Method and device for classifying carbon emission data of gas turbine and electronic equipment

Abstract

The invention discloses a gas turbine carbon emission data classification method and device and electronic equipment, and the method comprises the steps: constructing an EuSN reserve pool according to candidate hyper-parameter vectors; inputting gas turbine sensor data into the EuSN reserve pool, and obtaining a final hidden state of the EuSN reserve pool as a dynamic feature vector; according to gas turbine sensor data, global statistics of independent variables of all sensors are calculated to obtain static feature vectors; splicing the dynamic feature vector and the static feature vector to obtain a mixed feature vector; according to the mixed feature vector, calculating a classification objective function value on the verification set, and iteratively updating the candidate hyper-parameter vector through a CMA-ES optimization module to obtain an optimal hyper-parameter vector; an optimal EuSN reserve pool is determined according to the optimal hyper-parameter vector, and a ridge regression classifier is trained by using the mixed feature vector of the training set; and acquiring to-be-detected gas turbine sensor data, extracting a new mixed feature vector by using the optimal EuSN reserve pool, inputting the new mixed feature vector into the ridge regression classifier, and outputting a final gas turbine state classification result.

Description

Technical Field

[0001] This invention relates to the field of carbon emissions, and more particularly to a method, apparatus, and electronic equipment for classifying carbon emission data from gas turbines. Background Technology

[0002] As a core piece of equipment in the modern power industry and energy conversion field, the safety and efficiency of gas turbine operation are crucial for achieving global carbon emission reduction goals. During actual operation, the emission levels of nitrogen oxides and carbon monoxide are not only subject to stringent environmental regulations but are also key indicators for assessing the health of the unit's combustion chamber and overall combustion efficiency. Therefore, real-time and accurate classification monitoring of gas turbine emissions has become an essential means to ensure safe power plant operation and environmental compliance.

[0003] However, classifying the emission status of gas turbines faces extremely high data complexity. The unit's emission performance exhibits a complex nonlinear coupling relationship with environmental parameters. This multivariate data characteristic makes traditional monitoring methods based on single thresholds or simple logical judgments ineffective. Although data-driven intelligent diagnostic technologies have developed in recent years, classification algorithms for such high-dimensional, multivariate time-series data still have limitations. Traditional machine learning methods often struggle to capture long-term dynamic dependencies, while mainstream deep learning models, although highly accurate, are typically large in scale, consume significant computational resources, and their hyperparameter configuration relies on human experience, making rapid deployment and adaptive optimization difficult in industrial control environments with limited computing power.

[0004] Therefore, developing a gas turbine emission status classification method that can effectively integrate multi-dimensional sensor information, automatically capture time series features, and has low computational cost has significant engineering application value. Summary of the Invention

[0005] This invention provides a method, apparatus, and electronic device for classifying carbon emission data from gas turbines. This invention can classify different categories of carbon emission data from gas turbines, providing a reliable and efficient data analysis method for gas turbine operating scenarios. Details are described below:

[0006] A first aspect: a method for classifying carbon emission data from gas turbines, the method comprising:

[0007] Based on the gas turbine sensor data, the CMA-ES optimization module is initialized to generate a candidate hyperparameter vector containing network parameters;

[0008] Based on the candidate hyperparameter vectors, an EuSN reservoir is constructed; gas turbine sensor data is input into the EuSN reservoir to obtain its final hidden state as a dynamic feature vector; based on the gas turbine sensor data, global statistics of independent variables of each sensor are calculated to obtain static feature vectors.

[0009] The dynamic feature vector is concatenated with the static feature vector to obtain a mixed feature vector. Based on the mixed feature vector, the classification objective function value is calculated on the validation set, and the candidate hyperparameter vector is iteratively updated through the CMA-ES optimization module to obtain the optimal hyperparameter vector.

[0010] The optimal EuSN reservoir is determined based on the optimal hyperparameter vector, and a ridge regression classifier is trained using the mixed feature vectors of the training set.

