Construction method and system of thermal power plant carbon emission prediction model

Abstract

The invention provides a thermal power plant carbon emission prediction model construction method and system, and the method comprises the steps: obtaining the basic data of a thermal power station, and generating a training set and a test set according to the basic data; constructing an ELM topology, including configuring hidden layer nodes, initializing weight and bias from an input layer to a hidden layer, and obtaining an ELM model; importing the training set into a network architecture of an ELM (Extreme Learning Model), optimizing an input weight matrix and an offset vector of the ELM by adopting a PSO (Particle Swarm Optimization) algorithm, and obtaining a PSO-ELM model according to an optimization result; importing the test set into a PSO-ELM model for testing, and judging whether the PSO-ELM model meets a preset precision condition or not according to a test result; and if the PSO-ELM model meets a preset precision condition, obtaining a final PSO-ELM model. The method can adapt to carbon emission prediction requirements of the thermal power plant in different operation periods and under different working conditions.

Description

Technical Field

[0001] This invention relates to the field of carbon emission prediction technology, and in particular to a method and system for constructing a carbon emission prediction model for thermal power plants. Background Technology

[0002] As the main source of electricity supply in my country, thermal power plants account for 51.7% of the country's total carbon dioxide emissions from energy consumption, making them one of the largest sources of carbon dioxide emissions. Therefore, carbon accounting for thermal power plants is the core foundation for achieving the "dual carbon" goals and a key basis for formulating emission reduction policies and conducting assessments.

[0003] Currently, accurate carbon accounting faces many challenges. Traditional material balance methods rely on theoretical combustion efficiency, leading to errors exceeding 10%. Direct monitoring methods require retrofitting costs of up to tens of millions of yuan and struggle to differentiate emissions from multiple units. Emission factor methods ignore coal quality differences, resulting in a 20% error. None of these methods can adapt to the variable operating conditions of thermal power units, causing errors to further expand to over 15% under dynamic conditions. High-precision solutions such as carbon capture technology require retrofitting at a single plant cost of 200-300 million yuan, hindering widespread adoption. Therefore, there is an urgent need to develop low-cost, high-precision, and dynamically adaptive carbon emission prediction methods. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for constructing a carbon emission prediction model for thermal power plants, aiming to solve the problem of inaccurate predictions caused by traditional carbon accounting methods relying on static combustion models and ignoring dynamic characteristics under varying operating conditions.

[0005] In a first aspect, the present invention provides a method for constructing a carbon emission prediction model for thermal power plants, the method comprising: To acquire basic data of thermal power plants, outliers in the basic data are identified using the interquartile range method. At the same time, interpolation methods are used to fill missing values ​​in the basic data. A time-series dynamic feature extraction method is introduced to reconstruct the input feature matrix, and training and test sets are generated based on the input feature matrix. Constructing the ELM topology includes configuring hidden layer nodes, initializing the weights and biases from the input layer to the hidden layer, and obtaining the ELM model; The training set is imported into the network architecture of the ELM model, and the PSO algorithm is used to optimize the input weight matrix and bias vector of the ELM model. The PSO-ELM model is obtained based on the optimization results. The test set is imported into the PSO-ELM model for testing, and the PSO-ELM model is judged based on the test results to determine whether the preset accuracy condition is met. If the PSO-ELM model meets the preset accuracy conditions, the final PSO-ELM model is obtained.

[0006] In some embodiments, the steps of acquiring basic data of a thermal power plant, constructing an input feature matrix and an output target vector based on the basic data, and generating a training set and a test set based on the input feature matrix include: The basic data includes active power, reactive power, lower heating value of fuel entering the furnace, carbon content of fuel entering the furnace, and total daily carbon emissions. Constructing the original feature vector A time window with a duration of 3 days is defined based on the original feature vector. Calculate the time series feature matrix The input feature matrix X and the output target vector Y are reconstructed, with the following structure: ; in, This represents the original feature vector collected on day n. The active power collected on day n. The reactive power collected on day n. The lower heating value of the fuel entering the furnace is collected on day n. The carbon content of the fuel fed into the furnace is collected on day n. The expression for the timing window is:

