Data compression and reconstruction method and system for thermal power plant

Abstract

The invention discloses a data compression and reconstruction method and system for a thermal power plant, and belongs to the field of data management, and the method comprises the steps: determining thermal equipment of the thermal power plant, obtaining the historical data of the thermal equipment, and constructing a VAE model based on the historical data; interacting the thermal equipment, acquiring real-time data of the thermal equipment, inputting the real-time data into the VAE model for data compression, acquiring compressed data, and storing the compressed data into a database; and when the data of the thermal equipment needs to be checked, reading the compressed data from the database, and inputting the compressed data into the VAE model for data reconstruction to obtain reconstructed data. According to the method, the technical problems of high cost and low efficiency of mass data storage of the thermal equipment of the thermal power plant in the prior art are solved, and the technical effects of effectively compressing the data of the thermal equipment of the thermal power plant, reducing the storage cost and conveniently reconstructing and checking the data are achieved while the data quality is ensured.

Description

Technical Field

[0001] The invention relates to the field of data management, and in particular to a data compression and reconstruction method and system for a thermal power plant. Background Art

[0002] Thermal power plants contain many thermal equipment, such as condensers, coal mills, induced draft fans, and forced draft fans. In order to ensure the safe and stable operation of these equipment, real-time monitoring and data collection are required. With the development of monitoring technology and the increase in sampling frequency, thermal power plants generate massive amounts of equipment operation data every day. However, as time goes by, the storage and management of these massive data has gradually become a problem. Traditional data storage methods face challenges such as high storage costs and low retrieval efficiency. At the same time, due to the particularity of the power industry, these data often need to be stored for a long time for analysis and backtracking, further exacerbating storage pressure. At present, some common data compression methods can reduce storage space to a certain extent, but often lead to a decrease in data quality and cannot meet the high requirements of the power industry for data accuracy. In addition, these methods often have poor results when reconstructing data, affecting the availability and analysis value of the data. Therefore, the storage cost of massive data of thermal equipment in thermal power plants in the prior art is high and the efficiency is low. Summary of the invention

[0003] The present application provides a data compression and reconstruction method and system for a thermal power plant, aiming to solve the technical problems of high cost and low efficiency in storing massive data of thermal equipment in a thermal power plant in the prior art.

[0004] In view of the above problems, the present application provides a data compression and reconstruction method and system for a thermal power plant.

[0005] The first aspect disclosed in the present application provides a data compression and reconstruction method for a thermal power plant, the method comprising: determining the thermal equipment in the thermal power plant, obtaining historical data of the thermal equipment, and constructing a VAE model based on the historical data; interacting with the thermal equipment, obtaining real-time data of the thermal equipment, inputting the real-time data into the VAE model for data compression, obtaining compressed data, and storing it in a database; when it is necessary to view the data of the thermal equipment, reading the compressed data from the database, inputting the data into the VAE model for data reconstruction, and obtaining reconstructed data.