[0011] Acquire sensor data from the gas turbine to be tested, extract new hybrid feature vectors using the optimal EuSN reservoir, and input them into a ridge regression classifier to output the final gas turbine state classification result.

[0012] The candidate hyperparameter vector includes: the size of the reserve pool. Euler step size diffusion coefficient Circular connection rate and scaling factor.

[0013] The method includes: constructing a cyclic weight matrix for the reserve pool.

[0014] Generate dense random weight matrix and based on the circular connection rate The generated binary mask matrix Calculate the sparse intermediate matrix :

[0015]

[0016] The final cyclic weight matrix is ​​obtained:

[0017]

[0018] in, To represent the Hadamardi (a type of calculus); This represents the transpose of the sparse intermediate matrix; This is the cyclic weight matrix.

[0019] Wherein, the dynamic feature vector is: for input sample Hidden state By time step Iterative updates using the forward Euler method:

[0020]

[0021] in, For time steps The state vector of the reserve pool; This is the state vector of the reservoir at the previous time step; The Euler step size hyperparameter controls the update magnitude; This is the diffusion coefficient hyperparameter, used for stable discrete propagation; It is the identity matrix; The input weight matrix; For time steps The gas turbine sensor data vector at any given time; It is the bias vector; The hyperbolic tangent activation function is used. This is the cyclic weight matrix.

[0022] The static feature vector is:

[0023] make Indicates the first Calculate the mean of the entire time series vector of each sensor. Standard deviation Minimum value and maximum value ;

[0024] All By concatenating the statistics from each sensor, a static feature vector is obtained. :

[0025]

[0026] in, For dimension The static feature vector, For the first The average of data from each sensor. For the first Standard deviation of sensor data For the first The minimum value of sensor data, For the first The maximum value of data from each sensor.

[0027] The mixed feature vector is:

[0028]

[0029] in, The fused feature vector has a dimension of . ; : Represents vector concatenation operation.

[0030] The final gas turbine state classification result is as follows:

[0031]

[0032] in, This refers to the final predicted gas turbine state category; This represents the total number of gas turbine status categories; Weight matrix Corresponding category Column vectors; For bias vector Corresponding category Element; To determine the category that maximizes the result of the expression within the parentheses. The value of .

[0033] A second aspect is an apparatus for classifying carbon emission data from gas turbines, the apparatus comprising: a processor and a memory, the memory storing program instructions, the processor invoking the program instructions stored in the memory to cause the apparatus to perform the method described in any of the first aspects.

[0034] Third aspect: A computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method described in any one of the first aspects. Fourth aspect: An electronic device for classifying carbon emission data from a gas turbine, the electronic device including a processor, a memory, a network interface, a display screen, and an input device connected via a system bus, the processor invoking program instructions stored in the memory to cause the device to perform the method described in any one of the first aspects.

[0035] The beneficial effects of the technical solution provided by this invention are:

[0036] 1. This invention designs an innovative hybrid feature architecture that achieves deep collaboration between dynamic and static information. On the one hand, it is a dynamic path based on Eulerian state networks, which can capture complex nonlinear dynamic correlations within multivariate time series with extremely high efficiency and stability. On the other hand, it is a static path responsible for extracting the global statistical features of each independent variable, providing the model with a robust description of the overall distribution of the series.

[0037] 2. This invention integrates a highly efficient self-optimization module, which systematically solves the practicality problem of reservoir calculation; the invention framework embeds an automated optimization engine driven by an adaptive evolution strategy of covariance matrix, which can efficiently search in complex high-dimensional parameter space and automatically determine the optimal model configuration for a given task.

