Method for predicting load of an electrolytic aluminium plant
By decomposing load data from electrolytic aluminum plants and using multi-model forecasting, the problem of insufficient accuracy in medium- and long-term load forecasting for electrolytic aluminum plants has been solved, achieving high-precision load forecasting and supporting power grid optimization and green transformation of enterprises.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID HUBEI ELECTRIC POWER CO LTD WUHAN POWER SUPPLY CO
- Filing Date
- 2026-03-26
- Publication Date
- 2026-07-07
AI Technical Summary
Existing technologies are insufficient to accurately predict the medium- and long-term loads of electrolytic aluminum plants, resulting in inadequate prediction accuracy and failing to meet the high-precision requirements of the power market and grid dispatch.
A time series decomposition method based on Fourier series and moving average is adopted to decompose the load data into periodic components, trend components and residual components. The Informer model is used to predict the trend component and the AAE-LSTM model is used to predict the residual component. The load forecast results are obtained by superimposing them.
It improves the accuracy of load forecasting for electrolytic aluminum plants, accurately captures long-term trends and random fluctuations, provides more in-depth analysis of the underlying causes of load fluctuations, and supports grid optimization and the green and low-carbon transformation of electrolytic aluminum enterprises.
Smart Images
Figure CN122348501A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of load forecasting technology, and in particular relates to a load forecasting method for an electrolytic aluminum plant. Background Technology
[0002] Currently, the power system places higher demands on the refined management and flexible regulation of high-energy-consuming industrial loads. The electrolytic aluminum industry, as a major global industrial electricity consumer, accounts for approximately 40% of the global industrial load. Accurately predicting the medium- and long-term (monthly and annual) power load of electrolytic aluminum plants is a crucial prerequisite for unlocking demand-side response potential, supporting the grid's participation in ancillary services such as peak shaving and valley filling, and frequency regulation. It is also of great significance for optimizing grid investment layout, ensuring the safe and stable operation of the system under a high proportion of renewable energy grid integration, and promoting the green and low-carbon transformation of the electrolytic aluminum industry itself.
[0003] However, the medium- and long-term load of electrolytic aluminum plants exhibits complex and mixed characteristics: it includes both annual cyclicality and growth trends determined by production plans and industry expansion, and is frequently disturbed by internal random factors such as the anode effect, equipment start-up and shutdown, and raw material quality during the production process. Traditional forecasting methods (such as Fourier series, ARIMA, etc.) or single machine learning models are insufficient to fully capture and decouple these complex characteristics, resulting in limited forecast accuracy and failing to meet the urgent needs of the power market and grid dispatch for high-precision, interpretable load forecasting. Summary of the Invention
[0004] In view of this, embodiments of the present invention provide a method for predicting the load of an electrolytic aluminum plant to solve the problem of insufficient accuracy in the load prediction of electrolytic aluminum plants in the prior art.
[0005] A first aspect of this invention provides a method for predicting the load of an aluminum electrolytic plant, comprising: Obtain historical load data and historical production data for the same period in the electrolytic aluminum plant; A feature matrix is established based on the historical load data and the historical production data of the same period. The historical load data is decomposed into a time series to obtain periodic components, trend components, and residual components. Based on the trend component and the feature matrix, the predicted value of the trend component is determined using a pre-trained Informer model; Based on the residual components and the feature matrix, the predicted values of the residual components are determined using a pre-trained AAE-LSTM model. The predicted values of the periodic component, the trend component, and the residual component are superimposed to obtain the load prediction result of the electrolytic aluminum plant.
[0006] In one possible implementation, the historical production data from the same period includes parameters related to the anode effect during the aluminum electrolysis process.
[0007] In one possible implementation, the parameters related to the anode effect in the aluminum electrolysis process include at least one of the following: occurrence frequency, duration, voltage rise magnitude, and frequency of human intervention.
[0008] In one possible implementation, the step of performing time-series decomposition on the historical load data to obtain periodic components, trend components, and residual components includes: The historical load data is subjected to Fourier series decomposition to extract the periodic components, and the remaining periodic sequence is de-periodized. The trend component and the remaining residual component are extracted from the deperiodic sequence by means of a moving average.
[0009] In one possible implementation, determining the predicted value of the trend component based on the trend component and the feature matrix using a pre-trained Informer model includes: The trend components and the feature matrix are input as vectors into the Informer model; The sparsity measure of the vector is calculated through the ProbSparse self-attention layer; Key features of the vector are extracted through a self-attention distillation layer; The trend component prediction is output through a generative decoder.
