Carbon emission reduction supervision method based on artificial intelligence

By combining exponential function and adaptive parameter optimization technology, the shortcomings of the existing carbon emission reduction supervision methods in capturing nonlinear relationships and handling outliers are solved, the accuracy and stability of the carbon emission reduction supervision model are improved, and the more Efficient carbon emission reduction supervision.

CN118820693BActive Publication Date: 2024-11-15XIAMEN YIJUDA GRP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411307790.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-19
Publication Date
2024-11-15
Estimated Expiration
2044-09-19

AI Technical Summary

Technical Problem

Existing carbon emission reduction regulatory methods cannot effectively capture the complex nonlinear relationships in carbon emission reduction data, and cannot apply different carbon emission reduction scenarios. The ability to handle outliers or extreme values ​​in carbon emission reduction monitoring data is weak, resulting in carbon emission reduction The monitoring model has low accuracy, and the ability to optimize the carbon emission reduction supervision model parameters is poor, the global ability to optimize, the search efficiency is low and the convergence is weak, resulting in low stability and accuracy of carbon emission reduction supervision.

Method used

Using two different forms of exponential functions to design the activation function, introduce hyperparameter q to adjust the nonlinearity of the activation function, design adaptive scaling parameters and ranking scaling parameters, and perform fine-tuning operations to maintain the intersection points and exchange gene fragments. Genetic diversity, dynamically adjust the fine-tuning intensity, and build a carbon emission reduction regulatory model with high accuracy and stability.

Benefits of technology

The prediction and regulation capabilities of the carbon emission reduction supervision model have been improved, and the new carbon emission reduction data can be adapted to the new carbon emission reduction data more quickly, the stability and accuracy of the model have been enhanced, and the accuracy and global optimization capabilities of carbon emission reduction supervision have been improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118820693B_ABST
    Figure CN118820693B_ABST
Patent Text Reader

Abstract

The present invention discloses a carbon emission reduction supervision method based on artificial intelligence, and the method includes data collection, data preprocessing, establishing a carbon emission reduction supervision model and carbon emission reduction supervision. The present invention belongs to the field of carbon emission supervision technology, and specifically refers to a carbon emission reduction supervision method based on artificial intelligence. This scheme designs an activation function by combining two exponential functions to more accurately simulate the nonlinear behavior in the carbon emission reduction process; introduces a hyperparameter q to adjust the nonlinear degree of the activation function; promotes the generation of a better solution by introducing adaptive scaling parameters and ranking scaling parameters; helps the carbon emission reduction supervision model to explore the parameter space more comprehensively and improve the globality of the optimization by randomly selecting intersections and exchanging gene fragments; the fine-tuning operation makes the carbon emission reduction supervision model parameters widely explored in the early stage and concentratedly optimized in the later stage when optimizing the parameters, and finally constructs a carbon emission reduction supervision model with high accuracy and stability, thereby improving the accuracy of carbon emission reduction supervision.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission regulation, and specifically to a carbon emission reduction regulation method based on artificial intelligence. Background Art