[0038] 3. This invention develops a lightweight solution that combines excellent performance with extreme efficiency. While ensuring classification accuracy reaches or even surpasses many advanced deep learning models, it maintains a simplified training paradigm and computational efficiency. In the application scenario of gas turbine condition monitoring, it can effectively improve system deployment efficiency, reduce the consumption of computing resources, and provide reliable technical support for the efficient operation and maintenance of equipment. Attached Figure Description

[0039] Figure 1 A flowchart of a method for classifying carbon emission data from gas turbines;

[0040] Figure 2 Here is the algorithm architecture diagram;

[0041] Figure 3 This is a schematic diagram of a device for classifying carbon emission data from gas turbines.

[0042] Figure 4 A block diagram of an electronic device for classifying carbon emission data from gas turbines;

[0043] Figure 5 Another block diagram of an electronic device for classifying carbon emission data from gas turbines. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below.

[0045] Example 1

[0046] This invention proposes a method for classifying carbon emission data from gas turbines. (See also...) Figure 1 The specific steps of this method to solve the problem are as follows:

[0047] S101: Acquire gas turbine sensor data and divide it into training set, validation set and test set;

[0048] S102: Based on the gas turbine sensor data, initialize the CMA-ES optimization module and generate a candidate hyperparameter vector containing network parameters;

[0049] S103: Construct the EuSN reserve pool based on the candidate hyperparameter vector;

[0050] S104: Input the gas turbine sensor data into the EuSN storage pool and obtain its final hidden state as a dynamic feature vector;

[0051] S105: Based on the gas turbine sensor data, calculate the global statistics of the independent variables of each sensor to obtain the static feature vector;

[0052] S106: Concatenate the dynamic feature vector with the static feature vector to obtain a hybrid feature vector;

[0053] S107: Based on the mixed feature vector, calculate the classification objective function value on the validation set, and iteratively update the candidate hyperparameter vector through the CMA-ES optimization module to obtain the optimal hyperparameter vector;

[0054] S108: Based on the optimal EuSN reservoir determined by the optimal hyperparameter vector, train a ridge regression classifier using the mixed feature vector of the training set;

[0055] S109: Acquire sensor data of the gas turbine to be detected, extract new hybrid feature vectors using the optimal EuSN reservoir and input them into the ridge regression classifier, and output the final gas turbine state classification result.

[0056] In summary, the embodiments of the present invention can classify different categories of gas turbine carbon emission data through steps 101-109, providing a reliable and efficient data analysis method for gas turbine operating scenarios.

[0057] Example 2

[0058] The specific formula will be used below. Figure 2 The following example will be used to provide a detailed description of the solution in Embodiment 1, as detailed below:

[0059] In step S101 and in this embodiment of the invention, a multivariate time series of the gas turbine operation process is collected. According to a preset ratio, for example: 70% for the test set, 15% for the training set, and 15% for the validation set, the obtained sample set is divided into a training set for model training, a validation set for hyperparameter evaluation, and a test set for testing model performance.

[0060] : Indicates time-series data taken from gas turbine sensors;

[0061] : Indicates the time step of a time series;

[0062] : Indicates the number of variables, i.e. the number of gas turbine sensors.

[0063] In step S102 and in this embodiment of the invention, based on the gas turbine sensor data, the CMA-ES (Covariance Matrix Adaptive Evolution Strategy) optimization module is initialized to generate a candidate hyperparameter vector containing network parameters. Let a vector containing multiple hyperparameters be... , This includes: the size of the reserve pool Euler step size diffusion coefficient Circular connection rate and scaling factor. This module requires multiple rounds of evolution, assuming that in the... In generational evolution, sampling is performed from a multivariate Gaussian distribution. The candidate hyperparameter vectors are:

[0064]

[0065] : No. The generation produced A candidate hyperparameter vector;

[0066] : No. The mean vector of a multivariate Gaussian distribution;

[0067] Overall evolutionary step size;

[0068] : Multivariate standard Gaussian distribution;

[0069] The covariance matrix determines the shape and direction of the parameter search distribution.