[0006] Another aspect disclosed in the present application provides a data compression and reconstruction system for a thermal power plant, the system comprising: a model building module, used to determine the thermal equipment in the thermal power plant, obtain historical data of the thermal equipment, and build a VAE model based on the historical data; a data compression module, used to interact with the thermal equipment, obtain real-time data of the thermal equipment, input the real-time data into the VAE model for data compression, obtain compressed data, and store it in a database; a data reconstruction module, used to read compressed data from the database when it is necessary to view the data of the thermal equipment, input the compressed data into the VAE model for data reconstruction, and obtain reconstructed data.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The thermal equipment of the thermal power plant, such as condenser, coal mill, induced draft fan or forced draft fan, etc., is determined, and the historical data of the thermal equipment is obtained, which provides basic data for the subsequent model construction. Based on the historical data, the VAE model is constructed, which specifically includes: preprocessing the historical data, including data cleaning and normalization, to improve the data quality and unify the data scale; using the KPCA method to reduce the dimension of the training data and determine the hidden layer dimension of the VAE model, which helps to reduce data redundancy and extract key features; initializing the VAE model parameters and using the training data for model training; optimizing the VAE model parameters by minimizing the loss function including reconstruction loss and KL divergence, ensuring the performance of the model in data compression and reconstruction. Interact with the thermal equipment, obtain real-time data, and input it into the trained VAE model for data compression, obtain compressed data and store it in the database, realizing efficient compression and storage of real-time data. When the data of the thermal equipment needs to be viewed, the compressed data is read from the database, input into the VAE model again for data reconstruction, and obtain the reconstructed data, realizing the rapid restoration of compressed data, which is convenient for analysis and viewing. Among them, the VAE model consists of two parts: an encoder and a decoder. The encoder compresses the input data into a latent space of lower dimension to achieve data compression; the decoder restores the compressed data in the latent space to the original dimension to achieve a technical solution for data reconstruction. It solves the technical problems of high cost and low efficiency in storing massive data of thermal equipment in thermal power plants in the prior art, and achieves the technical effect of effectively compressing the data of thermal equipment in thermal power plants, reducing storage costs, and facilitating data reconstruction and viewing while ensuring data quality.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flow chart of a data compression and reconstruction method for a thermal power plant;

[0011] Figure 2 Schematic diagram of data collection variables for condenser equipment;

[0012] Figure 3 This is a schematic diagram of the KPCA dimensionality reduction result;

[0013] Figure 4 Schematic diagram of the VAE model structure;

[0014] Figure 5 Schematic diagram of VAE model parameter configuration;

[0015] Figure 6 Schematic diagram for comparing the reconstruction effects of VAE models;

[0016] Figure 7 The figure is a structural diagram of a data compression and reconstruction system for a thermal power plant.

[0017] Explanation of the accompanying drawings: model building module 11, data compression module 12, data reconstruction module 13. DETAILED DESCRIPTION

[0018] The overall idea of ​​the technical solution provided by this application is as follows:

[0019] The embodiments of the present application provide a data compression and reconstruction method and system for a thermal power plant, which utilizes a deep learning algorithm combined with dimensionality reduction technology to achieve efficient compression storage and high-quality reconstruction of massive data of thermal equipment in a thermal power plant, thereby solving the problems of high storage cost and low efficiency faced by traditional data storage methods, while ensuring the availability and analytical value of the data.

[0020] First, identify the target thermal equipment and obtain its historical data to lay the foundation for subsequent model construction. Then, through data preprocessing, KPCA dimensionality reduction and other steps, build and optimize the VAE model. The introduction of the KPCA method effectively determines the hidden layer dimension of the VAE model and improves the efficiency and performance of the model. Next, use the trained VAE model to compress the real-time collected equipment data, and store the compressed data in the database, reducing storage requirements. When the original data needs to be viewed, the data is reconstructed through the decoder of the VAE model to achieve rapid data restoration and viewing. Throughout the process, the encoder and decoder of the VAE model respectively undertake the important tasks of data compression and reconstruction to ensure the quality of the data during compression and reconstruction.

[0021] Through the combination of VAE model and KPCA method, efficient data compression and convenient reconstruction are achieved while ensuring data quality, which solves the data storage problem faced by thermal power plants and provides support for the long-term preservation and analytical utilization of data.

[0022] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0023] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides a data compression and reconstruction method for a thermal power plant, the method comprising:

[0024] S1: Determine the thermal equipment of the thermal power plant, obtain historical data of the thermal equipment, and build a VAE model based on the historical data.

