A carbon neutrality data processing method based on artificial intelligence

Abstract

The present invention relates to the field of carbon neutrality data processing technology, and in particular to a carbon neutrality processing method based on artificial intelligence. The method comprises the following steps: obtaining carbon emission related data, and performing data preprocessing on the carbon emission related data to obtain carbon emission data to be denoised; using an intelligent denoising algorithm to perform denoising on the carbon emission data to be denoised to obtain carbon emission denoised data; using a historical data tracing algorithm to perform historical tracing on the carbon emission denoised data to obtain carbon emission historical data; using a preset artificial intelligence-based multi-algorithm prediction model to predict and analyze the carbon emission historical data to obtain the initial results of carbon emission factor prediction; according to the initial results of carbon emission factor prediction, a weight balance algorithm is used to perform model fusion processing on the multi-algorithm prediction model to generate a fusion prediction model. The present invention continuously improves the carbon neutrality processing process by using artificial intelligence algorithms to monitor, analyze and optimize carbon emissions.

Description

Technical Field

[0001] The present invention relates to the field of carbon neutrality data processing technology, and in particular to a carbon neutrality processing method based on artificial intelligence. Background Art

[0002] With the development of the global economy and the improvement of human living standards, carbon emissions have become one of the important issues facing the world. In addition, with the rapid development of industrialization and urbanization, global carbon dioxide emissions have continued to increase, causing serious impacts on global climate change. In order to reduce greenhouse gas emissions, many countries, organizations and companies are actively promoting carbon neutrality technology, that is, removing carbon dioxide from the atmosphere or converting it into harmless substances through various means. At present, carbon neutrality technology mainly includes afforestation, marine carbon sinks, land management, carbon capture and storage, etc. However, these traditional carbon neutralization treatment methods have problems such as low efficiency and high cost. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide an artificial intelligence-based carbon neutrality data processing method to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above purpose, a carbon neutrality data processing method based on artificial intelligence includes the following steps:

[0005] Step S1: obtaining carbon emission related data, and performing data preprocessing on the carbon emission related data to obtain carbon emission noise reduction data; performing noise reduction processing on the carbon emission noise reduction data using an intelligent noise reduction algorithm to obtain carbon emission noise reduction data;

[0006] Step S2: Use the historical data tracing algorithm to perform historical tracing processing on the carbon emission noise reduction data to obtain the carbon emission historical data; use the preset artificial intelligence-based multi-algorithm prediction model to perform prediction and analysis on the carbon emission historical data to obtain the initial results of carbon emission factor prediction;

[0007] Step S3: Based on the initial prediction results of carbon emission factors, the multi-algorithm prediction models are fused using a weight balance algorithm to generate a fusion prediction model; and the carbon emission historical data are re-predicted and analyzed using the fusion prediction model to obtain the carbon emission factor prediction results;

[0008] Step S4: Formulate a corresponding carbon neutrality treatment plan based on the prediction results of carbon emission factors, and use the carbon neutrality treatment plan to execute the corresponding carbon neutrality treatment technology to obtain carbon neutrality treatment data; use the carbon neutrality monitoring model based on reinforcement learning to perform real-time monitoring and optimization processing on the carbon neutrality treatment data to obtain carbon neutrality treatment optimization results;

[0009] Step S5: Use adaptive decision-making technology to intelligently adjust the carbon neutrality processing optimization results to obtain a carbon neutrality adjustment decision; formulate an adjustment processing strategy based on the carbon neutrality adjustment decision, introduce a feedback mechanism, and use the adjustment processing strategy to improve and optimize the carbon neutrality processing process.

[0010] The present invention obtains carbon emission related data by using sensors and other data collection means, and performs data preprocessing on the obtained carbon emission related data, such as data cleaning, deduplication, normalization and other operations to eliminate unnecessary noise and interference. The preprocessed carbon emission noise reduction data is subjected to noise reduction processing using a suitable intelligent noise reduction algorithm to improve the accuracy and reliability of the carbon emission noise reduction data. The noise reduction carbon emission noise reduction data is historically restored using a historical data tracing algorithm to obtain more real and accurate carbon emission historical data. The carbon emission historical data is analyzed and processed by using a multi-algorithm prediction model based on artificial intelligence to obtain the initial results of carbon emission factor prediction. Then, according to the initial results of factor prediction, a weight balance algorithm is used to perform model fusion processing on multiple prediction models to generate a more accurate and stable fusion model. The carbon emission historical data is re-predicted and analyzed by using a fusion model to obtain a more accurate carbon emission factor prediction result. A carbon neutralization treatment plan is formulated according to the carbon emission factor prediction results, and different carbon neutralization treatment technologies are used to perform corresponding processing to achieve the effect of optimizing the carbon neutralization treatment process. By using the carbon neutrality monitoring model based on reinforcement learning to monitor and optimize the carbon neutrality processing data in real time, a more optimized and effective carbon neutrality processing optimization result can be obtained. Finally, by using adaptive decision-making technology to intelligently adjust the carbon neutrality processing optimization results, a more accurate and optimized carbon neutrality adjustment decision can be obtained, and corresponding adjustment processing strategies can be formulated according to the carbon neutrality adjustment decision. The carbon neutrality processing process is improved and optimized by introducing a feedback mechanism to continuously improve the effect and efficiency of carbon neutrality processing.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: Real-time monitoring and collection of carbon emission sources are performed through sensors to obtain measured carbon emission data;

[0013] Step S12: performing data cleaning processing on the carbon emission measured data to obtain carbon emission data to be denoised;

[0014] Step S13: using an intelligent noise reduction algorithm to perform noise reduction processing on the carbon emission noise reduction data to obtain carbon emission noise reduction data.