[0007] Define a time series window with a duration of 3 days. This timing window It includes data collected from day n-3 to day n. If n is less than 4, then use... Fill in the missing data; The expression for the time series feature matrix is: ; in, To be based on timing window The calculated time series feature matrix, , , , Based on timing window The mean of the corresponding characteristic quantities is calculated. , , , Based on timing window The standard deviation of the corresponding characteristic quantity is calculated. , , , Based on timing window The calculated rate of change of the corresponding characteristic quantity; The expression for the input feature matrix X is: ; The expression for the output target vector Y is: ; in, For the input feature matrix, To output the target vector, To be based on timing window The calculated time series feature matrix, , , These represent the total daily carbon emissions collected on day 1, day 2, and day n, respectively.

[0008] In some embodiments, the step of generating training and testing sets based on the input feature matrix includes: Normalize the input feature matrix: ; in, The input feature matrix is ​​normalized. This is a vector composed of the minimum values ​​of each column of feature data in the input feature matrix. It is a vector composed of the maximum values ​​of the feature data in each column of the input feature matrix; The normalized input feature matrix is ​​divided into a training set and a test set.

[0009] In some embodiments, the steps of constructing the ELM topology, including configuring hidden layer nodes, initializing the weights and biases from the input layer to the hidden layer, and obtaining the ELM model, include: Hidden layer nodes are determined using the following formula: ; in, The number of hidden layer nodes. The number of samples in the training set; Initialize the weights and biases from the input layer to the hidden layer. The output matrix of the hidden layer of the ELM model is as follows: ; in, This is the output matrix of the hidden layer. This is the weight matrix from the input layer to the hidden layer. The bias vector of the hidden layer. The activation function for the hidden layer; Calculate the weight matrix of the ELM model using the following formula:

[0010] in, It is the weight matrix of the output layer.

[0011] In some embodiments, the steps of importing the training set into the network architecture of the ELM model, optimizing the input weight matrix and bias vector of the ELM model using the PSO algorithm, and obtaining the PSO-ELM model based on the optimization results include: Initialize the position and velocity of the particles. The current position of each particle is represented by a set of ELM input weights and bias parameters: ; Update the particle velocity according to the following formula: ; Update the particle's position according to the following formula: ; in, For the i-th particle, , Let be the velocities of particle i at time t and time t+1, respectively. For inertial weights, and All are acceleration constants. and All numbers are random and between [0, 1]. This represents the best historical position for particle i. This represents the best historical position for all particles. The fitness value of the ELM model is calculated using the following formula: ; in, For fitness value, For true carbon emissions, This is a predicted value for carbon emissions; Calculate the fitness values ​​of all particles and update the particle states according to a greedy strategy: if the fitness value of a particle's current position is better than its individual historical best fitness value, then update the individual best position of the particle; if it is also better than the group's historical best fitness value, then update the global best position synchronously and obtain the optimal solution in the current iteration. Assign the optimal solution to the ELM model to complete the optimization of the ELM model.

[0012] In some embodiments, the step of importing the test set into the PSO-ELM model for testing and determining whether the PSO-ELM model meets the preset accuracy conditions based on the test results includes: The mean absolute percentage error of the PSO-ELM model is calculated using the following formula: ; The determination coefficient of the PSO-ELM model is calculated using the following formula: ; in, The mean absolute percentage error, As the coefficient of determination, The average of all actual carbon emissions; Determine whether the mean absolute percentage error and the coefficient of determination meet the preset accuracy conditions.

[0013] Secondly, the present invention provides a system for constructing a carbon emission prediction model for thermal power plants, the system comprising: The basic data acquisition module is used to acquire basic data of the thermal power plant, construct an input feature matrix and an output target vector based on the basic data, and generate a training set and a test set based on the input feature matrix. The model building module is used to build the ELM topology, including configuring hidden layer nodes, initializing the weights and biases from the input layer to the hidden layer, and obtaining the ELM model. The model optimization module is used to import the training set into the network architecture of the ELM model, optimize the input weight matrix and bias vector of the ELM model using the PSO algorithm, and obtain the PSO-ELM model based on the optimization results. The testing module is used to import the test set into the PSO-ELM model for testing, and to determine whether the PSO-ELM model meets the preset accuracy conditions based on the test results. The model output module is used to obtain the final PSO-ELM model if the PSO-ELM model meets the preset accuracy conditions.