[0002] Carbon emission reduction supervision method is a technology used to monitor, manage and optimize carbon emissions, involving the monitoring and management of carbon emissions in multiple fields such as industry, energy, transportation, agriculture, etc. However, the general carbon emission reduction supervision method has the problem that the model activation function cannot capture the complex nonlinear relationship in the carbon emission reduction data, cannot be applied to different carbon emission reduction scenarios, and has weak ability to handle outliers or extreme values ​​in carbon emission reduction monitoring data, resulting in low accuracy of the carbon emission reduction monitoring model; the general carbon emission reduction supervision method has poor ability to promote the generation of better solutions when optimizing the parameters of the carbon emission reduction supervision model, poor global optimization ability, low search efficiency and weak convergence, which leads to poor stability and low accuracy of carbon emission reduction supervision. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a carbon emission reduction supervision method based on artificial intelligence. In view of the fact that the model activation function of the general carbon emission reduction supervision method cannot capture the complex nonlinear relationship in the carbon emission reduction data, cannot be applied to different carbon emission reduction scenarios, and has weak ability to handle outliers or extreme values ​​in the carbon emission reduction monitoring data, resulting in low accuracy of the carbon emission reduction monitoring model, this scheme designs an activation function by combining two different forms of exponential functions to more accurately simulate the nonlinear behavior in the carbon emission reduction process and improve the prediction and adjustment capabilities of the model; the introduction of the hyperparameter q allows the nonlinear degree of the activation function to be adjusted according to different carbon emission reduction scenarios and data characteristics, thereby optimizing the model performance; the segmented design of the activation function can ensure stability and effectiveness under different data conditions, and is used to handle outliers and extreme values ​​in carbon emission reduction monitoring data; the convergence speed enables the model to adapt to new carbon emission reduction data more quickly, thereby improving the accuracy of carbon emission reduction supervision; for general carbon emission reduction monitoring The control method has poor ability to promote the generation of better solutions when optimizing the parameters of the carbon emission reduction supervision model, poor global optimization ability, low search efficiency and weak convergence, which leads to poor stability and low accuracy of carbon emission reduction supervision. This scheme introduces adaptive scaling parameters and ranking scaling parameters, so that the probability of individual selection can be adjusted according to the current iteration state when optimizing the parameters of the carbon emission reduction supervision model, which promotes the generation of better solutions; by randomly selecting intersections and exchanging gene fragments, it can effectively maintain gene diversity and avoid population degradation, which helps the carbon emission reduction supervision model to explore the parameter space more comprehensively and improve the globality of optimization; the fine-tuning operation performs local neighborhood operations on the optimal gene sequence, and the number of adaptive fine-tuning operations dynamically adjusts the fine-tuning intensity according to the number of iterations, so that the carbon emission reduction supervision model parameters are widely explored in the early stage and concentratedly optimized in the later stage, and finally achieves better convergence, and finally constructs a carbon emission reduction supervision model with high accuracy and stability, thereby improving the accuracy of carbon emission reduction supervision.

[0004] The technical solution adopted by the present invention is as follows: The carbon emission reduction supervision method based on artificial intelligence provided by the present invention comprises the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data preprocessing;

[0007] Step S3: Establishing a carbon emission reduction supervision model;

[0008] Step S4: Carbon emission reduction supervision.

[0009] Furthermore, in step S1, the data collection is to collect historical carbon emission reduction supervision data; the historical carbon emission reduction supervision data includes carbon emission data, energy consumption data, production data, equipment data, environmental data and carbon emission reduction assessment level; the higher the carbon emission reduction assessment level, the worse the carbon emission reduction effect; the carbon emission reduction assessment level is used as a data label.

[0010] Furthermore, in step S2, the data preprocessing is to perform data cleaning, data conversion and data set division on the collected data; the data cleaning is to process missing values, duplicate values ​​and outliers; the data conversion is to convert the data into vector form and perform standardization processing; the data set division is to randomly divide the converted data set into a test set and a training set.

[0011] Furthermore, in step S3, the establishment of the carbon emission reduction supervision model specifically includes the following steps:

[0012] Step S31: Design an activation function; the specific form is as follows:

[0013] ;

[0014] ;

[0015] ;

[0016] In the formula, is the activation function; and It is an auxiliary function; is the input variable of the activation function; q is a hyperparameter for controlling the nonlinearity of the activation function;

[0017] Step S32: verifying convergence; specifically including:

[0018] Step S321: define the Lyapunov function candidate as: ; Its derivative is expressed as: ;in, is the Lyapunov function; t is the time variable; is the error, corresponding to time t ; is the derivative of the Lyapunov function with respect to time t; k is a constant that adjusts the intensity of the error influence;

[0019] Step S322: Analysis ;for ,have ; ;but ;for ,have ; ;but ; Substitute get ;