[0015] The present invention uses sensors to monitor and collect carbon emission sources in real time to obtain carbon emission measured data. The sensor can collect real-time monitored carbon emission measured data according to different measurement methods and detection principles. Then, the carbon emission measured data collected by the sensor is cleaned. On the one hand, the carbon emission measured data collected by the sensor may be subject to various interferences, such as sensor failure, signal noise, etc., so it is necessary to remove and filter the carbon emission measured data; on the other hand, the carbon emission measured data can also be detected and repaired for outliers to ensure the quality and accuracy of the carbon emission measured data. After data cleaning, cleaner and more accurate carbon emission noise reduction data can be obtained. Finally, the obtained carbon emission noise reduction data is subjected to noise reduction by setting a suitable intelligent noise reduction algorithm to accurately calculate the amplitude and spatial distribution characteristics of the noise. On this basis, the carbon emission noise reduction data is subjected to noise reduction by filtering and other methods to eliminate the influence of noise on the carbon emission noise reduction data, thereby improving the reliability and accuracy of the carbon emission noise reduction data. After intelligent noise reduction, high-quality carbon emission noise reduction data can be obtained.

[0016] Preferably, step S13 comprises the following steps:

[0017] Step S131: using an intelligent noise reduction algorithm to calculate the noise value of the carbon emission data to be denoised, to obtain the carbon emission noise value;

[0018]

[0019] In the formula, e(x) is the carbon emission noise value, x is the carbon emission data set to be denoised, n is the number of Gaussian distributions, and a is the i is the weight of the i-th Gaussian distribution, σ i is the standard deviation of the ith Gaussian distribution, μ i is the mean of the i-th Gaussian distribution, y is the noisy carbon emission data to be denoised, is the l2 difference norm between the noisy carbon emission denoising data y and the carbon emission denoising data set x, m is the number of noise pulse signals in the carbon emission denoising data, c j is the amplitude value of the jth noise pulse signal in the carbon emission data to be denoised, x j is the position of the jth noise pulse signal in the carbon emission data to be de-noised, δ(x―x j ) is the Dirac pulse function, w(t-x) is the noise weight function of the adjacent carbon emission data to be de-noised in the time domain, t is the time variable in the time domain, and ε is the correction value of the carbon emission noise value;

[0020] The present invention constructs a formula of an intelligent denoising algorithm function, which is used to calculate the noise value of the carbon emission data to be denoised. In order to eliminate the influence of the noise source in the carbon emission data to be denoised on the subsequent construction process of the multi-algorithm prediction model, it is necessary to perform denoising on the carbon emission data to be denoised to obtain cleaner and more accurate carbon emission data to be denoised. The intelligent denoising algorithm can effectively remove the noise and interference data in the carbon emission data to be denoised, thereby improving the accuracy and reliability of the carbon emission data to be denoised. The algorithm function formula fully considers the carbon emission data set x to be denoised, the number of Gaussian distributions n, the weight a of the i-th Gaussian distribution i , the standard deviation σ of the i-th Gaussian distribution i , the mean μ of the i-th Gaussian distribution i , the noisy carbon emission denoising data y, the l2 difference norm between the noisy carbon emission denoising data y and the carbon emission denoising data set x The number m of noise pulse signals in the carbon emission data to be de-noised, the amplitude value c of the jth noise pulse signal in the carbon emission data to be de-noised j , the position x of the jth noise pulse signal in the carbon emission data to be denoised j , Dirac Delta function δ(x―x j ), the noise weight function w(t-x) of the adjacent carbon emission data to be de-noised in the time domain, the time variable t in the time domain, and the correlation between the carbon emission noise value e(x) and the above parameters form a functional relationship The algorithm function formula realizes the calculation of the noise value of the carbon emission data to be denoised. At the same time, the correction value ε of the carbon emission noise value in the algorithm function formula can be adjusted according to actual conditions, thereby improving the accuracy and applicability of the intelligent denoising algorithm.

[0021] Step S132: judging the carbon emission noise value according to a preset carbon emission noise threshold, and when the carbon emission noise value is greater than or equal to the preset carbon emission noise threshold, removing the carbon emission noise reduction data corresponding to the carbon emission noise value to obtain the carbon emission noise reduction data;

[0022] Step S133: the carbon emission noise value is judged according to the preset carbon emission noise threshold. When the carbon emission noise value is less than the preset carbon emission noise threshold, the carbon emission noise reduction data corresponding to the carbon emission noise value is directly defined as the carbon emission noise reduction data.