[0070] In step S103 and in this embodiment of the invention, an EuSN (Eulerian State Network) reservoir is constructed based on the candidate hyperparameter vector. The reservoir size is then determined using the hyperparameters. and cyclic connection rate Construct the cyclic weight matrix of the reserve pool To ensure system stability, it is constructed as an antisymmetric matrix with sparsity. First, a dense random weight matrix is ​​generated. and based on the circular connection rate The generated binary mask matrix Calculate the sparse intermediate matrix :

[0071]

[0072] This leads to the final cyclic weight matrix:

[0073]

[0074] : indicates the Hadamard volume;

[0075] : Represents the transpose of the sparse intermediate matrix.

[0076] : Represents the cyclic weight matrix.

[0077] In step S104 and in this embodiment of the invention, the gas turbine sensor data is input into the cyclic weight matrix of the EuSN storage pool. The final hidden state is obtained as a dynamic feature vector.

[0078] For input samples Hidden state By time step Iterative updates using the forward Euler method:

[0079]

[0080] Time step The state vector of the reserve pool;

[0081] : The state vector of the reservoir at the previous time step;

[0082] Euler step size hyperparameter controls the update magnitude;

[0083] : Diffusion coefficient hyperparameter, used for stable discrete propagation;

[0084] : Identity matrix;

[0085] Input weight matrix;

[0086] Time step The gas turbine sensor data vector at any given time;

[0087] Bias vector;

[0088] Hyperbolic tangent activation function.

[0089] : Cyclic weight matrix.

[0090] Complete traversal After each time step, retrieve the last hidden state. As a dynamic feature vector.

[0091] In step S105 and in this embodiment of the invention, global statistics of each sensor independent variable are calculated based on gas turbine sensor data to obtain static feature vectors.

[0092] make Indicates the first Collect the entire time series vector of each sensor. Calculate the mean of each. Standard deviation Minimum value and maximum value All By concatenating the statistics from each sensor, a static feature vector is obtained. :

[0093]

[0094] Dimensions are The static feature vector, For the first The average of data from each sensor. For the first Standard deviation of sensor data For the first The minimum value of sensor data, For the first The maximum value of data from each sensor.

[0095] In step S106 and in this embodiment of the invention, the dynamic feature vector and the static feature vector are concatenated to obtain a hybrid feature vector.

[0096]

[0097] The fused hybrid feature vector has the following dimensions: ;

[0098] : Represents vector concatenation operation.

[0099] In step S107 and in this embodiment of the invention, the classification objective function value is calculated on the validation set based on the hybrid feature vector, and the candidate hyperparameter vector is iteratively updated through the CMA-ES optimization module to obtain the optimal hyperparameter vector.

[0100] Calculate the classification accuracy of the model corresponding to each candidate hyperparameter vector on the validation set. Select the model with the best performance. Given a number of individuals, update the mean vector for the next generation. Covariance Matrix :

[0101]

[0102] The optimal number of individuals selected;

[0103] : Preset weighting, and satisfying ;

[0104] : indicates the first The performance of the generation ranks first A candidate hyperparameter vector for each bit.

[0105] Repeat the evolution process until the termination condition is met, and output the optimal hyperparameter vector that maximizes the accuracy on the validation set. .

[0106] In step S108 and in this embodiment of the invention, a ridge regression classifier is trained using the mixed feature vectors of the training set based on the optimal EuSN reservoir determined by the optimal hyperparameter vector.

[0107] On the complete training set, a linear readout layer is trained using a loss function with regularization:

[0108]

[0109] Loss function;

[0110] The number of training samples;

[0111] : No. A mixture of feature vectors from training samples;

[0112] : No. Each sample corresponds to a gas turbine status label;

[0113] The readout layer weight matrix to be solved;

[0114] The readout layer bias vector to be solved;

[0115] : Hyperparameters that control the strength of regularization;

[0116] The square of the L2 norm;

[0117] : The square of the Frobenius norm.

[0118] By solving the closed-form solution of this convex optimization problem, the fixed... and .

[0119] In step S109 and in this embodiment of the invention, sensor data of the gas turbine to be detected is acquired, a new hybrid feature vector is extracted using the optimal EuSN reservoir and input into the ridge regression classifier, and the final gas turbine state classification result is output.