[0014] Thirdly, the present invention provides a storage medium that stores one or more programs, which, when executed by a processor, implement the above-described method for constructing a carbon emission prediction model for thermal power plants.

[0015] Fourthly, the present invention provides an electronic device, the electronic device comprising a memory and a processor, wherein: The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the above-mentioned method for constructing a carbon emission prediction model for thermal power plants.

[0016] Compared with the prior art, the present invention has the following advantages: 1. This invention constructs an extreme learning machine network with an adaptive hidden layer structure, where the number of hidden layer nodes is dynamically configured according to the scale of the training samples, thus creating a matching mechanism between the network topology and data complexity. Then, a particle swarm optimization algorithm is used to globally search for the optimal solution of the input weight matrix and bias vector within the constrained space, eliminating parameter sensitivity caused by random initialization. Based on this, the input-output layer mapping relationship effectively suppresses the interference of matrix ill-conditioned problems on prediction stability. Finally, through a multi-source parameter collaborative modeling architecture, dynamic operating parameters are synchronously fused to establish a functional mapping relationship, solving the fundamental defect that static models cannot adapt to dynamic operating conditions.

[0017] 2. This invention optimizes the parameters of the Extreme Learning Machine through particle swarm optimization, enabling the model to not only perform well on the training set, but also to provide relatively reliable prediction results when faced with new carbon emission data that has not been used in the training. This reduces overfitting and can adapt to the carbon emission prediction needs of thermal power plants under different operating conditions and at different stages of operation.

[0018] 3. This invention introduces a time-series dynamic feature extraction method. Since the operating load and fuel characteristics of thermal power plants are not constant, the features extracted by the time-series window can encode this changing information. When the unit's state begins to change, its trend and fluctuation characteristics also change, thereby providing early warning signals for the model, making its predictions more adaptable to changes in operating conditions, rather than only performing well under stable conditions.

[0019] 4. This invention patent constructs a PSO-ELM model by introducing a time-series dynamic feature extraction method, which demonstrates excellent performance in predicting carbon emissions from thermal power units. The correlation between the predicted results and the true values ​​is as high as 0.984, and the average absolute percentage error is as low as 2.79%, which is better than the traditional extreme learning machine algorithm (correlation 0.938, error 6.08%). This proves that the invention can capture the dynamic evolution law of the system extremely effectively and improve the prediction accuracy and reliability. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a method for constructing a carbon emission prediction model for thermal power plants according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of a system for constructing a carbon emission prediction model for thermal power plants, as proposed in an embodiment of the present invention.

[0021] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, but does not exclude other elements or objects.

[0023] like Figure 1 As shown, this invention proposes a method for constructing a carbon emission prediction model for thermal power plants. This method includes steps S101 to S105, wherein: Step S101: Obtain basic data of the thermal power plant, construct an input feature matrix and an output target vector based on the basic data, and generate a training set and a test set based on the input feature matrix; It should be noted that this step first obtains the basic data of the thermal power plant, then uses the interquartile range method to identify outliers in the basic data, and at the same time uses interpolation methods to fill in the missing values ​​in the basic data. Based on the basic data, an input feature matrix and an output target vector are constructed. In order to overcome the shortcomings of existing prediction methods that only use static daily data and cannot effectively capture the dynamic evolution of the system, this invention introduces a time-series dynamic feature extraction method to reconstruct the input feature matrix, and generates training and test sets based on the reconstructed input feature matrix. In addition, in some embodiments, the basic data includes active power, reactive power, lower heating value of fuel entering the furnace, carbon content of fuel entering the furnace, and total daily carbon emissions. Constructing the original feature vector A time window with a duration of 3 days is defined based on the original feature vector. Calculate the time series feature matrix The input feature matrix X and the output target vector Y are reconstructed, with the following structure: ; in, This represents the original feature vector collected on day n. The active power collected on day n. The reactive power collected on day n. The lower heating value of the fuel entering the furnace is collected on day n. The carbon content of the fuel fed into the furnace is collected on day n. The expression for the timing window is:

[0024] Define a time series window with a duration of 3 days. This timing window It includes data collected from day n-3 to day n. If n is less than 4, then use... Fill in the missing data; The expression for the time series feature matrix is: ; in, To be based on timing window The calculated time series feature matrix, , , , Based on timing window The mean of the corresponding characteristic quantities is calculated. , , , Based on timing window The standard deviation of the corresponding characteristic quantity is calculated. , , , Based on timing window The calculated rate of change of the corresponding characteristic quantity; For example, Specifically, timing window The mean of active power within the range; other values ​​with the same subscript have the same meaning. resemblance, Specifically, timing window The standard deviation of active power within the range, and the meanings of other values ​​with the same subscript are the same as those in the range. resemblance, Specifically, it is obtained by fitting a linear trend using the least squares method, and the specific expression is: ; in, This is the average of all n values. Other formulas with the same subscripts are... Similar, but it needs to be and Replace with the corresponding feature values.

[0025] The expression for the input feature matrix X is: ; The expression for the output target vector Y is: ; in, For the input feature matrix, To output the target vector, To be based on timing window The calculated time series feature matrix, , , These represent the total daily carbon emissions collected on day 1, day 2, and day n, respectively.

[0026] In summary, this invention introduces a time-series dynamic feature extraction method. Since the operating load and fuel characteristics of thermal power plants are not static, the features extracted by the time-series window can encode this changing information. When the unit's state begins to change, its trend and fluctuation characteristics also change, thereby providing early warning signals for the model, making its predictions more adaptable to changes in operating conditions, rather than only performing well under stable conditions.

[0027] Furthermore, in some embodiments, the input feature matrix is ​​normalized: ; in, The input feature matrix is ​​normalized. This is a vector composed of the minimum values ​​of each column of feature data in the input feature matrix. It is a vector composed of the maximum values ​​of the feature data in each column of the input feature matrix; The normalized input feature matrix is ​​divided into a training set and a test set.

[0028] It should be noted that the features in the basic data vary greatly in size and magnitude. If directly input into the model, the features with larger magnitudes will dominate the optimization process, thus affecting the model's convergence speed and final accuracy. To address this issue, data normalization preprocessing is introduced to eliminate the influence of data size, bringing all features to the same magnitude. This accelerates the convergence speed and ensures that each feature contributes equally to the model training.

[0029] Step S102: Construct the ELM topology, including configuring hidden layer nodes, initializing the weights and biases from the input layer to the hidden layer, and obtaining the ELM model; It should be noted that in this step, the hidden layer nodes are determined specifically according to the following formula: ; in, The number of hidden layer nodes. The number of samples in the training set; Initialize the weights and biases from the input layer to the hidden layer. The output matrix of the hidden layer of the ELM model is as follows: ; in, This is the output matrix of the hidden layer. This is the weight matrix from the input layer to the hidden layer. The bias vector of the hidden layer. The activation function for the hidden layer; Calculate the weight matrix of the ELM model using the following formula:

[0030] in, It is the weight matrix of the output layer.

[0031] In summary, traditional ELM models have a fixed number of hidden layer nodes, making it difficult to adapt to the dynamic changes in data characteristics under different operating conditions of thermal power plants. This can easily lead to overfitting or underfitting under certain conditions, limiting the model's generalization ability. This invention introduces an empirical formula for training set size to dynamically and reasonably determine the optimal number of hidden layer nodes, providing the model with an initial network structure that matches the problem complexity, thereby maintaining good generalization performance under complex operating conditions.