[0023] Since the present invention may have noise interference and outliers in the acquired carbon emission data to be de-noised, which may have an adverse effect on the accuracy and reliability of the subsequent model building work, a suitable intelligent noise reduction algorithm is set to calculate the noise value of the carbon emission data to be de-noised, so that the noise and interference signals existing in the carbon emission data to be de-noised can be identified and measured, and the noise signal can be removed from the source, and the signal-to-noise ratio of the carbon emission data to be de-noised can be enhanced, thereby improving the accuracy and reliability of the carbon emission data to be de-noised. The intelligent noise reduction algorithm optimizes the noise reduction process by combining the weight of the Gaussian distribution, the standard deviation of the Gaussian distribution, the mean of the Gaussian distribution, the l2 difference norm, the Dirac pulse function, and the noise weight function, and adjusts and optimizes the intelligent noise reduction algorithm by setting appropriate noise pulse signal parameters and related noise reduction function parameters to obtain the best noise reduction effect and calculation results, thereby more accurately calculating the carbon emission noise value. Then, according to the specific data processing requirements and quality standards, an appropriate carbon emission noise threshold is set to judge the calculated carbon emission noise value, which can effectively eliminate the carbon emission noise reduction data with large carbon emission noise values, avoid the impact of these carbon emission noise reduction data with large noise values ​​on the overall data, and help to further improve the quality of the data, reduce unnecessary interference and errors, thereby ensuring the accuracy and reliability of the carbon emission noise reduction data. Finally, the carbon emission noise value is judged using the preset carbon emission noise threshold, and the carbon emission noise reduction data with small carbon emission noise values ​​are defined as carbon emission noise reduction data, which can obtain more accurate and reliable carbon emission noise reduction data, which are less affected by noise and can provide a more stable data basis for subsequent multi-algorithm prediction models, thereby improving the availability and effectiveness of carbon emission noise reduction data.

[0024] Preferably, step S2 comprises the following steps:

[0025] Step S21: tracing the carbon emission noise reduction data historically using a historical data tracing algorithm to obtain carbon emission historical data;

[0026] Step S22: preprocessing the carbon emission historical data using feature extraction and conversion technology to obtain a carbon emission historical feature data set;

[0027] Step S23: Predictive analysis is performed on the historical characteristic data set of carbon emissions using a preset artificial intelligence-based multi-algorithm prediction model to obtain an initial result of carbon emission factor prediction.

[0028] The present invention uses an appropriate historical data tracing algorithm to trace the carbon emission noise reduction data to obtain accurate carbon emission historical data, which can ensure that subsequent predictions and analyses are based on real historical data. Then, the carbon emission historical data is processed by using feature extraction and conversion technology to obtain a carbon emission historical feature data set. Feature extraction can convert carbon emission historical data into feature information data that can be used for modeling and prediction, so as to facilitate subsequent predictions and analysis. Finally, by using a preset multi-algorithm prediction model based on artificial intelligence, the carbon emission historical feature data set is predictively analyzed and processed to obtain the initial results of carbon emission factor prediction. Through the combined use of multi-algorithm prediction models, the influence of multiple factors can be comprehensively considered to obtain more accurate initial results of carbon emission factor prediction, which can provide important reference and data support for the implementation of subsequent carbon neutralization processing optimization processes.

[0029] Preferably, the historical data tracing algorithm function formula in step S21 is specifically:

[0030]

[0031]

[0032] Where D(T) is the historical data of carbon emissions at the historical tracing time T, N is the number of tracing exponential decay functions, τ is the integral time component, and α k is the attenuation oscillation amplitude of the kth traceable exponential decay function, β k is the decay rate of the kth traceable exponential decay function, γ k is the tracing time control parameter of the kth tracing exponential decay function, exp is the exponential function, M is the number of tracing discontinuous functions, and f r (u r (T―τ)) is the rth traceable discontinuous function, u r (T―τ) is the time step decay function corresponding to the rth traceable discontinuous function, is the tracing time control function of the historical tracing time T, R is the number of historical tracing in the carbon emission noise reduction data, p l is the lth carbon emission noise reduction data, ξ is the traceability ratio of carbon emission noise reduction data per unit time, T0 is the initial time of historical traceability, T R is the end time of historical tracing, and η is the correction value of historical carbon emission data.

[0033] The present invention constructs a formula of a historical data tracing algorithm function, which is used for historical tracing of carbon emission noise reduction data. The historical data tracing algorithm analyzes and processes the historical tracing information of carbon emission noise reduction data, extracts the implicit laws and trends behind the carbon emission noise reduction data, and thus provides strong data support for the prediction of future trends. The core of the historical data tracing algorithm function is the tracing exponential decay function, which describes the decay law of historical data. The tracing exponential decay function is composed of the superposition of several exponential functions and the setting of appropriate decay oscillation amplitude value, decay rate and tracing time control parameters. In addition to the tracing exponential decay function, the historical data tracing algorithm function also includes a discontinuous function. Compared with the tracing exponential decay function, the discontinuous function is more in line with the actual situation. In the actual application process, historical data is often not completely smooth, and some discontinuous situations are likely to occur. Through the introduction of the discontinuous function, the marker function can be used to describe the mutation points in the data, so as to more accurately describe the evolution law of historical data. In addition, in the historical data traceability algorithm, the traceability time of historical data can also be modeled through the traceability time control function, which is related to the number of samples of historical data and the frequency of traceability. The historical data traceability algorithm can extract the inherent regularity information of the data through in-depth mining of historical data, providing an important reference for future predictions. The algorithm function formula fully considers the number N of traceability exponential decay functions, the integral time component τ, and the attenuation oscillation amplitude value α of the kth traceability exponential decay function. k , the decay rate β of the kth traceable exponential decay function k , the tracing time control parameter γ of the kth tracing exponential decay function k , the number of traceable discontinuous functions M, the rth traceable discontinuous function f r (u r (T―τ)), the time step decay function u corresponding to the rth traceable discontinuous function r (T―τ), the traceability time control function of the historical traceability time T The number of historical traceability in carbon emission noise reduction data R, the lth carbon emission noise reduction data p l , the carbon emission noise reduction data traceability ratio value ξ per unit time, the initial time of historical traceability T0, the number of historical traceability through carbon emission noise reduction data R, the lth carbon emission noise reduction data p l , the carbon emission noise reduction data traceability ratio value ξ per unit time, the historical traceability initial time T0 and the historical traceability time T constitute the traceability time control function relation According to the relationship between the historical carbon emission data D(T) at the historical traceability time T and the above parameters, a functional relationship is formed: The algorithm function formula realizes the historical traceability of carbon emission noise reduction data. At the same time, by introducing the correction value η of carbon emission historical data, it can be adjusted according to actual conditions, thereby improving the accuracy and applicability of the historical data traceability algorithm.