[0120] New mixed feature vectors are extracted from the data to be detected using the methods in steps S104 to S106. Substitute the values ​​into the decision function to obtain the predicted state category. :

[0121]

[0122] The final predicted gas turbine condition category;

[0123] The total number of gas turbine status categories;

[0124] Weight matrix Corresponding category Column vectors;

[0125] Bias vector Corresponding category Element;

[0126] : Select the category that maximizes the result of the expression within the parentheses. The value of .

[0127] In summary, the embodiments of the present invention adopt a combination of dynamic and static features, combined with a self-optimizing hyperparameter search algorithm, which solves the problems of low accuracy and low efficiency of existing methods that rely on only a single perspective. It can complete the analysis and monitoring of actual gas turbine operation scenarios with high efficiency, and improve the intelligence and efficiency of the unit operation process.

[0128] Example 3

[0129] This invention uses nine datasets from a publicly available time-series classification dataset and one gas turbine sensor dataset to verify the performance of a proposed method for classifying gas turbine carbon emission data. For all data, a pre-defined training and test set was used. The average rank of four algorithms was compared, and the results are shown in Table 1. The method of this invention achieved the lowest average rank value, ranking first among all algorithms. This fully demonstrates that the dynamic and static feature combination strategy proposed in this invention can effectively classify gas turbine state data.

[0130] The Gas Turbine dataset originates from gas turbine data in a thermal power plant. The data is collected by multiple sensors placed within the system to measure the concentrations of carbon monoxide (CO) and nitrogen oxides (NOx) in emissions in real time. This is a binary classification dataset; based on the changes in sensor data over a period of time, it determines whether the current emission from the system is carbon monoxide or nitrogen oxides.

[0131] Table 1. Comparison of classification accuracy of the method of this invention with four other algorithms on nine datasets.

[0132]

[0133] Table 2. Classification accuracy information of the method of the present invention on the Gas Turbine dataset.

[0134]

[0135] As can be seen from the results in Tables 1 and 2, the embodiments of the present invention have a high classification accuracy and can achieve the classification of carbon emissions from gas turbines.

[0136] Example 4

[0137] An apparatus for classifying carbon emission data from gas turbines, the apparatus comprising: a processor and a memory, the memory storing program instructions, the processor calling the program instructions stored in the memory to cause the apparatus to perform the following method in Embodiment 1:

[0138] Based on the gas turbine sensor data, the CMA-ES optimization module is initialized to generate a candidate hyperparameter vector containing network parameters;

[0139] Based on the candidate hyperparameter vectors, an EuSN reservoir is constructed; gas turbine sensor data is input into the EuSN reservoir to obtain its final hidden state as a dynamic feature vector; based on the gas turbine sensor data, global statistics of independent variables of each sensor are calculated to obtain static feature vectors.

[0140] The dynamic feature vector is concatenated with the static feature vector to obtain a mixed feature vector. Based on the mixed feature vector, the classification objective function value is calculated on the validation set, and the candidate hyperparameter vector is iteratively updated through the CMA-ES optimization module to obtain the optimal hyperparameter vector.

[0141] The optimal EuSN reservoir is determined based on the optimal hyperparameter vector, and a ridge regression classifier is trained using the mixed feature vectors of the training set.

[0142] Acquire sensor data from the gas turbine to be tested, extract new hybrid feature vectors using the optimal EuSN reservoir, and input them into a ridge regression classifier to output the final gas turbine state classification result.

[0143] Among them, the candidate hyperparameter vector includes: reservoir size Euler step size diffusion coefficient Circular connection rate and scaling factor.

[0144] This includes: constructing the cyclic weight matrix for the reserve pool,

[0145] Generate dense random weight matrix and based on the circular connection rate The generated binary mask matrix Calculate the sparse intermediate matrix :

[0146]

[0147] The final cyclic weight matrix is ​​obtained:

[0148]

[0149] in, To represent the Hadamardi (a type of calculus); This represents the transpose of the sparse intermediate matrix.

