum level, electrolyte level, tank temperature, fluoride addition times, and daily fluoride salt consumption. Based on the measurement and control system, sample the above data and establish the process variable set X=[x1,x2,…,x 15 ], based on the existence of a certain nonlinear mapping relationship between the process variable set and the target set Y = [y1, y2].

[0058] In this embodiment, if the target variable is directly modeled using the entire process variable set, it will not only seriously affect the modeling accuracy, but also increase the model calculation complexity and reduce the real-time prediction performance of the model due to the high dimension of the input variables.

[0059] In order to eliminate the influence of different dimensions on feature extraction, the sample data X and Y should be normalized.

[0060] y=(y max -y min )*(xx min ) / (x max -x min )+y min ;

[0061] Among them, x is the data before transformation; x min and x max are the minimum and maximum values ​​of x respectively; y is the normalized value, y max and min is the maximum and minimum value of the normalized variable interval. If the normalized interval is [-1,1], then y max and min are 1 and -1 respectively; if the normalized interval is [0,1], then y max and min1 and 0 respectively.

[0062] Step S302, input sample data to the model to be trained, and obtain reconstruction error function information and target prediction error function information respectively.

[0063] In this embodiment, the structure of the model to be trained includes a layer of autoencoder network, a layer of denoising autoencoder network and ANN neural network, wherein the autoencoder network outputs to the denoising autoencoder network outputs to the ANN neural network as input, and outputs the final result. The autoencoder network and the denoising autoencoder network are combined to form a distinguishing denoising autoencoder network.

[0064] Step S303: fusing the reconstruction error function and the target prediction error function into a distinguishable composite error function.

[0065] Step S304, training the model to be trained until the error meets the preset value, extracting the feature extraction vector in the sample data, and inputting the feature extraction vector into the neural network for training until the error of the feature extraction vector meets the preset value.

[0066] In this embodiment, the hidden layer in the denoising autoencoder network is used as the input layer of the ANN. In the specific processing process, the reconstruction error function and the target prediction error function are fused to form a distinguishing composite error function, which is mainly based on constructing a distinguishing denoising autoencoder network, wherein the distinguishing denoising autoencoder network is the same as the standard denoising autoencoder network from the input layer to the reconstruction layer. The calculation process from the new hidden layer to the output layer is as follows:

[0067]

[0068]

[0069] Among them, newh t is the output of the tth neuron node in the new hidden layer; f still uses the tansig function; h j is the jth input in the ANN; is the threshold in the new hidden layer; is the weight matrix from the new hidden layer to the hidden layer in ANN; is the weight matrix from the output layer to the new hidden layer in the ANN; is the threshold of the output layer; is the true value y of the output layer k The predicted output or estimate of .

[0070]

[0071]

[0072] Where γ is the weight coefficient between the original error function and the target prediction error function; The difference between the target predicted value and the true measured value is described, and then the discriminative denoising autoencoder regression neural network can maintain a balance between the unsupervised reconstruction error and the target prediction.

[0073] In this embodiment, the number of hidden layers is set to 9, the number of input layers is 15, the number of reconstruction layers is 15, and the number of prediction output layers is 2; the hidden layers and reconstruction layers use the tansig function, and the output layer uses the purelin linear function.

[0074] See also Figure 4 , Figure 4 A feature extraction method in one embodiment of the present application is shown, the method comprising:

[0075] S401, acquiring real-time operation data to a feature extraction model.

[0076] S402, obtaining corresponding features by distinguishing hidden layer processing in the autoencoder network.