[0034] Preferably, step S23 includes the following steps:

[0035] Step S231: construct a multi-algorithm prediction model based on an artificial intelligence algorithm, wherein the multi-algorithm prediction model includes a neural network prediction model, a support vector machine classification model, a logistic regression prediction model, and a decision tree prediction model;

[0036] Step S232: using a neural network prediction model to perform autonomous learning prediction processing on the carbon emission historical characteristic data set to obtain a carbon emission factor prediction data set;

[0037] Step S233: using a support vector machine classification model to perform factor classification processing on the carbon emission factor prediction data set to obtain a carbon emission factor type data set;

[0038] Step S234: using a logistic regression prediction model to perform factor weight prediction on the carbon emission factor type data set to obtain a carbon emission factor weight data set;

[0039] Step S235: Use the decision tree prediction model to perform weighted tree classification processing on the carbon emission factor weight data set to obtain the initial results of carbon emission factor prediction.

[0040] The present invention constructs a multi-algorithm prediction model by utilizing a variety of artificial intelligence algorithms, and the multi-algorithm prediction model includes a neural network prediction model, a support vector machine classification model, a logistic regression prediction model, and a decision tree prediction model. The multi-algorithm prediction model comprehensively considers the impact of multiple factors on carbon emissions, and can predict and optimize according to the historical characteristic data of carbon emissions. The combined use of these artificial intelligence algorithm models can improve the accuracy and reliability of the prediction. The carbon emission historical characteristic data set is processed by using a neural network prediction model to obtain a carbon emission factor prediction data set. The neural network can explore potential laws and patterns from historical data through autonomous learning, and can process multi-dimensional characteristic data to obtain more accurate prediction results. Then, the carbon emission factor prediction data set is subjected to factor classification processing using a support vector machine classification model to obtain a carbon emission factor type data set. The support vector machine can classify the data into different categories according to the characteristics of the data attributes, so that the factors can be further classified and processed to provide more information for subsequent prediction and optimization. The carbon emission factor type data set is subjected to factor weight prediction using a logistic regression prediction model to obtain a carbon emission factor weight data set. The logistic regression model can predict the impact of different factors on carbon emissions by analyzing the factor type data set, so as to determine the weight of each factor and provide more accurate information for subsequent optimization. Finally, the decision tree prediction model is used to perform weight tree classification on the carbon emission factor weight data set to obtain the initial results of carbon emission factor prediction. The decision tree model can classify and process the weight data set in more detail to obtain more accurate prediction results. Accurate carbon emission factor prediction results are obtained through the multi-algorithm prediction model, providing accurate data support for subsequent carbon neutralization processing.

[0041] Preferably, the weight balancing algorithm function formula in step S3 is specifically:

[0042]

[0043]

[0044] Where V(θ) is the weight balancing algorithm function, θ is the time variable of the model fusion process, ω q is the weight coefficient of the qth algorithm model in the multi-algorithm prediction model, is the control granularity parameter of the time exponential function, b is the time variable harmonic smoothing parameter, F q is the prediction function of the qth algorithm model in the multi-algorithm prediction model, λ q is the harmonic smoothing parameter of the prediction function of the qth algorithm model in the multi-algorithm prediction model, J is the number of parameters of the prediction function, φ q,g is the gth parameter value of the prediction function of the qth algorithm model in the multi-algorithm prediction model, Xθ―g is the model input data of the multi-algorithm prediction model at time θ―g, exp is the exponential function, X q is the model input data of the qth algorithm model in the multi-algorithm prediction model, X′ q is the model output data of the qth algorithm model in the multi-algorithm prediction model, d q (X q ,X′ q ) is the model input data X of the qth algorithm model in the multi-algorithm prediction model q With the model output data X′ q The distance function between is the weight controlling smoothing parameter, It is the correction value of the weight balancing algorithm function.