[0025] Specifically, first, determine the thermal equipment that needs to be processed. Thermal equipment is the key equipment related to the thermal cycle in a thermal power plant, such as a condenser, a coal mill, an induced draft fan or a forced draft fan. Then, obtain the historical data of the selected thermal equipment from the SIS (station-level monitoring information system) of the thermal power unit. Preferably, three months of historical data are collected with a sampling frequency of 1 minute to ensure the comprehensiveness and representativeness of the data. These historical data record the various operating parameters of the thermal equipment at different time points, providing a basis for subsequent model construction. Afterwards, a VAE (variational autoencoder) model is constructed based on the acquired historical data. VAE is a generative model based on deep learning, consisting of two parts: an encoder and a decoder. The encoder compresses the input data into a low-dimensional latent space, while the decoder reconstructs the data in the latent space back to the original dimension. The use of the VAE model enables this method to achieve effective data compression and accurate data reconstruction while ensuring data quality.

[0026] S2: Interact with the thermal equipment to obtain real-time data of the thermal equipment, input the real-time data into the VAE model for data compression, obtain compressed data, and store it in a database.

[0027] Specifically, first, by interacting with the determined thermal equipment in real time, the real-time data of the equipment during its operation is obtained. These real-time data reflect the current operating status and various parameters of the thermal equipment. Then, the real-time data is input into the constructed VAE model. Specifically, the real-time data passes through the encoder part of the VAE model, and the encoder compresses the high-dimensional input data into a low-dimensional latent space, realizing effective data compression, significantly reducing the storage space requirements of the data, and retaining the key feature information of the data. Then, the data compressed by the VAE model is obtained, that is, the compressed data. These compressed data contain the main information of the original real-time data, but have a smaller data volume. After that, the obtained compressed data is stored in the database, realizing the efficient storage of the real-time data of the thermal equipment. Since the compressed data is stored, the data storage cost is greatly reduced and the storage efficiency is improved. At the same time, since the VAE model is used for compression, the quality and reconstructibility of the stored data are guaranteed.

[0028] Through the efficient compression and storage of real-time data of thermal equipment, a foundation is provided for subsequent data analysis and equipment monitoring, while effectively solving the cost and efficiency problems caused by massive data storage.

[0029] S3: When the data of the thermal equipment needs to be viewed, the compressed data is read from the database, input into the VAE model for data reconstruction, and the reconstructed data is obtained.

[0030] Specifically, when the original data of the thermal equipment needs to be viewed, the corresponding compressed data is read from the database. Then, the compressed data is input into the VAE model for data reconstruction. Specifically, the compressed data is processed by the decoder part of the VAE model. The function of the decoder is to reconstruct the compressed data in the low-dimensional latent space back to the original high-dimensional space, that is, to restore the compressed data to a form that approximates the original data. Then, through the decoding process of the VAE model, the reconstructed data is obtained. After that, the obtained reconstructed data can be viewed and analyzed by relevant personnel.

[0031] By quickly reconstructing and viewing compressed stored data, we can save storage space while ensuring the accessibility and practicality of the data. This not only meets the needs of long-term data storage, but also ensures that the data can be easily viewed and analyzed when needed, effectively balancing the relationship between storage efficiency and data availability.

[0032] Furthermore, based on the historical data, a VAE model is constructed, including:

[0033] Performing data preprocessing on the historical data to obtain training data;

[0034] Using the KPCA method to reduce the dimension of the training data and determine the hidden layer dimension of the VAE model;

[0035] Initializing the parameters of the VAE model in combination with the hidden layer dimensions, and performing model training using the training data;

[0036] The parameters of the VAE model are optimized by minimizing the loss function to complete the construction of the VAE model.