[0020] Step S323: Convergence analysis; for linear models , the derivative of the Lyapunov function is ; To verify that convergence is faster, it is necessary to prove Right now For positive ,get For negative ,have , then it is proved that Right now ,have ;

[0021] Step S33: Design of the carbon emission reduction supervision model architecture; the carbon emission reduction supervision model includes: an input layer for receiving pre-processed historical carbon emission reduction supervision data; a convolution layer, the convolution layer uses a convolution kernel to extract local features from the input data, and the convolution kernel gradually convolves on the data through a sliding window operation to generate a feature map; a pooling layer, the pooling layer is used to reduce the dimension of the feature map, thereby reducing the computational complexity of the model; a fully connected layer, the fully connected layer receives high-level features from the convolution layer and the pooling layer, and maps these features to the output space; an output layer, the final output of the model is the classification result of the input data, that is, the carbon emission reduction assessment level; the designed activation function is used after the convolution layer and after the fully connected layer; the model uses a cross entropy loss function;

[0022] Step S34: model determination; the carbon emission reduction supervision model is trained based on the training set. If the loss of the carbon emission reduction supervision model on the training set converges, the training of the carbon emission reduction supervision model is completed; an accuracy threshold is set in advance. If the prediction accuracy of the trained carbon emission reduction supervision model on the test set is higher than the accuracy threshold, the carbon emission reduction supervision model is established. Otherwise, the model parameters are optimized and the data set is re-divided to train the carbon emission reduction supervision model;

[0023] Step S35: Model parameter optimization; specifically including:

[0024] Step S351: Initialization; establishing an optimization space based on the activation function hyperparameter q, the initial bias and initial weight of the carbon emission reduction supervision model; randomly generating a gene set, where each gene represents a parameter configuration of the model; each gene sequence represents the configuration of the bias and weight of the model, using numerical encoding; using the prediction accuracy of the carbon emission reduction supervision model trained based on the gene sequence for the test set as the individual fitness value; setting the maximum number of iterations;

[0025] Step S352: selection operation; using the roulette wheel selection algorithm, selecting individuals based on the adaptive scaling parameter and the fitness value; expressed as follows:

[0026] ;

[0027] ;

[0028] In the formula, is the probability of selecting the gth individual; is the fitness value of the g-th individual; α and β are the adaptive scaling parameter and the ranking scaling parameter respectively; is the descending ranking value of the g-th individual based on fitness value; T is the current number of iterations; Tmax is the maximum number of iterations; is a smoothing term;

[0029] Step S353: crossover operation; randomly select two parent gene sequences P1 and P2 from the current gene pool, and randomly select two crossover points in the parent gene sequences, i.e., determine the starting point and the end point of the gene fragment to be exchanged; extract the gene fragment located at the starting point and the end point from P1, find the position of this fragment in P2, delete it, and insert the fragment extracted from P1 into the corresponding position in P2; perform the same operation on P2;

[0030] Step S354: mutation operation; randomly select a gene sequence from the current gene pool as the mutation object; randomly select Nm positions in the gene sequence, and exchange the gene values ​​at the selected positions in pairs; after completing the exchange operation, the mutated gene is obtained, and the fitness of the mutated gene sequence is evaluated. If the fitness value after mutation is improved, the mutation is retained, otherwise the gene before mutation is returned;

[0031] Step S355: fine-tuning operation; select the gene sequence with the highest fitness value from the current gene pool as the optimal solution for the fine-tuning operation; define neighborhood operations, which include: insertion, randomly selecting a gene, removing it from its current position and reinserting it into another random position; swapping, randomly selecting several pairs of genes and swapping their positions; flipping, randomly selecting a gene fragment and reversing its order; fragment swapping, randomly selecting two non-overlapping gene fragments and swapping their positions; random swapping, randomly selecting several positions in the gene sequence and swapping the gene values ​​at these positions; fragment movement, randomly selecting a gene fragment and moving it to another random position in the sequence; define the number of adaptive fine-tuning operations as ;in, is the number of T-th adaptation fine-tuning operations, is the number of basic operations, is a parameter that controls the amplitude; the number of neighborhood operations that are performed during each iteration is the number of fine-tuning operations;

[0032] Step S356: Optimization judgment; a fitness threshold is set in advance; when there is an individual fitness value higher than the fitness threshold, a carbon emission reduction supervision model is established based on the parameters corresponding to the individual gene sequence, and the carbon emission reduction supervision model is now established; if the maximum number of iterations is reached, return to step S351; otherwise, return to step S352.