[0045] The present invention constructs a formula of a weighted balance algorithm function for model fusion processing of multiple algorithm prediction models. The weighted balance algorithm is a fusion technology for multiple algorithm prediction models. By assigning different weight coefficients to each algorithm model, it is fused into a more accurate and robust fusion prediction model, thereby improving the accuracy and stability of the prediction results. The weighted balance algorithm function contains several important parameters, among which the weight coefficient determines the contribution of each algorithm model to the final result. The weight coefficient of each algorithm model is determined according to its prediction performance and error index. The better the performance of the algorithm model, the larger the corresponding weight coefficient. The weighted balance algorithm function is a function that integrates the prediction results of multiple algorithm models. It can control the process and results of model fusion through multiple factors such as weight coefficients, prediction functions, and harmonic smoothing parameters. In the application of carbon emission factor prediction, the function can be used to fuse the results of multiple prediction models, thereby improving the prediction accuracy and reliability. The algorithm function formula fully considers the model fusion process time variable θ, the weight coefficient ω of the qth algorithm model in the multi-algorithm prediction model q , the time exponential function controls the granularity parameter The time variable harmonic smoothing parameter b, the prediction function F of the qth algorithm model in the multi-algorithm prediction model q , the harmonic smoothing parameter λ of the prediction function of the qth algorithm model in the multi-algorithm prediction model q , the number of parameters J of the prediction function, the gth parameter value φ of the prediction function of the qth algorithm model in the multi-algorithm prediction model q,g , the model input data X of the multi-algorithm prediction model at time θ―g θ―g , the model input data X of the qth algorithm model in the multi-algorithm prediction model q , the model output data X′ of the qth algorithm model in the multi-algorithm prediction model q , the model input data X of the qth algorithm model in the multi-algorithm prediction modelq With the model output data X′ q The distance function d q (X q ,X′ q ), weight controls the smoothing parameter Correction value of weight balancing algorithm function The weight coefficient is the core of the weight balance algorithm function, and the model input data X of the qth algorithm model in the multi-algorithm prediction model is q , the model output data X′ of the qth algorithm model in the multi-algorithm prediction model q , the model input data X of the qth algorithm model in the multi-algorithm prediction model q With the model output data X′ q The distance function d q (X q ,X′ q ) and weights to control the smoothing parameter A functional relationship According to the relationship between the weight balance algorithm function V(θ) and the above parameters, a functional relationship is formed: The algorithm function formula realizes the model fusion of multiple algorithm prediction models. At the same time, the correction value of the algorithm function is balanced by weight. The introduction of can be used to adjust special situations that occur during the model fusion process, further improving the applicability and stability of the weight balancing algorithm function, thereby improving the generalization ability and robustness of the fusion prediction model.

[0046] Preferably, step S4 comprises the following steps:

[0047] Step S41: formulating a corresponding carbon neutralization processing scheme according to the carbon emission factor prediction result, and using the carbon neutralization processing scheme to perform a corresponding carbon neutralization processing technology on the carbon emission noise reduction data to obtain carbon neutralization processing data;

[0048] Step S42: performing data preprocessing on the carbon neutrality processing data to obtain a carbon neutrality processing data set;

[0049] Step S43: Use the carbon neutrality monitoring model based on reinforcement learning to perform real-time monitoring and optimization processing on the carbon neutrality processing data set to obtain the carbon neutrality processing optimization results.

[0050] The present invention formulates a corresponding carbon neutralization treatment scheme according to the prediction results of carbon emission factors, and performs corresponding carbon neutralization treatment technology on carbon emission noise reduction data by using the formulated carbon neutralization treatment scheme. Through the application of carbon neutralization treatment technology, carbon emissions can be reduced to a certain extent, thereby obtaining carbon neutralization treatment data. Then, the obtained carbon neutralization treatment data is preprocessed so that it can be subsequently input into the carbon neutralization monitoring model for analysis and optimization. The purpose of data preprocessing is to remove duplicate invalid data, make the carbon neutralization treatment data cleaner and easier to analyze, and at the same time, feature selection and conversion operations can be performed to improve the accuracy and generalization ability of the carbon neutralization monitoring model. After data preprocessing, a carbon neutralization treatment data set suitable for monitoring and analysis can be obtained. Finally, by using a carbon neutralization monitoring model based on reinforcement learning to perform real-time monitoring and optimization of the carbon neutralization treatment data set, the optimization result of the carbon neutralization treatment process can be obtained. By real-time updating and adjusting the carbon neutralization treatment process of the carbon neutralization monitoring model, the carbon neutralization treatment process can be optimized and controlled in real time, thereby obtaining a better carbon neutralization treatment effect, and having higher real-time performance and carbon neutralization treatment efficiency.

[0051] Preferably, step S43 includes the following steps:

[0052] Step S431: dividing the carbon neutrality processing data set into a carbon neutrality training data set, a carbon neutrality verification data set, and a carbon neutrality test data set according to a preset division rule;

[0053] Step S432: constructing a carbon neutrality monitoring model based on reinforcement learning, wherein the carbon neutrality monitoring model includes model training, model verification and model evaluation;

[0054] Step S433: inputting the carbon neutrality training data set into the carbon neutrality monitoring model based on reinforcement learning for model training, and tuning the model parameters by monitoring the loss function to generate a verification model; inputting the carbon neutrality verification data set into the verification model for model verification to generate a test model;

[0055] Among them, the monitoring loss function is as follows:

[0056]