[0150] The dynamic feature vector is: for the input sample Hidden state By time step Iterative updates using the forward Euler method:

[0151]

[0152] in, For time steps The state vector of the reserve pool; This is the state vector of the reservoir at the previous time step; The Euler step size hyperparameter controls the update magnitude; This is the diffusion coefficient hyperparameter, used for stable discrete propagation; It is the identity matrix; The input weight matrix; For time steps The gas turbine sensor data vector at any given time; It is the bias vector; It is the hyperbolic tangent activation function; This is the cyclic weight matrix.

[0153] The static feature vector is:

[0154] make Indicates the first Calculate the mean of the entire time series vector of each sensor. Standard deviation Minimum value and maximum value ;

[0155] All By concatenating the statistics from each sensor, a static feature vector is obtained. :

[0156]

[0157] in, For dimension The static feature vector, For the first The average of data from each sensor. For the first Standard deviation of sensor data For the first The minimum value of sensor data, For the first The maximum value of data from each sensor.

[0158] The mixed feature vector is:

[0159]

[0160] in, The fused feature vector has a dimension of . ; This represents the vector concatenation operation.

[0161] The final gas turbine state classification result is output as follows:

[0162]

[0163] in, This refers to the final predicted gas turbine state category; This represents the total number of gas turbine status categories; Weight matrix Corresponding category Column vectors; For bias vector Corresponding category Element; To determine the category that maximizes the result of the expression within the parentheses. The value of .

[0164] It should be noted that the device descriptions in the above embodiments correspond to the method descriptions in the embodiments, and the embodiments of the present invention will not be repeated here.

[0165] The execution entities of the aforementioned processor and memory can be devices with computing functions such as computers, microcontrollers, and single-chip microcomputers. In specific implementations, the embodiments of the present invention do not limit the execution entities and can select them according to the needs of actual applications.

[0166] Data signals are transmitted between the memory and the processor via a bus, which will not be elaborated upon in this embodiment of the invention.

[0167] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium, the storage medium including a stored program, which, when the program is running, controls the device where the storage medium is located to execute the method steps in the above embodiments.

[0168] The computer-readable storage medium includes, but is not limited to, flash memory, hard disk, solid-state drive, etc.

[0169] It should be noted that the description of the readable storage medium in the above embodiments corresponds to the description of the method in the embodiments, and the embodiments of the present invention will not be repeated here.

[0170] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated.

[0171] A computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in or transmitted through a computer-readable storage medium. A computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic or semiconductor, etc.

[0172] Example 5

[0173] Figure 4 This is a block diagram illustrating an electronic device for classifying carbon emission data from a gas turbine, according to an exemplary embodiment. The electronic device can be a terminal, and its internal structure diagram can be as follows: Figure 4 As shown, the electronic device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with external terminals via a network connection. When executed by the processor, the computer program implements a method for classifying carbon emission data from gas turbines. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0174] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0175] Figure 5 This is a block diagram of an electronic device for classifying carbon emission data from a gas turbine, according to an exemplary embodiment. The electronic device may be a server. The electronic device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the electronic device is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a system for classifying carbon emission data from a gas turbine.

[0176] Those skilled in the art will understand that Figure 5The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0177] In an exemplary embodiment, an electronic device is also provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement a method for classifying carbon emission data from a gas turbine as described in an embodiment of this application.

[0178] In an exemplary embodiment, a computer-readable storage medium is also provided, which, when executed by a processor of an electronic device, enables the electronic device to perform a method for classifying carbon emission data of a gas turbine according to an embodiment of this application. The computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc.

[0179] In an exemplary embodiment, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute a method for classifying carbon emission data of a gas turbine according to an embodiment of this application.

[0180] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0181] Unless otherwise specified, the model numbers of the various devices in this embodiment of the invention are not limited, and any device that can perform the above functions is acceptable.