[0010] In one possible implementation, determining the predicted values of the residual components based on the residual components and the feature matrix using a pre-trained AAE-LSTM model includes: The residual components and the feature matrix are input as vectors into the AAE network in the AAE-LSTM model, and the reconstructed sequence is output. The reconstructed sequence is input into an LSTM network, which outputs the predicted values of the residual components.
[0011] In one possible implementation, the AAE network is trained using a loss function that minimizes the weighted sum of the reconstruction loss and the adversarial loss.
[0012] A second aspect of the present invention provides a load prediction device for an aluminum electrolytic plant, comprising: The acquisition module is used to acquire historical load data and historical production data for the same period in the electrolytic aluminum plant. The processing module is used to establish a feature matrix based on the historical load data and the historical production data of the same period. The decomposition module is used to perform time series decomposition on the historical load data to obtain periodic components, trend components, and residual components. The trend component prediction module is used to determine the predicted value of the trend component based on the trend component and the feature matrix using a pre-trained Informer model. The residual component prediction module is used to determine the predicted value of the residual component based on the residual component and the feature matrix using a pre-trained AAE-LSTM model. The superposition module is used to superimpose the periodic component, the predicted value of the periodic component, and the predicted value of the residual component to obtain the load prediction result of the electrolytic aluminum plant.
[0013] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the first aspect or any possible implementation thereof described above.
[0014] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the first aspect or any possible implementation thereof described above.
[0015] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: This invention establishes a feature matrix based on historical load data and historical production data from the same period, serving as an auxiliary input for subsequent models. This enables the models to learn and reflect the unique load dynamics of the electrolytic aluminum industry, thereby improving the model's prediction accuracy. Furthermore, the historical load data is decomposed into time series components, yielding periodic components, trend components, and residual components. For the trend components, which reflect the long-term development of the industry, an Informer model is introduced for prediction. This reduces the computational complexity of long-sequence modeling while enhancing the ability to focus on key time steps, thus accurately capturing load growth or decline trends spanning multiple years. For the residual components, which characterize random fluctuations in the production process, an AAE-LSTM ensemble model is designed to achieve high-precision simulation of non-Gaussian and nonlinear random fluctuations. Finally, the periodic components, their predicted values, and the predicted values of the residual components are superimposed to obtain the load prediction results for the electrolytic aluminum plant. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a flowchart illustrating the load forecasting method for electrolytic aluminum plants provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the load prediction device for an electrolytic aluminum plant provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0018] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.
[0019] To illustrate the technical solution described in this invention, specific embodiments are described below.
[0020] Figure 1 This is a schematic diagram illustrating the implementation process of the electrolytic aluminum plant load forecasting method provided in this embodiment of the invention. See also... Figure 1 As shown, this method includes: Step S101: Obtain historical load data and historical production data for the same period of the electrolytic aluminum plant.
[0021] This section collects historical monthly or yearly load data from electrolytic aluminum plants. This includes historical production data from the same period, such as the cell type technology and parameters related to the anode effect during aluminum electrolysis. For example, parameters related to the anode effect during aluminum electrolysis may include, but are not limited to: frequency of occurrence, duration, voltage rise amplitude, and frequency of human intervention.
[0022] Step S102: Establish a feature matrix based on historical load data and historical production data for the same period.
[0023] Step S103: Perform time series decomposition on the historical load data to obtain periodic components, trend components, and residual components.
[0024] The hierarchical decomposition mechanism based on Fourier series and moving average is a hierarchical time series decomposition method that first extracts the period, then analyzes the trend, and finally leaves the residual. The core logic is to use the frequency domain analysis capability of Fourier series to accurately extract multi-scale periodic components, and then use the time domain smoothing characteristics of moving average to separate the trend components, finally obtaining the irregular residual components.
[0025] The specific process includes: (1) Periodic component extraction: Extraction of raw load data Perform Fourier series decomposition to extract harmonic components related to the annual cycle, thus obtaining the periodic component. :
[0026]
[0027] in, The mean of the sequence. 、 These are the Fourier coefficients. The sequence length is given.
[0028] (2) Trend component extraction: Calculate non-periodic components And apply a centralized moving average smoothing method to it to obtain the trend component. :
[0029] in, The moving average window width, This represents the number of valid items within the corresponding window.
[0030] (3) Calculation of residual components: After removing the periodic and trend components from the original load, the residual sequence is obtained. This component represents random fluctuations that are not explained by cycles and trends.