[0057] Where L(θ) is the monitoring loss function, θ is the carbon neutrality monitoring model parameter, H is the number of training rounds for model training, and s h is the hth model training round, ρ(s h ) is the time weight coefficient of the hth model training round, z is the CO2 concentration, z max is the maximum CO2 concentration in the carbon neutrality training dataset, is the CO2 concentration disturbance control space, is the disturbance space variable, is the h-th model training round s h , CO2 concentration z and disturbance space The predicted output result of the model parameter θ under the condition of is the actual output of the model, is the number of rounds s of model training at a given hth time h , CO2 concentration z and disturbance space The probability density function of the actual output result of the model under the condition of , ∈ is the correction value of the monitoring loss function;

[0058] The present invention constructs a formula of a monitoring loss function for tuning the parameters of a carbon neutrality monitoring model. When training a carbon neutrality training data set by using a carbon neutrality monitoring model, in order to help the carbon neutrality monitoring model fit the data as much as possible, it is necessary to use a suitable monitoring loss function as an indicator for model parameter optimization. The main function of the monitoring loss function is to monitor the output results of the carbon neutrality monitoring model, and then adjust the model parameters according to the monitoring results to improve the prediction accuracy and stability of the carbon neutrality monitoring model. The monitoring loss function is an integral form of formula, which requires comparing the prediction results and actual output results of the carbon neutrality monitoring model, and taking into account the disturbance factor of CO2 concentration. In the monitoring loss function, the time weight coefficient is used to weight the model training effect of the historical training rounds, so that the more recent training results have a greater impact on the carbon neutrality monitoring model. At the same time, the introduction of disturbance space variables can increase the adaptability of the carbon neutrality monitoring model to abnormal situations, thereby improving the robustness of the carbon neutrality monitoring model. When the predicted results of the carbon neutrality monitoring model are inconsistent with the actual output results, the value of the monitoring loss function will increase, thereby guiding the carbon neutrality monitoring model parameters to be adjusted in a more optimized direction. Therefore, the monitoring loss function improves the training effect and stability of the carbon neutrality monitoring model from multiple aspects by introducing the CO2 concentration disturbance factor, weighting the historical training effect, and taking into account the actual output results, so that it can adapt to the input data in different scenarios and accurately monitor and effectively control the carbon neutrality processing data set to be monitored. The algorithm function formula fully considers the carbon neutrality monitoring model parameters θ, the number of training rounds H for model training, and the hth model training round s h , the time weight coefficient ρ(s h ), CO2 concentration z, the maximum CO2 concentration z in the carbon neutral training dataset max , CO2 concentration disturbance control space Perturbation space variables In the hth model training round s h , CO2 concentration z and disturbance space The predicted output of the model parameter θ under the condition The actual output of the model At a given h-th model training round s h The probability density function of the actual output of the model under the conditions of CO2 concentration z and disturbance space g According to the monitoring loss function L(θ), a functional relationship is formed The algorithm function formula realizes the tuning of the parameters of the carbon neutrality monitoring model. At the same time, by introducing the correction value ∈ of the monitoring loss function, adjustments can be made to special situations that arise during model training, further improving the applicability and stability of the monitoring loss function, thereby improving the generalization ability and robustness of the carbon neutrality monitoring model.

[0059] Step S434: Input the carbon neutrality test data set into the test model for model evaluation to obtain an optimized carbon neutrality monitoring model; re-input the carbon neutrality treatment data set into the optimized carbon neutrality monitoring model for real-time monitoring, and optimize the carbon emission sources in the carbon neutrality treatment data set by setting up an incentive mechanism to obtain carbon neutrality treatment optimization results.

[0060] The present invention divides the carbon neutrality processing data set by a preset division rule, and can better utilize the carbon neutrality processing data set to train, verify and evaluate the carbon neutrality monitoring model. At the same time, by separating the carbon neutrality processing data set, overfitting of the carbon neutrality monitoring model and improving the generalization ability of the carbon neutrality monitoring model can be avoided. By using a method based on reinforcement learning, a carbon neutrality monitoring model suitable for carbon neutrality processing optimization is constructed. By performing model training, model verification and model evaluation on the carbon neutrality monitoring model, all-round monitoring and optimization of the carbon neutrality monitoring model can be achieved. Then, by inputting the carbon neutrality training data set and the carbon neutrality verification data set, and adaptively adjusting the model parameters through the calculation of the appropriate monitoring loss function, the carbon neutrality monitoring model can more accurately predict the carbon neutrality effect. Through cyclic iterative training, verification and tuning, a more optimized test model can be generated for subsequent monitoring and optimization of the carbon neutrality processing process. Finally, by inputting the carbon neutrality test data set, the generated test model is evaluated to obtain an optimized carbon neutrality monitoring model. The carbon neutrality monitoring model can be used to monitor the carbon neutrality treatment process in real time, and optimize the carbon emission source by using the set reward mechanism to obtain a better carbon neutrality treatment effect, which can further improve the real-time and efficiency of the carbon neutrality treatment process. Using the carbon neutrality monitoring model based on reinforcement learning, through the steps of carbon neutrality treatment data set division, model training, model verification and model evaluation, the carbon neutrality treatment process can be more effectively monitored and optimized, thereby achieving a more complete carbon neutrality treatment effect.

[0061] Preferably, step S5 comprises the following steps:

[0062] Step S51: using adaptive decision-making technology to intelligently adjust and evaluate the carbon neutrality processing optimization results to obtain a carbon neutrality adjustment decision;

[0063] Step S52: Formulate a corresponding adjustment processing strategy according to the carbon neutrality adjustment decision, and use the adjustment processing strategy to dynamically adjust and optimize the carbon neutrality processing optimization result accordingly;

[0064] Step S53: Improve and optimize the carbon neutrality processing process by introducing a feedback mechanism into the carbon neutrality adjustment decision and utilizing the adjustment processing strategy.