[0033] Furthermore, in step S4, the carbon emission reduction supervision is based on the established carbon emission reduction supervision model, and real-time carbon emission data, energy consumption data, production data, equipment data and environmental data are collected, and input into the carbon emission reduction supervision model after preprocessing. The carbon emission reduction assessment level output by the model is used as the assessment result of the real-time carbon emission reduction supervision data, and a level threshold is set in advance. If the carbon emission reduction assessment level output by the model is higher than the level threshold, early warning processing is performed.

[0034] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0035] (1) In view of the fact that the general carbon emission reduction supervision method has the problem that the model activation function cannot capture the complex nonlinear relationship in the carbon emission reduction data, cannot be applied to different carbon emission reduction scenarios, and has weak ability to handle outliers or extreme values ​​in the carbon emission reduction monitoring data, resulting in low accuracy of the carbon emission reduction monitoring model, this scheme designs the activation function by combining two different forms of exponential functions to more accurately simulate the nonlinear behavior in the carbon emission reduction process and improve the prediction and adjustment capabilities of the model; the introduction of the hyperparameter q allows the nonlinear degree of the activation function to be adjusted according to different carbon emission reduction scenarios and data characteristics, thereby optimizing the model performance; the segmented design of the activation function can ensure stability and effectiveness under different data conditions, and is used to handle outliers and extreme values ​​in carbon emission reduction monitoring data; the convergence speed enables the model to adapt to new carbon emission reduction data more quickly, thereby improving the accuracy of carbon emission reduction supervision.

[0036] (2) In view of the fact that general carbon emission reduction regulatory methods have poor ability to promote the generation of better solutions when optimizing the parameters of carbon emission reduction regulatory models, poor global optimization ability, low search efficiency and weak convergence, which lead to poor stability and low accuracy of carbon emission reduction regulation, this scheme introduces adaptive scaling parameters and ranking scaling parameters, so that the probability of individual selection can be adjusted according to the current iteration state when optimizing the parameters of the carbon emission reduction regulatory model, thereby promoting the generation of better solutions; by randomly selecting intersections and exchanging gene fragments, it can effectively maintain gene diversity and avoid population degradation, which helps the carbon emission reduction regulatory model to explore the parameter space more comprehensively and improve the globality of optimization; the fine-tuning operation performs local neighborhood operations on the optimal gene sequence, and the number of adaptive fine-tuning operations dynamically adjusts the fine-tuning intensity according to the number of iterations, so that when optimizing the parameters of the carbon emission reduction regulatory model, extensive exploration is carried out in the early stage and concentrated optimization is carried out in the later stage, ultimately achieving better convergence, and finally constructing a carbon emission reduction regulatory model with high accuracy and stability, thereby improving the accuracy of carbon emission reduction regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A schematic diagram of the process of the carbon emission reduction supervision method based on artificial intelligence provided by the present invention;

[0038] Figure 2 is a schematic diagram of the process of step S3;

[0039] Figure 3 It is a flowchart diagram of step S35.

[0040] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0042] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0043] Example 1, see Figure 1 The present invention provides an artificial intelligence-based carbon emission reduction supervision method, which comprises the following steps:

[0044] Step S1: Data collection, collecting historical carbon emission reduction supervision data;

[0045] Step S2: Data preprocessing, data cleaning, data conversion and data set division of the collected data;

[0046] Step S3: Establish a carbon emission reduction supervision model, design an activation function by combining two different forms of exponential functions, introduce a hyperparameter q to adjust the nonlinearity of the activation function; introduce an adaptive scaling parameter and a ranking scaling parameter; design a fine-tuning operation by performing a local neighborhood operation on the optimal gene sequence, and design the number of adaptive fine-tuning operations to dynamically adjust the fine-tuning intensity according to the number of iterations; thereby realizing the establishment of a carbon emission reduction supervision model;

[0047] Step S4: Carbon emission reduction supervision: perform carbon emission reduction supervision on real-time carbon emission reduction data based on the established carbon emission reduction supervision model.