[0032] Step S103: Import the training set into the network architecture of the ELM model, and use the PSO algorithm to optimize the input weight matrix and bias vector of the ELM model, and obtain the PSO-ELM model based on the optimization results; First, initialize the position and velocity of the particles. The current position of each particle is represented by a set of ELM input weights and bias parameters: ; Update the particle velocity according to the following formula: ; Update the particle's position according to the following formula: ; in, For the i-th particle, , Let be the velocities of particle i at time t and time t+1, respectively. For inertial weights, and All are acceleration constants. and All numbers are random and between [0, 1]. This represents the best historical position for particle i. This represents the best historical position for all particles. The fitness value of the ELM model is calculated using the following formula: ; in, For fitness value, For true carbon emissions, This is the predicted value of carbon emissions. The average of all actual carbon emissions; Calculate the fitness values ​​of all particles and update the particle states according to a greedy strategy: if the fitness value of a particle's current position is better than its individual historical best fitness value, then update the individual best position of the particle; if it is also better than the group's historical best fitness value, then update the global best position synchronously and obtain the optimal solution in the current iteration. Assign the optimal solution to the ELM model to complete the optimization of the ELM model.

[0033] In summary, to address the problem of unstable model training results and large fluctuations in prediction performance caused by the random generation of input weights and biases in traditional ELM models, the swarm intelligence search mechanism of the PSO algorithm is introduced into the parameter optimization process of ELM models. This optimizes the core parameters of the ELM model, ensuring that a stable and high-performance model that approaches the global optimum can be obtained in each training session.

[0034] Step S104: Import the test set into the PSO-ELM model for testing, and determine whether the PSO-ELM model meets the preset accuracy conditions based on the test results; Specifically, the mean absolute percentage error of the PSO-ELM model is calculated using the following formula: ; The determination coefficient of the PSO-ELM model is calculated using the following formula: ; in, The mean absolute percentage error, The coefficient of determination; Determine whether the mean absolute percentage error and the coefficient of determination meet the preset accuracy conditions.

[0035] It should be noted that, in response to the lack of objective and unified performance evaluation standards for optimized prediction models, this invention introduces two key indicators, mean absolute percentage error and coefficient of determination, which respectively measure prediction accuracy and goodness of fit, providing quantifiable and reproducible objective evaluation standards for model performance.

[0036] Step S105: If the PSO-ELM model meets the preset accuracy conditions, the final PSO-ELM model is obtained.

[0037] In summary, the construction method for the carbon emission prediction model of thermal power plants described above has the following advantages: 1. This invention constructs an extreme learning machine network with an adaptive hidden layer structure, where the number of hidden layer nodes is dynamically configured according to the scale of the training samples, thus creating a matching mechanism between the network topology and data complexity. Then, a particle swarm optimization algorithm is used to globally search for the optimal solution of the input weight matrix and bias vector within the constrained space, eliminating parameter sensitivity caused by random initialization. Based on this, the input-output layer mapping relationship effectively suppresses the interference of matrix ill-conditioned problems on prediction stability. Finally, through a multi-source parameter collaborative modeling architecture, dynamic operating parameters are synchronously fused to establish a functional mapping relationship, solving the fundamental defect that static models cannot adapt to dynamic operating conditions.

[0038] 2. This invention optimizes the parameters of the Extreme Learning Machine through particle swarm optimization, enabling the model to not only perform well on the training set, but also to provide relatively reliable prediction results when faced with new carbon emission data that has not been used in the training. This reduces overfitting and can adapt to the carbon emission prediction needs of thermal power plants under different operating conditions and at different stages of operation.

[0039] 3. This invention introduces a time-series dynamic feature extraction method. Since the operating load and fuel characteristics of thermal power plants are not constant, the features extracted by the time-series window can encode this changing information. When the unit's state begins to change, its trend and fluctuation characteristics also change, thereby providing early warning signals for the model, making its predictions more adaptable to changes in operating conditions, rather than only performing well under stable conditions.

[0040] 4. This invention patent constructs a PSO-ELM model by introducing a time-series dynamic feature extraction method, which demonstrates excellent performance in predicting carbon emissions from thermal power units. The correlation between the predicted results and the true values ​​is as high as 0.984, and the average absolute percentage error is as low as 2.79%, which is better than the traditional extreme learning machine algorithm (correlation 0.938, error 6.08%). This proves that the invention can capture the dynamic evolution law of the system extremely effectively and improve the prediction accuracy and reliability.