[0065] The present invention uses adaptive decision-making technology to intelligently adjust and evaluate the optimization results of carbon neutralization treatment to obtain a more optimized carbon neutralization effect. The adaptive decision-making technology introduces the technology of reinforcement learning to adaptively evaluate and adjust the carbon neutralization treatment process. By drawing experience from historical carbon neutralization treatment data, different decisions are constantly tried, and the decision-making strategy is self-adjusted and improved through continuous experiments and feedback to achieve the optimal carbon neutralization adjustment decision. Then, after the carbon neutralization adjustment decision is obtained, a corresponding adjustment processing strategy is formulated according to the carbon neutralization adjustment decision and applied to the carbon neutralization treatment process. Through the adjustment processing strategy, some key parameters, equipment, processes and other steps in the carbon neutralization treatment process are dynamically adjusted and optimized. These dynamic adjustments and optimization measures are based on real-time data analysis and decision feedback, and can adaptively change the carbon neutralization treatment process to further improve carbon neutralization efficiency and reduce costs. Finally, a feedback mechanism is established to improve the carbon neutrality adjustment decision and optimize the carbon neutrality treatment process. After applying the carbon neutrality adjustment decision to the carbon neutrality treatment optimization results, the carbon neutrality treatment optimization results will be continuously monitored, analyzed and feedback will be utilized to evaluate the effectiveness of the determined carbon neutrality adjustment decision. The carbon neutrality treatment process can be adjusted in real time to achieve the optimal carbon neutrality treatment effect. At the same time, improving and optimizing the carbon neutrality treatment process based on the feedback mechanism can further improve the carbon neutrality treatment efficiency and ultimately achieve the optimization of the carbon neutrality treatment process. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0067] Figure 1 This is a schematic diagram of the steps of the carbon neutrality data processing method based on artificial intelligence of the present invention;

[0068] Figure 2 for Figure 1 Detailed step flow diagram of step S1;

[0069] Figure 3 for Figure 2 Detailed step flow diagram of step S13;

[0070] Figure 4 for Figure 1 Detailed step flow diagram of step S2;

[0071] Figure 5 for Figure 4 Detailed step flow chart of step S23 in FIG. DETAILED DESCRIPTION

[0072] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0073] To achieve this, please refer to Figures 1 to 5 The present invention provides a carbon neutrality data processing method based on artificial intelligence, the method comprising the following steps:

[0074] Step S1: obtaining carbon emission related data, and performing data preprocessing on the carbon emission related data to obtain carbon emission noise reduction data; performing noise reduction processing on the carbon emission noise reduction data using an intelligent noise reduction algorithm to obtain carbon emission noise reduction data;

[0075] Step S2: Use the historical data tracing algorithm to perform historical tracing processing on the carbon emission noise reduction data to obtain the carbon emission historical data; use the preset artificial intelligence-based multi-algorithm prediction model to perform prediction and analysis on the carbon emission historical data to obtain the initial results of carbon emission factor prediction;

[0076] Step S3: Based on the initial prediction results of carbon emission factors, the multi-algorithm prediction models are fused using a weight balance algorithm to generate a fusion prediction model; and the carbon emission historical data are re-predicted and analyzed using the fusion prediction model to obtain the carbon emission factor prediction results;

[0077] Step S4: Formulate a corresponding carbon neutrality treatment plan based on the prediction results of carbon emission factors, and use the carbon neutrality treatment plan to execute the corresponding carbon neutrality treatment technology to obtain carbon neutrality treatment data; use the carbon neutrality monitoring model based on reinforcement learning to perform real-time monitoring and optimization processing on the carbon neutrality treatment data to obtain carbon neutrality treatment optimization results;

[0078] Step S5: Use adaptive decision-making technology to intelligently adjust the carbon neutrality processing optimization results to obtain a carbon neutrality adjustment decision; formulate an adjustment processing strategy based on the carbon neutrality adjustment decision, introduce a feedback mechanism, and use the adjustment processing strategy to improve and optimize the carbon neutrality processing process.

[0079] In the embodiment of the present invention, please refer to Figure 1 As shown, it is a schematic diagram of the steps of the carbon neutralization treatment method based on artificial intelligence of the present invention. In this example, the steps of the carbon neutralization treatment method based on artificial intelligence include:

[0080] Step S1: obtaining carbon emission related data, and performing data preprocessing on the carbon emission related data to obtain carbon emission noise reduction data; performing noise reduction processing on the carbon emission noise reduction data using an intelligent noise reduction algorithm to obtain carbon emission noise reduction data;

[0081] The embodiment of the present invention obtains measured data related to carbon emissions by selecting appropriate sensors, and obtains carbon emission data to be de-noised by performing data cleaning, removing duplicate data, abnormal data, invalid data, and filling missing values ​​on the measured carbon emission data. Then, an appropriate intelligent denoising algorithm is selected to eliminate the influence of noise sources in the carbon emission data to be de-noised, so as to improve the reliability and stability of the carbon emission data to be de-noised, and finally obtain carbon emission de-noised data.