[0037] In a feasible implementation, first, the original historical data obtained from the SIS system is cleaned and standardized. Data cleaning includes removing outliers and processing missing values ​​to improve data quality. Standardization unifies the data to the same scale through methods such as normalization, which is conducive to subsequent model training. Through preprocessing, a data set with higher quality and more suitable for model training is obtained. Then, the KPCA method is used to reduce the dimension of the training data and determine the hidden layer dimension of the VAE model. KPCA (kernel principal component analysis) is a nonlinear dimensionality reduction method that can effectively extract the main features of the data. The characteristic distribution of the training data is analyzed by KPCA to determine the hidden layer dimension, which provides a basis for the structural design of the VAE model. Subsequently, based on the determined hidden layer dimension, the network structure and parameters of the VAE model are designed and initialized. Next, the VAE model is trained using the preprocessed training data so that the model gradually learns the intrinsic characteristics and distribution of the data. After that, the parameters of the VAE model are optimized by minimizing the loss function to complete the construction of the VAE model. The loss function of the VAE model includes two parts: reconstruction error and KL divergence. Through iterative optimization, the loss function is minimized to achieve the best performance in data compression and reconstruction. When the loss function converges to the preset threshold, the construction of the VAE model is completed.

[0038] By constructing a VAE model based on historical data, the foundation is laid for subsequent data compression and reconstruction, and an efficient VAE model suitable for data processing of specific thermal equipment is obtained.

[0039] Further, the historical data is preprocessed, including:

[0040] Performing data cleaning on the historical data to remove outliers and missing values;

[0041] Performing data normalization processing on the cleaned historical data through a normalization formula to obtain normalized historical data;

[0042] Wherein, the normalization formula is: x is the original data, x * To normalize the historical data, max(x) and min(x) are the maximum and minimum values ​​of the historical data, respectively.

[0043] In a feasible implementation, the historical data is preprocessed. First, outliers in the historical data are identified, such as data points that are obviously deviated from the normal range, and they are deleted or replaced with reasonable values. At the same time, for missing values ​​in the data set, interpolation, mean filling and other methods are used according to the specific situation to improve data quality and reduce interference factors in subsequent model training. Then, the cleaned historical data is normalized by the normalization formula to obtain normalized historical data. Among them, the normalization formula is: Among them, is the original data, x * To normalize historical data, max(x) and min(x) are the maximum and minimum values ​​of the historical data, respectively. Through this formula, data of different scales are uniformly mapped to the interval [0, 1]. Normalization helps to eliminate the dimensional differences between different variables, so that each feature has the same weight in subsequent model training, thereby improving the training effect and generalization ability of the model.

[0044] By adopting the above data preprocessing, the quality and consistency of historical data can be significantly improved, creating favorable conditions for subsequent KPCA dimensionality reduction and VAE model training, improving the efficiency of model training, and enhancing the performance of the final VAE model in data compression and reconstruction.

[0045] Furthermore, the KPCA method is used to reduce the dimension of the training data to determine the hidden layer dimension of the VAE model, including:

[0046] Mapping the training data to a high-dimensional feature space, and calculating a covariance matrix of the training data in the high-dimensional feature space;

[0047] Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, and arranging the eigenvectors in descending order according to the magnitude of the eigenvalues;

[0048] According to the descending order results, the first n eigenvectors are selected and used as the principal components of KPCA, where n is the hidden layer dimension of the VAE model.

[0049] In a preferred embodiment, first, the training data is mapped from the original space to the high-dimensional feature space through the kernel function, so as to capture the nonlinear characteristics of the data, which is conducive to the subsequent dimensionality reduction analysis. In the high-dimensional feature space, the covariance matrix of the mapped data is calculated, which reflects the distribution and correlation of the data in each dimension. Then, a series of eigenvalues ​​and their corresponding eigenvectors are obtained by performing eigenvalue decomposition on the covariance matrix. The size of the eigenvalue reflects the importance of the corresponding eigenvector in the data change. The eigenvectors are arranged in descending order according to the size of the corresponding eigenvalues, so that the important features are in front. Subsequently, according to the descending order of the eigenvalues, the first n most important eigenvectors are selected as principal components. These n principal components constitute the main features of the data, and the value of n also determines the hidden layer dimension of the VAE model.

[0050] Through the KPCA dimensionality reduction process, the main features of the training data can be effectively extracted and the appropriate hidden layer dimensions of the VAE model can be determined. This not only reduces the redundancy of the data, but also retains the key information of the data, providing an important basis for the subsequent construction and training of the VAE model.