[0041] like Figure 2 As shown, one embodiment of the present invention proposes a system for constructing a carbon emission prediction model for thermal power plants, the system comprising: The basic data acquisition module 10 is used to acquire basic data of the thermal power plant, construct an input feature matrix and an output target vector based on the basic data, and generate a training set and a test set based on the input feature matrix. Model building module 20 is used to build the ELM topology, including configuring hidden layer nodes, initializing the weights and biases from the input layer to the hidden layer, and obtaining the ELM model; The model optimization module 30 is used to import the training set into the network architecture of the ELM model, optimize the input weight matrix and bias vector of the ELM model using the PSO algorithm, and obtain the PSO-ELM model based on the optimization results. The test module 40 is used to import the test set into the PSO-ELM model for testing, and to determine whether the PSO-ELM model meets the preset accuracy conditions based on the test results. The model output module 50 is used to obtain the final PSO-ELM model if the PSO-ELM model meets the preset accuracy conditions.

[0042] In another aspect, the present invention also proposes a storage medium on which one or more programs are stored, which, when executed by a processor, implement the above-described method for constructing a carbon emission prediction model for thermal power plants.

[0043] In another aspect, the present invention also proposes an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to realize the above-mentioned method for constructing a carbon emission prediction model for thermal power plants.

[0044] Those skilled in the art will understand that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain stored, communicated, propagated, or transmitted programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0045] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0046] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0047] While embodiments of the present invention have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations fall within the scope and spirit of the invention as set forth in the claims. Furthermore, the invention described herein may have other embodiments and can be implemented or carried out in various ways.

Claims

1. A method for constructing a carbon emission prediction model of a thermal power plant, characterized in that, The method comprises: acquiring basic data of a thermal power plant, constructing an input feature matrix and an output target vector according to the basic data, and generating a training set and a test set according to the input feature matrix; constructing an ELM topology, including configuring hidden layer nodes, initializing weights and biases from an input layer to a hidden layer, and obtaining an ELM model; inputting the training set into a network architecture of the ELM model, optimizing input weight matrices and bias vectors of the ELM model by using a PSO algorithm, and obtaining a PSO-ELM model according to an optimization result; testing the test set in the PSO-ELM model, and judging whether the PSO-ELM model meets a preset accuracy condition according to a test result; if the PSO-ELM model meets the preset accuracy condition, obtaining a final PSO-ELM model.

2. The method of claim 1, wherein, The step of acquiring basic data of a thermal power plant, constructing an input feature matrix and an output target vector according to the basic data, and generating a training set and a test set according to the input feature matrix comprises: the basic data comprises active power, reactive power, low calorific value of fuel into a furnace, carbon content of fuel into the furnace, and total daily carbon emission; Constructing the original feature vector and define a time length of 3 days of timing window according to the original feature vector , calculate the timing feature matrix , reconstruct the input feature matrix X and the output target vector Y, the structure is: ； wherein, is the original feature vector collected on the nth day, is the active power collected on the nth day, is the reactive power collected on the nth day, is the low calorific value of the fuel into the furnace collected on the nth day, is the carbon content of the fuel into the furnace collected on the nth day; an expression of a time sequence window is: ； Define a time series window of length 3 days The time series window contains data collected from day n-3 to day n, if n is less than 4, then fill in the missing data with an expression of a time sequence feature matrix is: ； wherein, is a time series feature matrix calculated according to a time series window , , , , is a mean of the corresponding feature quantity calculated according to a time series window , , , , is a standard deviation of the corresponding feature quantity calculated according to a time series window , , , , is a rate of change of the corresponding feature quantity calculated according to a time series window . an expression of the input feature matrix X is: ； an expression of the output target vector Y is: ； wherein, is an input feature matrix, is an output target vector, is a time series feature matrix calculated according to a time series window is a time series feature matrix calculated according to a time series window , , is the total daily carbon emission amount collected on the first day, the second day, and the nth day, respectively.