[0077] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0078] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0079] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0080] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0081] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A feature extraction model training method, characterized in that: The method is applied to the electrolytic aluminum production process, comprising: Obtain sample data to be used for training; The obtaining of sample data to be used for training comprises: obtaining historical sample data contained in electrolytic aluminum process equipment, marking high-frequency data as initial key data, and obtaining two key data in combination with the electrolytic aluminum manufacturing process: DC power consumption per ton of aluminum and current efficiency; setting an ideal power consumption-current correspondence model in the electrolytic aluminum process, and the power consumption-current correspondence model obtains process variables: series current, tank working voltage, AE times, NB times, RC times, AE waiting time, NB interval, AE interval, molecular ratio, aluminum output, aluminum level, electrolyte level, tank temperature, fluoride addition times, and daily fluoride salt consumption; sampling the above data based on the measurement and control system, establishing a process variable set and a target set, and normalizing the process variable set and the target set; Input the sample data into the model to be trained to obtain reconstruction error function information and target prediction error function information respectively; fusing the reconstruction error function and the target prediction error function into a distinguishable composite error function; The model to be trained is trained until the error meets a preset value, and a feature extraction vector is extracted from the sample data, and the feature extraction vector is input into a neural network for training until the error of the feature extraction vector meets a preset value.

2. The feature extraction model training method according to claim 1, characterized in that: The model structure to be trained includes distinguishing a denoising autoencoder network and an ANN neural network.

3. The feature extraction model training method according to claim 2, characterized in that: Input the sample data into the model to be trained to obtain reconstruction error function information and target prediction error function information, including: Inputting the sample data into the autoencoder network to obtain reconstruction error function information; The sample data is input into the ANN neural network after encoding to obtain target prediction error function information.

4. The feature extraction model training method according to claim 2, characterized in that: Training the model to be trained until the error meets a preset value, including: Based on the distinguishable composite error function, a gradient descent algorithm is used to train the distinguishable denoising autoencoder network until the obtained error meets a preset error value.

5. The feature extraction model training method according to claim 4, characterized in that: The distinguishable denoising autoencoder network includes a hidden layer.

6. The feature extraction model training method according to claim 5, characterized in that: Extracting a feature extraction vector from the sample data includes: When the error in the distinguishing denoising autoencoder network meets the preset error value, a feature extraction vector of the input sample data is obtained from the hidden layer.

7. The feature extraction model training method according to claim 2, characterized in that: Inputting the feature extraction vector into a neural network for training until the feature extraction vector error meets a preset value, including: The extracted feature extraction vector is input into the ANN neural network model for training until the feature extraction vector error meets the preset error value.

8. A feature extraction method, characterized in that: include: Based on the feature extraction model obtained in the feature extraction model training method according to any one of claims 1 to 7, specifically The following methods are included: Inputting real-time operation data into the feature extraction model; The corresponding feature expression is obtained by processing the hidden layer in the distinguishable autoencoder network.

9. A feature extraction model training device, characterized in that: The device is used for electrolytic aluminum production, comprising: An acquisition module is used to acquire sample data to be used for training; the acquisition of sample data to be used for training includes: acquiring historical sample data contained in electrolytic aluminum process equipment, marking high-frequency data as initial key data, and acquiring two key data in combination with the electrolytic aluminum manufacturing process: DC power consumption per ton of aluminum and current efficiency; setting an ideal power consumption-current correspondence model in the electrolytic aluminum process, and the power consumption-current correspondence model acquires process variables: series current, tank working voltage, AE times, NB times, RC times, AE waiting time, NB interval, AE interval, molecular ratio, aluminum output, aluminum level, electrolyte level, tank temperature, fluoridation times, and daily fluoride salt consumption; sampling the above data based on the measurement and control system, establishing a process variable set and a target set, and normalizing the process variable set and the target set; A first function acquisition module is used to obtain reconstruction error function information and target prediction error function information; A fusion module, used for fusing the reconstruction error function and the target prediction error function into a distinguishable composite error function; The training module is used to train the model to be trained until the error meets the preset value, and extract the feature extraction vector in the sample data, and input the feature extraction vector into the neural network for training until the error of the feature extraction vector meets the preset value.

10. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; A processor is used to implement the feature extraction model training method described in any one of claims 1 to 8 when running the executable instructions stored in the memory.

Images