[0031] Step S104: Based on the trend component and the feature matrix, determine the predicted value of the trend component using a pre-trained Informer model.
[0032] For the trend components decomposed from the original load sequence, this invention uses the Informer model for modeling and prediction, and the specific steps are as follows: (1) Data preparation: The trend component sequences obtained through hierarchical decomposition are processed. As training data.
[0033] (2) Construct the Informer model, which mainly includes: ProbSparse Self-Attention Layer: This layer calculates a sparsity metric between the query and the key. Attention is calculated only on the top-u query-key pairs with the highest attention, thus significantly reducing the computational complexity of processing long sequences. The sparsity metric is defined as follows:
[0034] Self-attention distillation layer: Convolutional operations are introduced between multiple attention modules to progressively compress sequence length, extract key features, and effectively reduce the space complexity of the model.
[0035] Generative decoder: This method generates the entire prediction sequence in a single forward propagation, avoiding stepwise recursive prediction and significantly improving the efficiency of long sequence prediction. The decoder input is represented as:
[0036] in, The starting marker sequence, Embed the timestamp into the sequence.
[0037] (3) Model Training and Prediction: The Informer model was trained using historical trend data to obtain an optimized trend prediction model. This model can receive future time features and output the predicted trend components for the corresponding time periods. .
[0038] Step S105: Based on the residual components and the feature matrix, determine the predicted values of the residual components using the pre-trained AAE-LSTM model.
[0039] AAE-LSTM (Adversarial Autoencoder-LSTM) is a hybrid deep learning model that combines an adversarial autoencoder (AAE) and a long short-term memory (LSTM) network. It enhances the latent space representation ability through adversarial training, while using LSTM to capture temporal dependencies.
[0040] The modeling and prediction process of residual components based on AAE-LSTM is as follows: (1) AAE network Historical residual components As input for adversarial autoencoders.
[0041] The AAE structure includes an encoder. decoder And discriminator. The encoder maps the input to latent variables. The decoder performs reconstruction. The discriminator is used to distinguish between the actual residual data and the reconstructed data.
[0042] By jointly minimizing the reconstruction loss and combat losses The AAE is trained to learn and approximate the underlying data distribution of the residual components.
[0043] (2) LSTM network The historical residual components are reconstructed using a trained AAE (Advanced Array Optimizer), yielding a reconstructed sequence containing their distribution characteristics. This reconstructed sequence is then used as training data to construct and train an LSTM (Laser-Based LSTM) network to learn the temporal evolution of the residual components. The AAE and LSTM together constitute an ensemble prediction model for the residual components, used to output predicted residual values for future time periods. .
[0044] Step S106: The predicted values of the periodic component, trend component, and residual component are superimposed to obtain the load prediction result of the electrolytic aluminum plant.
[0045] For any future time Its final load forecast The result is obtained by superimposing the predicted values of the three decomposed components: ; in, These are known periodic components extrapolated based on periodic patterns.
[0046] This embodiment uses root mean square error (RMSE) and mean absolute percentage error (MAPE) as the main evaluation metrics. The performance of this method is compared with mainstream benchmark models such as LSTM, Transformer, and the basic Informer on the test set. The results show that on load data from three electrolytic aluminum enterprises with different technical characteristics (A1 and A2 are prebaked cells, and A3 is a Soderberg cell), the RMSE and MAPE of this method are significantly lower than all the comparison models, demonstrating the best prediction accuracy. Furthermore, the predicted curve of this method has the highest good fit to the actual load curve, further validating its superiority.
[0047] The method provided in this invention first employs a hierarchical decomposition mechanism based on Fourier series and moving averages to decompose the original load sequence into periodic components, trend components, and residual components. This step aims to separate load driving factors at different time scales, laying the foundation for subsequent targeted modeling. For the trend component reflecting the long-term development of the industry, an Informer model is introduced for modeling. Through the ProbSparse self-attention mechanism and self-attention distillation layer, the computational complexity of long sequence modeling is reduced while enhancing the ability to focus on key time steps, thereby accurately capturing load growth or decline trends spanning multiple years. For the residual component characterizing random fluctuations in the production process, an AAE-LSTM ensemble model is designed. The AAE learns the latent distribution of the residual component through adversarial training, generating a reconstructed sequence that more closely reflects the complex characteristics of reality. The LSTM further captures the temporal dependencies in the residuals. Together, they achieve high-precision simulation of non-Gaussian and nonlinear random fluctuations. By constructing a data matrix containing key production parameters as auxiliary input to the model, the model can learn and reflect the unique load dynamic patterns of the electrolytic aluminum industry, thereby significantly improving the model's prediction accuracy.