[0051] Furthermore, the VAE model includes an encoder and a decoder, including:

[0052] Configuring the VAE model according to the hidden layer dimension, wherein the hidden layer dimension is smaller than the input data dimension of the VAE model;

[0053] When the VAE model receives the real-time data, the encoder compresses the real-time data to the hidden layer dimension to obtain compressed data;

[0054] When the data of the thermal equipment needs to be viewed, the compressed data is restored from the hidden layer dimension to the original input data dimension through the decoder to obtain reconstructed data.

[0055] In a feasible implementation, first, the network structure of the VAE model is constructed based on the hidden layer dimension n determined by KPCA dimensionality reduction. The input layer dimension of the VAE model is the same as the original data dimension, while the hidden layer dimension is set to n, and n is smaller than the input dimension. This structural design ensures that the data can be effectively compressed during the encoding process. After the VAE model receives the real-time data, it is input into the encoder part of the VAE model. The encoder maps the high-dimensional input data to the low-dimensional latent space through a series of nonlinear transformations to achieve data compression. The dimension of the compressed data is the same as the set hidden layer dimension n, which greatly reduces the storage space requirement for the data. When the data of the thermal equipment needs to be viewed, the stored compressed data is input into the decoder part of the VAE model. The decoder remaps the data in the low-dimensional latent space to the original high-dimensional space through the opposite process of the encoder, thereby reconstructing a form that approximates the original data. This reconstruction process enables the compressed and stored data to be easily viewed and analyzed.

[0056] By utilizing the encoder and decoder structure of the VAE model, efficient data compression and accurate reconstruction are achieved, which not only significantly reduces the data storage space but also ensures the quality of data reconstruction, providing an efficient and reliable solution for data management of thermal equipment in thermal power plants.

[0057] Furthermore, the loss function includes reconstruction loss and KL divergence, and the calculation formula is:

[0058]

[0059] in, Represents the reconstruction loss, which is used to measure the reconstruction output of the VAE model The difference between the input data x, ∑ j KL(q j (z|x)||p(z)) represents the KL divergence, which is used to measure the latent space distribution q learned by the encoder of the VAE model. j The difference between (z|x) and the true distribution p(z).

[0060] In a preferred embodiment, the training process of the VAE model involves minimizing the loss function. The loss function includes two key parts: reconstruction loss and KL divergence. Specifically, the calculation formula of the loss function is as follows:

[0061]

[0062] in, Represents the reconstruction loss, which is used to measure the reconstruction output of the VAE model The difference between the original input data x. The reconstruction loss reflects the performance of the model in data reconstruction. The smaller its value, the higher the accuracy of the model reconstruction. In practical applications, methods such as mean square error or cross entropy can be selected to calculate the reconstruction loss. j KL(q j (z|x)||p(z)) represents the KL divergence. This term is used to measure the latent space distribution q learned by the encoder of the VAE model. j The difference between (z|x) and the preset true distribution p(x). The introduction of KL divergence ensures that the hidden layer representation generated by the encoder follows the predetermined distribution (standard normal distribution), which helps to improve the generation ability and generalization performance of the model.

[0063] By minimizing the above loss function, the VAE model can strike a balance between data compression and reconstruction. The reconstruction loss term encourages the model to reconstruct the input data as accurately as possible, while the KL divergence term normalizes the distribution of the latent space, prevents overfitting and enhances the model's generation ability. The design of this loss function enables the VAE model to not only effectively compress data, but also ensure the quality and diversity of the reconstructed data. During the model training process, the model parameters are continuously adjusted through the backpropagation algorithm and optimizer (such as stochastic gradient descent) to minimize the loss function. When the loss function converges to a stable low value, it can be considered that the VAE model has completed training and can effectively perform data compression and reconstruction tasks.