[0048] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the historical carbon emission reduction supervision data includes carbon emission data, energy consumption data, production data, equipment data, environmental data and carbon emission reduction assessment level; the higher the carbon emission reduction assessment level, the worse the carbon emission reduction effect; the carbon emission reduction assessment level is used as a data label.

[0049] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, data cleaning is to process missing values, duplicate values ​​and outliers; data conversion is to convert data into vector form and perform standardization; data set partitioning is to randomly divide the converted data set into a test set and a training set.

[0050] Example 4, see Figure 1 and Figure 2 Based on the above embodiment, in step S3, establishing a carbon emission reduction supervision model specifically includes the following steps:

[0051] Step S31: Design an activation function; the specific form is as follows:

[0052] ;

[0053] ;

[0054] ;

[0055] In the formula, is the activation function; and It is an auxiliary function; is the input variable of the activation function; q is a hyperparameter for controlling the nonlinearity of the activation function;

[0056] Step S32: verifying convergence; specifically including:

[0057] Step S321: define the Lyapunov function candidate as: ; Its derivative is expressed as: ;in, is the Lyapunov function; t is the time variable; is the error, corresponding to time t ; is the derivative of the Lyapunov function with respect to time t; k is a constant that adjusts the intensity of the error influence;

[0058] Step S322: Analysis ;for ,have ; ;but ;for ,have ; ;but ; Substitute get ;

[0059] Step S323: Convergence analysis; for linear models , the derivative of the Lyapunov function is ; To verify that convergence is faster, it is necessary to prove Right now For positive ,get For negative ,have , then it is proved that Right now ,have ;

[0060] Step S33: Design of the carbon emission reduction supervision model architecture; the carbon emission reduction supervision model includes: an input layer for receiving pre-processed historical carbon emission reduction supervision data; a convolution layer, the convolution layer uses a convolution kernel to extract local features from the input data, and the convolution kernel gradually convolves on the data through a sliding window operation to generate a feature map; a pooling layer, the pooling layer is used to reduce the dimension of the feature map, thereby reducing the computational complexity of the model; a fully connected layer, the fully connected layer receives high-level features from the convolution layer and the pooling layer, and maps these features to the output space; an output layer, the final output of the model is the classification result of the input data, that is, the carbon emission reduction assessment level; the designed activation function is used after the convolution layer and after the fully connected layer; the model uses a cross entropy loss function;

[0061] Step S34: model determination; the carbon emission reduction supervision model is trained based on the training set. If the loss of the carbon emission reduction supervision model on the training set converges, the training of the carbon emission reduction supervision model is completed; an accuracy threshold is set in advance. If the prediction accuracy of the trained carbon emission reduction supervision model on the test set is higher than the accuracy threshold, the carbon emission reduction supervision model is established. Otherwise, the model parameters are optimized and the data set is re-divided to train the carbon emission reduction supervision model;

[0062] Step S35: Model parameter optimization.

[0063] By performing the above operations, the general carbon emission reduction supervision method has the problem that the model activation function cannot capture the complex nonlinear relationship in the carbon emission reduction data, cannot be applied to different carbon emission reduction scenarios, and has weak ability to handle outliers or extreme values ​​in the carbon emission reduction monitoring data, resulting in low accuracy of the carbon emission reduction monitoring model. This scheme designs the activation function by combining two different forms of exponential functions to more accurately simulate the nonlinear behavior in the carbon emission reduction process and improve the prediction and adjustment capabilities of the model; the introduction of the hyperparameter q allows the nonlinear degree of the activation function to be adjusted according to different carbon emission reduction scenarios and data characteristics, thereby optimizing the model performance; the segmented design of the activation function can ensure stability and effectiveness under different data conditions, and is used to handle outliers and extreme values ​​in carbon emission reduction monitoring data; the convergence speed enables the model to adapt to new carbon emission reduction data more quickly, thereby improving the accuracy of carbon emission reduction supervision.