[0048] Furthermore, in this embodiment, the load forecasting results can be clearly decomposed into three parts: period, trend, and residual. This helps operators understand the underlying causes of load fluctuations and provides deeper insights for production planning, energy dispatch, and market decisions. Meanwhile, high-precision medium- and long-term load forecasting is a crucial prerequisite for electrolytic aluminum loads to participate in the electricity market and implement demand-side response. This method helps improve the grid's capacity to absorb renewable energy, enhances system operational stability, and provides effective technical support for electrolytic aluminum companies to optimize energy efficiency, reduce electricity costs, and minimize their carbon footprint.
[0049] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0050] Figure 2 This is a schematic diagram of the load prediction device for an electrolytic aluminum plant provided in an embodiment of the present invention. Figure 2 As shown, the load forecasting device 2 for an electrolytic aluminum plant includes: Module 21 is used to acquire historical load data and historical production data of the same period in the electrolytic aluminum plant. Processing module 22 is used to establish a feature matrix based on the historical load data and the historical production data of the same period; Decomposition module 23 is used to perform time series decomposition on the historical load data to obtain periodic components, trend components and residual components; The trend component prediction module 24 is used to determine the predicted value of the trend component based on the trend component and the feature matrix using a pre-trained Informer model. The residual component prediction module 25 is used to determine the predicted value of the residual component based on the residual component and the feature matrix using a pre-trained AAE-LSTM model. The superposition module 26 is used to superimpose the periodic component, the predicted value of the periodic component, and the predicted value of the residual component to obtain the load prediction result of the electrolytic aluminum plant.
[0051] In one possible implementation, the historical production data from the same period includes parameters related to the anode effect during the aluminum electrolysis process.
[0052] In one possible implementation, the parameters related to the anode effect in the aluminum electrolysis process include at least one of the following: occurrence frequency, duration, voltage rise magnitude, and frequency of human intervention.
[0053] In one possible implementation, the step of performing time-series decomposition on the historical load data to obtain periodic components, trend components, and residual components includes: The historical load data is subjected to Fourier series decomposition to extract the periodic components, and the remaining periodic sequence is de-periodized. The trend component and the remaining residual component are extracted from the deperiodic sequence by means of a moving average.
[0054] In one possible implementation, determining the predicted value of the trend component based on the trend component and the feature matrix using a pre-trained Informer model includes: The trend components and the feature matrix are input as vectors into the Informer model; The sparsity measure of the vector is calculated through the ProbSparse self-attention layer; Key features of the vector are extracted through a self-attention distillation layer; The trend component prediction is output through a generative decoder.
[0055] In one possible implementation, determining the predicted values of the residual components based on the residual components and the feature matrix using a pre-trained AAE-LSTM model includes: The residual components and the feature matrix are input as vectors into the AAE network in the AAE-LSTM model, and the reconstructed sequence is output. The reconstructed sequence is input into an LSTM network, which outputs the predicted values of the residual components.
[0056] In one possible implementation, the AAE network is trained using a loss function that minimizes the weighted sum of the reconstruction loss and the adversarial loss.
[0057] This invention establishes a feature matrix based on historical load data and historical production data from the same period, serving as an auxiliary input for subsequent models. This enables the models to learn and reflect the unique load dynamics of the electrolytic aluminum industry, thereby improving the model's prediction accuracy. Furthermore, the historical load data is decomposed into time series components, yielding periodic components, trend components, and residual components. For the trend components, which reflect the long-term development of the industry, an Informer model is introduced for prediction. This reduces the computational complexity of long-sequence modeling while enhancing the ability to focus on key time steps, thus accurately capturing load growth or decline trends spanning multiple years. For the residual components, which characterize random fluctuations in the production process, an AAE-LSTM ensemble model is designed to achieve high-precision simulation of non-Gaussian and nonlinear random fluctuations. Finally, the periodic components, their predicted values, and the predicted values of the residual components are superimposed to obtain the load prediction results for the electrolytic aluminum plant.
[0058] Figure 3 This is a schematic diagram of an electronic device 30 provided in an embodiment of the present invention. For example... Figure 3 As shown, the electronic device 30 of this embodiment includes: a processor 31, a memory 32, and a computer program 33 stored in the memory 32 and executable on the processor 31. When the processor 31 executes the computer program 33, it implements the steps in the various method embodiments described above. Alternatively, when the processor 31 executes the computer program 33, it implements the functions of each module in the various device embodiments described above.