[0064] Furthermore, the thermal equipment is a condenser, a coal mill, an induced draft fan or a forced draft fan.

[0065] In a feasible implementation, the thermal equipment includes but is not limited to a condenser, a coal mill, an induced draft fan or a forced draft fan. The condenser is an important equipment in the steam turbine system of a thermal power plant. Its main function is to condense the exhaust steam discharged from the steam turbine into water to maintain the vacuum of the steam turbine and improve thermal efficiency. The coal mill is used to crush the raw coal into fine powder for full combustion in the boiler. The induced draft fan is mainly used to extract the flue gas in the boiler, control the negative pressure in the furnace, and ensure the normal operation of the boiler. The forced draft fan is responsible for delivering the air required for combustion to the boiler, which is crucial to maintaining the normal combustion of the boiler.

[0066] These thermal equipment will generate a large amount of real-time data during operation, such as temperature, pressure, flow, speed, etc. These data are of great value in monitoring the operating status of equipment, optimizing operating parameters, predicting failures, and performing preventive maintenance. In view of the characteristics of these thermal equipment, the VAE model can effectively compress the massive data generated by these equipment, significantly reduce storage costs, and ensure the accuracy of data reconstruction, providing reliable data support for equipment monitoring, fault diagnosis, and optimization control.

[0067] Next, taking the condenser as an example, the embodiment of the present application is further described:

[0068] In a specific embodiment, the condenser equipment of unit 01 of a thermal power plant is taken as the research object, and the operation data of the equipment is collected for three months from June 1, 2023 to August 31, 2023. The collected variables are as follows: Figure 2 As shown in the figure, 22 key parameters are included, including condenser vacuum, circulating water outlet temperature, circulating water inlet temperature, etc. The sampling interval is set to 1 minute, and a total of 132,480 data samples are obtained as historical data. The historical data is preprocessed to obtain training data, which provides a sufficient data basis for subsequent model training and verification.

[0069] First, the KPCA method was used to reduce the dimension of 80% of the training data. By analyzing the variance explanation ratio, the number of principal components when the variance contribution rate reached 90% was selected and determined to be 6. These 6 principal components were used as the hidden layer dimensions in the VAE model, achieving a data compression ratio of 3.7:1. The results of KPCA dimensionality reduction are shown in Figure 2. Figure 3 As shown in Figure 1, we can intuitively observe the distribution characteristics of the data after dimensionality reduction. Then, we build a VAE model, whose structure is as follows: Figure 4 As shown in the figure. The input layer dimension of the model is 22, which corresponds to the number of features of the original data. Two hidden layers are set with dimensions of 49 and 42 respectively to enhance the expressiveness of the model. The hidden layer dimension is set to 6, which is consistent with the KPCA dimensionality reduction result. The detailed parameter configuration of the VAE model is shown in the figure. Figure 5 As shown in the figure, it includes key information such as activation functions of each layer and optimizer selection. In order to intuitively show the reconstruction effect of the VAE model, five important indicators, namely condenser vacuum, circulating water outlet temperature, circulating water inlet temperature, condenser hot well liquid level and condensate pump current, are selected for visualization. The reconstruction results are shown in the figure. Figure 6 As shown in the figure, it can be clearly seen that the data reconstructed by the VAE model is highly consistent with the original data, which fully demonstrates the excellent performance of this method in data compression and reconstruction.