[0064] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S35, the model parameters are optimized; specifically, the following steps are performed:

[0065] Step S351: Initialization; establishing an optimization space based on the activation function hyperparameter q, the initial bias and initial weight of the carbon emission reduction supervision model; randomly generating a gene set, where each gene represents a parameter configuration of the model; each gene sequence represents the configuration of the bias and weight of the model, using numerical encoding; using the prediction accuracy of the carbon emission reduction supervision model trained based on the gene sequence for the test set as the individual fitness value; setting the maximum number of iterations;

[0066] Step S352: selection operation; using the roulette wheel selection algorithm, selecting individuals based on the adaptive scaling parameter and the fitness value; expressed as follows:

[0067] ;

[0068] ;

[0069] In the formula, is the probability of selecting the gth individual; is the fitness value of the g-th individual; α and β are the adaptive scaling parameter and the ranking scaling parameter respectively; is the descending ranking value of the g-th individual based on fitness value; T is the current number of iterations; Tmax is the maximum number of iterations; is a smoothing term;

[0070] Step S353: crossover operation; randomly select two parent gene sequences P1 and P2 from the current gene pool, and randomly select two crossover points in the parent gene sequences, i.e., determine the starting point and the end point of the gene fragment to be exchanged; extract the gene fragment located at the starting point and the end point from P1, find the position of this fragment in P2, delete it, and insert the fragment extracted from P1 into the corresponding position in P2; perform the same operation on P2;

[0071] Step S354: mutation operation; randomly select a gene sequence from the current gene pool as the mutation object; randomly select Nm positions in the gene sequence, and exchange the gene values ​​at the selected positions in pairs; after completing the exchange operation, the mutated gene is obtained, and the fitness of the mutated gene sequence is evaluated. If the fitness value after mutation is improved, the mutation is retained, otherwise the gene before mutation is returned;

[0072] Step S355: fine-tuning operation; select the gene sequence with the highest fitness value from the current gene pool as the optimal solution for the fine-tuning operation; define neighborhood operations, which include: insertion, randomly selecting a gene, removing it from its current position and reinserting it into another random position; swapping, randomly selecting several pairs of genes and swapping their positions; flipping, randomly selecting a gene fragment and reversing its order; fragment swapping, randomly selecting two non-overlapping gene fragments and swapping their positions; random swapping, randomly selecting several positions in the gene sequence and swapping the gene values ​​at these positions; fragment movement, randomly selecting a gene fragment and moving it to another random position in the sequence; define the number of adaptive fine-tuning operations as ;in, is the number of T-th adaptation fine-tuning operations, is the number of basic operations, is a parameter that controls the amplitude; the number of neighborhood operations that are performed during each iteration is the number of fine-tuning operations;

[0073] Step S356: Optimization judgment; a fitness threshold is set in advance; when there is an individual fitness value higher than the fitness threshold, a carbon emission reduction supervision model is established based on the parameters corresponding to the individual gene sequence, and the carbon emission reduction supervision model is now established; if the maximum number of iterations is reached, return to step S351; otherwise, return to step S352.