[0182] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0183] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for classifying carbon emission data from gas turbines, characterized in that, The method includes: Based on the gas turbine sensor data, the CMA-ES optimization module is initialized to generate a candidate hyperparameter vector containing network parameters; Based on the candidate hyperparameter vectors, an EuSN reservoir is constructed; gas turbine sensor data is input into the EuSN reservoir to obtain its final hidden state as a dynamic feature vector; based on the gas turbine sensor data, global statistics of independent variables of each sensor are calculated to obtain static feature vectors. The dynamic feature vector is concatenated with the static feature vector to obtain a mixed feature vector. Based on the mixed feature vector, the classification objective function value is calculated on the validation set, and the candidate hyperparameter vector is iteratively updated through the CMA-ES optimization module to obtain the optimal hyperparameter vector. The optimal EuSN reservoir is determined based on the optimal hyperparameter vector, and a ridge regression classifier is trained using the mixed feature vectors of the training set. Acquire sensor data from the gas turbine to be tested, extract new hybrid feature vectors using the optimal EuSN reservoir, and input them into a ridge regression classifier to output the final gas turbine state classification result.

2. The method for classifying carbon emission data from gas turbines according to claim 1, characterized in that, The candidate hyperparameter vector includes: reserve pool size. Euler step size diffusion coefficient Circular connection rate and scaling factor.

3. The method for classifying carbon emission data from gas turbines according to claim 1, characterized in that, The method includes: constructing a cyclic weight matrix for the reserve pool, Generate dense random weight matrix and based on the circular connection rate The generated binary mask matrix Calculate the sparse intermediate matrix : ； The final cyclic weight matrix is ​​obtained: ； in, To represent the Hadamardi (a type of calculus); Let be the transpose of the sparse intermediate matrix.

4. The method for classifying carbon emission data of gas turbines according to claim 1, characterized in that, The dynamic feature vector is: for the input sample Hidden state By time step Iterative updates using the forward Euler method: ； in, For time steps The state vector of the reserve pool; This is the state vector of the reservoir at the previous time step; The Euler step size hyperparameter controls the update magnitude; This is the diffusion coefficient hyperparameter, used for stable discrete propagation; It is the identity matrix; The input weight matrix; For time steps The gas turbine sensor data vector at any given time; It is the bias vector; It is the hyperbolic tangent activation function; This is the cyclic weight matrix.

5. The method for classifying carbon emission data from gas turbines according to claim 1, characterized in that, The static feature vector is: make Indicates the first Calculate the mean of the entire time series vector of each sensor. Standard deviation Minimum value and maximum value ; All By concatenating the statistics from each sensor, a static feature vector is obtained. : ； For dimension The static feature vector, For the first The average of data from each sensor. For the first Standard deviation of sensor data For the first The minimum value of sensor data, For the first The maximum value of data from each sensor.

6. The method for classifying carbon emission data from gas turbines according to claim 1, characterized in that, The mixed feature vector is: ； in, The fused feature vector has a dimension of . ; This represents the vector concatenation operation.

7. The method for classifying carbon emission data of gas turbines according to claim 1, characterized in that, The final gas turbine state classification result is as follows: ； in, This refers to the final predicted gas turbine state category; This represents the total number of gas turbine status categories; Weight matrix Corresponding category Column vectors; For bias vector Corresponding category Element; To determine the category that maximizes the result of the expression within the parentheses. The value of .

8. A device for classifying carbon emission data from gas turbines, characterized in that, The device includes a processor and a memory, the memory storing program instructions, the processor calling the program instructions stored in the memory to cause the device to perform the method according to any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the method described in any one of claims 1-7.

10. An electronic device for classifying carbon emission data from gas turbines, characterized in that, The electronic device includes a processor, a memory, a network interface, a display screen, and an input device connected via a system bus, wherein the processor invokes program instructions stored in the memory to cause the device to perform the method described in any one of claims 1-7.

Images