[0070] In summary, the data compression and reconstruction method for a thermal power plant provided in the embodiment of the present application has the following technical effects:

[0071] Determine the thermal equipment of the thermal power plant, obtain the historical data of the thermal equipment, build the VAE model based on the historical data, determine the data processing object, provide the basic data required for model training, and complete the construction of the VAE model to ensure that the subsequent steps have reliable model support. Interact with thermal equipment, obtain the real-time data of the thermal equipment, input the real-time data into the VAE model for data compression, obtain the compressed data, and store it in the database, so as to realize the collection and compression storage of real-time data. By using the constructed VAE model, the data storage space is effectively reduced, the storage cost is reduced, and the real-time nature of the data is guaranteed. When it is necessary to view the data of the thermal equipment, read the compressed data from the database, input the VAE model for data reconstruction, obtain the reconstructed data, and realize the rapid restoration of the compressed data, so as to ensure that the original data can be easily viewed when needed, and the availability and analysis value of the data are guaranteed, while the advantages brought by the compressed storage are not affected, so as to achieve the technical effect of effectively compressing the thermal equipment data of the thermal power plant, reducing the storage cost, and conveniently reconstructing and viewing the data while ensuring the data quality.

[0072] Embodiment 2 is based on the same inventive concept as the data compression and reconstruction method for thermal power plants in the above embodiment. Figure 7 As shown, the embodiment of the present application provides a data compression and reconstruction system for a thermal power plant, the system comprising:

[0073] The model building module 11 is used to determine the thermal equipment of the thermal power plant, obtain the historical data of the thermal equipment, and build a VAE model based on the historical data;

[0074] A data compression module 12 is used to interact with the thermal equipment, obtain real-time data of the thermal equipment, input the real-time data into the VAE model for data compression, obtain compressed data, and store it in a database;

[0075] The data reconstruction module 13 is used to read the compressed data from the database when it is necessary to view the data of the thermal equipment, input the compressed data into the VAE model to reconstruct the data, and obtain the reconstructed data.

[0076] Furthermore, the model building module 11 includes the following execution steps:

[0077] Performing data preprocessing on the historical data to obtain training data;

[0078] Using the KPCA method to reduce the dimension of the training data and determine the hidden layer dimension of the VAE model;

[0079] Initializing the parameters of the VAE model in combination with the hidden layer dimensions, and performing model training using the training data;

[0080] The parameters of the VAE model are optimized by minimizing the loss function to complete the construction of the VAE model.

[0081] Furthermore, the model building module 11 also includes the following execution steps:

[0082] Performing data cleaning on the historical data to remove outliers and missing values;

[0083] Performing data normalization processing on the cleaned historical data through a normalization formula to obtain normalized historical data;

[0084] Wherein, the normalization formula is: x is the original data, x * To normalize the historical data, max(x) and min(x) are the maximum and minimum values ​​of the historical data, respectively.

[0085] Furthermore, the model building module 11 also includes the following execution steps:

[0086] Mapping the training data to a high-dimensional feature space, and calculating a covariance matrix of the training data in the high-dimensional feature space;

[0087] Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, and arranging the eigenvectors in descending order according to the magnitude of the eigenvalues;

[0088] According to the descending order results, the first n eigenvectors are selected and used as the principal components of KPCA, where n is the hidden layer dimension of the VAE model.

[0089] Furthermore, the model building module 11 also includes the following execution steps:

[0090] Configuring the VAE model according to the hidden layer dimension, wherein the hidden layer dimension is smaller than the input data dimension of the VAE model;

[0091] When the VAE model receives the real-time data, the encoder compresses the real-time data to the hidden layer dimension to obtain compressed data;

[0092] When the data of the thermal equipment needs to be viewed, the compressed data is restored from the hidden layer dimension to the original input data dimension through the decoder to obtain reconstructed data.

[0093] Furthermore, the model building module 11 also includes the following execution steps:

[0094]

[0095] in, Represents the reconstruction loss, which is used to measure the reconstruction output of the VAE model The difference between the input data x, ∑ j KL(q j (z|x)||p(z)) represents the KL divergence, which is used to measure the latent space distribution q learned by the encoder of the VAE model. j The difference between (z|x) and the true distribution p(z).

[0096] Furthermore, the thermal equipment is a condenser, a coal mill, an induced draft fan or a forced draft fan.

[0097] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0098] Furthermore, the first or second mentioned above may not only represent an order relationship, but may also represent a specific concept, and / or refer to multiple elements that can be selected individually or in full. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these changes and variations.