[0074] By performing the above operations, the general carbon emission reduction supervision method has the problems of poor ability to promote the generation of better solutions when optimizing the parameters of the carbon emission reduction supervision model, poor global optimization ability, low search efficiency and weak convergence, which leads to poor stability and low accuracy of carbon emission reduction supervision. This scheme introduces adaptive scaling parameters and ranking scaling parameters, so that the probability of individual selection can be adjusted according to the current iteration state when optimizing the parameters of the carbon emission reduction supervision model, thereby promoting the generation of better solutions; by randomly selecting intersections and exchanging gene fragments, it can effectively maintain gene diversity and avoid population degradation, which is helpful for the carbon emission reduction supervision model to explore the parameter space more comprehensively and improve the globality of optimization; the fine-tuning operation performs local neighborhood operations on the optimal gene sequence, and the number of adaptive fine-tuning operations dynamically adjusts the fine-tuning intensity according to the number of iterations, so that the carbon emission reduction supervision model parameters are widely explored in the early stage and concentratedly optimized in the later stage, and finally achieves better convergence, and finally constructs a carbon emission reduction supervision model with high accuracy and stability, thereby improving the accuracy of carbon emission reduction supervision.

[0075] Example 6, see Figure 1This embodiment is based on the above embodiment. In step S4, carbon emission reduction supervision is based on the established carbon emission reduction supervision model. Carbon emission data, energy consumption data, production data, equipment data and environmental data are collected in real time. After preprocessing, they are input into the carbon emission reduction supervision model. The carbon emission reduction assessment level output by the model is used as the assessment result of the real-time carbon emission reduction supervision data. A level threshold is set in advance. If the carbon emission reduction assessment level output by the model is higher than the level threshold, early warning processing is performed.

[0076] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0077] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0078] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A carbon emission reduction supervision method based on artificial intelligence, characterized by: The method comprises the following steps: Step S1: Data collection, collecting historical carbon emission reduction supervision data; Step S2: data preprocessing; Step S3: Establish a carbon emission reduction supervision model, design an activation function by combining two different forms of exponential functions, introduce a hyperparameter q to adjust the nonlinearity of the activation function; introduce an adaptive scaling parameter and a ranking scaling parameter; design a fine-tuning operation by performing a local neighborhood operation on the optimal gene sequence, and design the number of adaptive fine-tuning operations to dynamically adjust the fine-tuning intensity according to the number of iterations; thereby realizing the establishment of a carbon emission reduction supervision model; Step S4: Carbon emission reduction supervision, performing carbon emission reduction supervision on real-time carbon emission reduction data based on the established carbon emission reduction supervision model; In step S3, the establishment of the carbon emission reduction supervision model specifically includes the following steps: Step S31: Design an activation function; the specific form is as follows: ; ; ; In the formula, is the activation function; and It is an auxiliary function; is the input variable of the activation function; q is a hyperparameter for controlling the nonlinearity of the activation function; Step S32: verifying the convergence of the activation function; Step S33: Design of the carbon emission reduction supervision model architecture; the carbon emission reduction supervision model includes: an input layer for receiving preprocessed historical carbon emission reduction supervision data; a convolution layer, the convolution layer uses a convolution kernel to extract local features from the input data, and the convolution kernel gradually convolves on the data through a sliding window operation to generate a feature map; a pooling layer, the pooling layer is used to reduce the dimension of the feature map; a fully connected layer, the fully connected layer receives high-level features from the convolution layer and the pooling layer, and maps these features to the output space; an output layer, the final output of the model is the classification result of the input data, that is, the carbon emission reduction assessment level; the designed activation function is used after the convolution layer and after the fully connected layer; the model uses a cross entropy loss function; Step S34: model determination; the carbon emission reduction supervision model is trained based on the training set. If the loss of the carbon emission reduction supervision model on the training set converges, the training of the carbon emission reduction supervision model is completed; an accuracy threshold is set in advance. If the prediction accuracy of the trained carbon emission reduction supervision model on the test set is higher than the accuracy threshold, the carbon emission reduction supervision model is established. Otherwise, the model parameters are optimized and the data set is re-divided to train the carbon emission reduction supervision model; Step S35: Model parameter optimization; specifically including: Step S351: Initialization; establishing an optimization space based on the activation function hyperparameter q, the initial bias and initial weight of the carbon emission reduction supervision model; randomly generating a gene set, where each gene represents a parameter configuration of the model; each gene sequence represents the configuration of the bias and weight of the model, using numerical encoding; using the prediction accuracy of the carbon emission reduction supervision model trained based on the gene sequence for the test set as the individual fitness value; setting the maximum number of iterations; Step S352: selection operation; using the roulette wheel selection algorithm, selecting individuals based on the adaptive scaling parameter and the fitness value; expressed as follows: ; ; In the formula, is the probability of selecting the gth individual; is the fitness value of the g-th individual; α and β are the adaptive scaling parameter and the ranking scaling parameter respectively; is the descending ranking value of the g-th individual based on fitness value; T is the current number of iterations; Tmax is the maximum number of iterations; is a smoothing term; Step S353: crossover operation; randomly select two parent gene sequences P1 and P2 from the current gene pool, and randomly select two crossover points in the parent gene sequences, i.e., determine the starting point and the end point of the gene fragment to be exchanged; extract the gene fragment located at the starting point and the end point from P1, find the position of this fragment in P2, delete it, and insert the fragment extracted from P1 into the corresponding position in P2; perform the same operation on P2; Step S354: mutation operation; randomly select a gene sequence from the current gene pool as the mutation object; randomly select Nm positions in the gene sequence, and exchange the gene values ​​at the selected positions in pairs; after completing the exchange operation, the mutated gene is obtained, and the fitness of the mutated gene sequence is evaluated. If the fitness value after mutation is improved, the mutation is retained, otherwise the gene before mutation is returned; Step S355: fine-tuning operation; select the gene sequence with the highest fitness value from the current gene pool as the optimal solution for the fine-tuning operation; define neighborhood operations, which include: insertion, randomly selecting a gene, removing it from its current position and reinserting it into another random position; swapping, randomly selecting several pairs of genes and swapping their positions; flipping, randomly selecting a gene fragment and reversing its order; fragment swapping, randomly selecting two non-overlapping gene fragments and swapping their positions; random swapping, randomly selecting several positions in the gene sequence and swapping the gene values ​​at these positions; fragment movement, randomly selecting a gene fragment and moving it to another random position in the sequence; define the number of adaptive fine-tuning operations as ;in, is the number of adaptation fine-tuning operations at the Tth iteration, is the number of basic operations, is a parameter that controls the amplitude; the number of neighborhood operations that are performed during each iteration is the number of fine-tuning operations; Step S356: optimization judgment; a fitness threshold is set in advance; when there is an individual fitness value higher than the fitness threshold, a carbon emission reduction supervision model is established based on the parameters corresponding to the individual gene sequence, and the carbon emission reduction supervision model is now established; if the maximum number of iterations is reached, return to step S351; otherwise, return to step S352.