3. The method of claim 2, wherein, The step of generating a training set and a test set according to the input feature matrix comprises: normalizing the input feature matrix: ； wherein, is a normalized input feature matrix, is a vector of minimum values of each column of the input feature matrix, is a vector of maximum values of each column of the input feature matrix; dividing the normalized input feature matrix into the training set and the test set.

4. The method of claim 3, wherein, The step of constructing an ELM topology, including configuring hidden layer nodes, initializing weights and biases from an input layer to a hidden layer, and obtaining an ELM model comprises: determining hidden layer nodes according to the following formula: ； wherein, is the number of hidden layer nodes, is the number of training set samples; initializing weights and biases from an input layer to a hidden layer, and an output matrix of the ELM model hidden layer is: ； wherein, is an output matrix of the hidden layer, is a weight matrix from the input layer to the hidden layer, is a bias vector of the hidden layer, is an activation function of the hidden layer; calculating a weight matrix of the ELM model according to the following formula: ； wherein, is a weight matrix of the output layer.

5. The method of claim 4, wherein, The step of inputting the training set into a network architecture of the ELM model, optimizing input weight matrices and bias vectors of the ELM model by using a PSO algorithm, and obtaining a PSO-ELM model according to an optimization result comprises: initializing positions and speeds of particles, and a current position of each particle is represented as a set of input weights and bias parameters of ELM: ； updating a speed of a particle according to the following formula: ； updating a position of the particle according to the following formula: ； wherein, is the i-th particle, , are the velocities of the i-th particle at time t, t+1 respectively, is the inertia weight, and are acceleration constants, and are random numbers between [0, 1], is the historical best position of the i-th particle, is the historical best position of all particles; calculating a fitness value of the ELM model according to the following formula: ； wherein, is the fitness value, is the real carbon emission value, is the carbon emission prediction value; calculating fitness values of all particles, and updating a particle state according to a greedy strategy: if a fitness value of a current position of the particle is better than an individual historical optimal fitness value of the particle, the individual optimal position of the particle is updated; if the fitness value of the current position of the particle is better than a group historical optimal fitness value at the same time, a global optimal position is updated synchronously, and an optimal solution in a current iteration is obtained, the optimal solution is assigned to the ELM model, and optimization of the ELM model is completed.

6. The method of claim 5, wherein, The step of testing the test set in the PSO-ELM model, and judging whether the PSO-ELM model meets a preset accuracy condition according to a test result comprises: calculating a mean absolute percentage error of the PSO-ELM model according to the following formula: ； The determination coefficient of the PSO-ELM model is calculated according to the following formula: ； wherein, is the mean absolute percentage error, is the coefficient of determination, is the mean of all true carbon emission values; It is judged whether the average absolute percentage error and the determination coefficient meet the preset accuracy condition.

7. A system for constructing a carbon emission prediction model of a thermal power plant, characterized in that, The system comprises: The basic data acquisition module is configured to acquire basic data of the thermal power plant, construct an input feature matrix and an output target vector according to the basic data, and generate a training set and a test set according to the input feature matrix; The model construction module is configured to construct an ELM topology, including configuring hidden layer nodes, initializing weights and biases from an input layer to a hidden layer, and obtaining an ELM model; The model optimization module is configured to import the training set into a network architecture of the ELM model, optimize an input weight matrix and a bias vector of the ELM model by using a PSO algorithm, and obtain a PSO-ELM model according to an optimization result; The test module is configured to import the test set into the PSO-ELM model for testing, and judge whether the PSO-ELM model meets a preset accuracy condition according to a test result; The model output module is configured to obtain a final PSO-ELM model if the PSO-ELM model meets the preset accuracy condition.

8. A storage medium, characterized by The storage medium stores one or more programs, which are executed by the processor to implement the method for constructing a carbon emission prediction model of a thermal power plant according to any one of claims 1-6.

9. An electronic device, comprising: The electronic device comprises a memory and a processor, wherein: The memory is configured to store a computer program; The processor is configured to execute the computer program stored on the memory to implement the method for constructing a carbon emission prediction model of a thermal power plant according to any one of claims 1-6.

Images