Claims

1. A data compression and reconstruction method for a thermal power plant, characterized in that: include: Determine the thermal equipment of the thermal power plant, obtain historical data of the thermal equipment, and build a VAE model based on the historical data; Interact with the thermal equipment, obtain real-time data of the thermal equipment, input the real-time data into the VAE model for data compression, obtain compressed data, and store it in a database; When the data of the thermal equipment needs to be viewed, the compressed data is read from the database, input into the VAE model for data reconstruction, and the reconstructed data is obtained.

2. The data compression and reconstruction method for a thermal power plant according to claim 1, characterized in that: Based on the historical data, a VAE model is constructed, including: Performing data preprocessing on the historical data to obtain training data; Using the KPCA method to reduce the dimension of the training data and determine the hidden layer dimension of the VAE model; Initializing the parameters of the VAE model in combination with the hidden layer dimensions, and performing model training using the training data; The parameters of the VAE model are optimized by minimizing the loss function to complete the construction of the VAE model.

3. The data compression and reconstruction method for a thermal power plant according to claim 2, characterized in that: The historical data is preprocessed, including: Performing data cleaning on the historical data to remove outliers and missing values; Performing data normalization processing on the cleaned historical data through a normalization formula to obtain normalized historical data; Wherein, the normalization formula is: x is the original data, x * To normalize the historical data, max(x) and min(x) are the maximum and minimum values ​​of the historical data, respectively.

4. The data compression and reconstruction method for a thermal power plant according to claim 2, characterized in that: The KPCA method is used to reduce the dimension of the training data to determine the hidden layer dimension of the VAE model, including: Mapping the training data to a high-dimensional feature space, and calculating a covariance matrix of the training data in the high-dimensional feature space; Performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, and arranging the eigenvectors in descending order according to the magnitude of the eigenvalues; According to the descending order results, the first n eigenvectors are selected and used as the principal components of KPCA, where n is the hidden layer dimension of the VAE model.

5. The data compression and reconstruction method for a thermal power plant according to claim 4, characterized in that: The VAE model includes an encoder and a decoder, including: Configuring the VAE model according to the hidden layer dimension, wherein the hidden layer dimension is smaller than the input data dimension of the VAE model; When the VAE model receives the real-time data, the encoder compresses the real-time data to the hidden layer dimension to obtain compressed data; When the data of the thermal equipment needs to be viewed, the compressed data is restored from the hidden layer dimension to the original input data dimension through the decoder to obtain reconstructed data.

6. The data compression and reconstruction method for a thermal power plant according to claim 2, characterized in that: The loss function includes reconstruction loss and KL divergence, and the calculation formula is: in, Represents the reconstruction loss, which is used to measure the reconstruction output of the VAE model The difference between the input data x, ∑ j KL(q j (z|x)||p(z)) represents the KL divergence, which is used to measure the latent space distribution q learned by the encoder of the VAE model. j The difference between (z|x) and the true distribution p(z).

7. The data compression and reconstruction method for a thermal power plant according to claim 1, characterized in that: The thermal equipment is a condenser, a coal mill, an induced draft fan or a forced draft fan.

8. A data compression and reconstruction system for a thermal power plant, characterized in that: The method for data compression and reconstruction for a thermal power plant according to any one of claims 1 to 7 comprises: A model building module, the model building module is used to determine the thermal equipment of the thermal power plant, obtain the historical data of the thermal equipment, and build a VAE model based on the historical data; A data compression module, the data compression module is used to interact with the thermal equipment, obtain real-time data of the thermal equipment, input the real-time data into the VAE model for data compression, obtain compressed data, and store it in a database; A data reconstruction module is used to read the compressed data from the database when it is necessary to view the data of the thermal equipment, input the compressed data into the VAE model for data reconstruction, and obtain the reconstructed data.

Images