[0059] For example, the computer program 33 may be divided into one or more modules / units, which are stored in the memory 32 and executed by the processor 31 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 33 in the electronic device 30.
[0060] The electronic device 30 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. The electronic device 30 may include, but is not limited to, a processor 31 and a memory 32. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 30 and does not constitute a limitation on electronic device 30. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 30 may also include input / output devices, network access devices, buses, etc.
[0061] The processor 31 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.
[0062] The memory 32 can be an internal storage unit of the electronic device 30, such as a hard disk or memory of the electronic device 30. The memory 32 can also be an external storage device of the electronic device 30, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 30. Furthermore, the memory 32 can include both internal and external storage units of the electronic device 30. The memory 32 is used to store the computer program and other programs and data required by the electronic device 30. The memory 32 can also be used to temporarily store data that has been output or will be output.
[0063] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0064] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0065] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0066] In the embodiments provided by this invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.
[0067] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0068] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0069] If the integrated module / unit is implemented as 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, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.
[0070] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A method for predicting the load of an electrolytic aluminum plant, characterized in that, include: Obtain historical load data and historical production data for the same period in the electrolytic aluminum plant; A feature matrix is established based on the historical load data and the historical production data of the same period. The historical load data is decomposed into a time series to obtain periodic components, trend components, and residual components. Based on the trend component and the feature matrix, the predicted value of the trend component is determined using a pre-trained Informer model; Based on the residual components and the feature matrix, the predicted values of the residual components are determined using a pre-trained AAE-LSTM model. The predicted values of the periodic component, the trend component, and the residual component are superimposed to obtain the load prediction result of the electrolytic aluminum plant.
2. The method for predicting the load of an electrolytic aluminum plant as described in claim 1, characterized in that, The historical production data from the same period includes parameters related to the anode effect during the electrolytic aluminum process.
3. The method for predicting the load of an electrolytic aluminum plant as described in claim 2, characterized in that, The parameters related to the anode effect in the electrolytic aluminum process include at least one of the following: occurrence frequency, duration, voltage rise amplitude, and frequency of human intervention.
4. The method for predicting the load of an electrolytic aluminum plant as described in any one of claims 1 to 3, characterized in that, The process of performing time series decomposition on the historical load data to obtain periodic components, trend components, and residual components includes: The historical load data is subjected to Fourier series decomposition to extract the periodic components, and the remaining periodic sequence is de-periodized. The trend component and the remaining residual component are extracted from the deperiodic sequence by means of a moving average.
5. The method for predicting the load of an electrolytic aluminum plant as described in any one of claims 1 to 3, characterized in that, The step of determining the predicted value of the trend component based on the trend component and the feature matrix using a pre-trained Informer model includes: The trend components and the feature matrix are input as vectors into the Informer model; The sparsity measure of the vector is calculated through the ProbSparse self-attention layer; Key features of the vector are extracted through a self-attention distillation layer; The trend component prediction is output through a generative decoder.
6. The method for predicting the load of an electrolytic aluminum plant as described in any one of claims 1 to 3, characterized in that, The step of determining the predicted values of the residual components based on the residual components and the feature matrix using a pre-trained AAE-LSTM model includes: The residual components and the feature matrix are input as vectors into the AAE network in the AAE-LSTM model, and the reconstructed sequence is output. The reconstructed sequence is input into an LSTM network, which outputs the predicted values of the residual components.
7. The method for predicting the load of an electrolytic aluminum plant as described in claim 6, characterized in that, During training, the AAE network uses the weighted sum of the reconstruction loss and the adversarial loss as its loss function.
8. A load prediction device for an electrolytic aluminum plant, characterized in that, include: The acquisition module is used to acquire historical load data and historical production data for the same period in the electrolytic aluminum plant. The processing module is used to establish a feature matrix based on the historical load data and the historical production data of the same period. The decomposition module is used to perform time series decomposition on the historical load data to obtain periodic components, trend components, and residual components. The trend component prediction module is used to determine the predicted value of the trend component based on the trend component and the feature matrix using a pre-trained Informer model. The residual component prediction module is used to determine the predicted value of the residual component based on the residual component and the feature matrix using a pre-trained AAE-LSTM model. The superposition module is used to superimpose the periodic component, the predicted value of the periodic component, and the predicted value of the residual component to obtain the load prediction result of the electrolytic aluminum plant.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.