2. The method for monitoring carbon emission reduction based on artificial intelligence according to claim 1 is characterized in that: In step S1, the historical carbon emission reduction supervision data includes carbon emission data, energy consumption data, production data, equipment data, environmental data and carbon emission reduction assessment level; the higher the carbon emission reduction assessment level, the worse the carbon emission reduction effect; the carbon emission reduction assessment level is used as a data label.

3. The method for monitoring carbon emission reduction based on artificial intelligence according to claim 1 is characterized in that: In step S4, the carbon emission reduction supervision is based on the established carbon emission reduction supervision model, and real-time carbon emission data, energy consumption data, production data, equipment data and environmental data are collected. After pre-processing, they are input into the carbon emission reduction supervision model. The carbon emission reduction assessment level output by the model is used as the assessment result of the real-time carbon emission reduction supervision data. A level threshold is set in advance. If the carbon emission reduction assessment level output by the model is higher than the level threshold, early warning processing is performed.

4. The method for monitoring carbon emission reduction based on artificial intelligence according to claim 1 is characterized in that: In step S2, the data preprocessing is to clean the collected data, convert the data and divide the data set; the data cleaning is to process missing values, duplicate values ​​and abnormal values; The data conversion is to convert the data into vector form and perform standardization processing; the data set division is to randomly divide the data set after data conversion into a test set and a training set.