[0082] Step S2: Use the historical data tracing algorithm to perform historical tracing processing on the carbon emission noise reduction data to obtain the carbon emission historical data; use the preset artificial intelligence-based multi-algorithm prediction model to perform prediction and analysis on the carbon emission historical data to obtain the initial results of carbon emission factor prediction;

[0083] The embodiment of the present invention performs historical tracing processing on the denoised carbon emission denoising data by setting an appropriate historical data tracing algorithm to obtain the carbon emission historical data. Then, according to actual needs, a multi-algorithm prediction model based on artificial intelligence is constructed to predict and analyze the carbon emission historical characteristic data, and combined with the carbon emission historical data and real-time data for training and tuning to generate high-quality prediction results, and finally obtain the initial results of carbon emission factor prediction.

[0084] Step S3: Based on the initial prediction results of carbon emission factors, the multi-algorithm prediction models are fused using a weight balance algorithm to generate a fusion prediction model; and the carbon emission historical data are re-predicted and analyzed using the fusion prediction model to obtain the carbon emission factor prediction results;

[0085] The embodiment of the present invention constructs a suitable weight balancing algorithm based on the initial results of carbon emission factor prediction through weight coefficients, prediction functions, harmonic smoothing parameters and other parameters, and fuses each algorithm model in the multi-algorithm prediction model into a more accurate and robust fusion prediction model by assigning different weight coefficients to the models to control the process and results of model fusion. Then, the historical carbon emission data is re-predicted and analyzed through the fusion prediction model to finally obtain the prediction results of carbon emission factors.

[0086] Step S4: Formulate a corresponding carbon neutrality treatment plan based on the prediction results of carbon emission factors, and use the carbon neutrality treatment plan to execute the corresponding carbon neutrality treatment technology to obtain carbon neutrality treatment data; use the carbon neutrality monitoring model based on reinforcement learning to perform real-time monitoring and optimization processing on the carbon neutrality treatment data to obtain carbon neutrality treatment optimization results;

[0087] The embodiment of the present invention analyzes and classifies carbon emission factors through the prediction results of carbon emission factors, selects corresponding carbon neutralization treatment schemes for different carbon emission sources, and performs corresponding carbon neutralization treatment technologies on carbon emission noise reduction data to obtain carbon neutralization treatment data. Then, the carbon neutralization treatment data is monitored and optimized in real time by using a carbon neutrality monitoring model based on reinforcement learning, and an incentive reward mechanism is set in combination with reinforcement learning to reward the carbon emission reduction contribution of the carbon emission sources in the carbon neutralization treatment data, so as to promote the optimization of the carbon neutralization treatment process and finally obtain the carbon neutralization treatment optimization result.

[0088] Step S5: Use adaptive decision-making technology to intelligently adjust the carbon neutrality processing optimization results to obtain a carbon neutrality adjustment decision; formulate an adjustment processing strategy based on the carbon neutrality adjustment decision, introduce a feedback mechanism, and use the adjustment processing strategy to improve and optimize the carbon neutrality processing process.

[0089] The embodiment of the present invention uses the adaptive decision-making technology that introduces reinforcement learning to intelligently adjust the carbon neutrality processing optimization results, intelligently evaluate and predict the carbon emission status, and obtain the carbon neutrality adjustment decision. Then, according to the carbon neutrality adjustment decision, the corresponding adjustment processing strategy is selected, and the carbon neutrality processing optimization results are adjusted according to different adjustment processing strategies. By introducing a feedback mechanism to monitor various indicators in the carbon neutrality processing process, the real-time monitored data is transmitted to the adaptive decision-making technology to re-formulate the optimal carbon neutrality adjustment decision, and the adjustment processing strategy is improved and the carbon neutrality processing process is optimized according to the optimal carbon neutrality adjustment decision.

[0090] The present invention obtains carbon emission related data by using sensors and other data collection means, and performs data preprocessing on the obtained carbon emission related data, such as data cleaning, deduplication, normalization and other operations to eliminate unnecessary noise and interference. The preprocessed carbon emission noise reduction data is subjected to noise reduction processing using a suitable intelligent noise reduction algorithm to improve the accuracy and reliability of the carbon emission noise reduction data. The noise reduction carbon emission noise reduction data is historically restored using a historical data tracing algorithm to obtain more real and accurate carbon emission historical data. The carbon emission historical data is analyzed and processed by using a multi-algorithm prediction model based on artificial intelligence to obtain the initial results of carbon emission factor prediction. Then, according to the initial results of factor prediction, a weight balance algorithm is used to perform model fusion processing on multiple prediction models to generate a more accurate and stable fusion model. The carbon emission historical data is re-predicted and analyzed by using a fusion model to obtain a more accurate carbon emission factor prediction result. A carbon neutralization treatment plan is formulated according to the carbon emission factor prediction results, and different carbon neutralization treatment technologies are used to perform corresponding processing to achieve the effect of optimizing the carbon neutralization treatment process. By using the carbon neutrality monitoring model based on reinforcement learning to monitor and optimize the carbon neutrality processing data in real time, a more optimized and effective carbon neutrality processing optimization result can be obtained. Finally, by using adaptive decision-making technology to intelligently adjust the carbon neutrality processing optimization results, a more accurate and optimized carbon neutrality adjustment decision can be obtained, and corresponding adjustment processing strategies can be formulated according to the carbon neutrality adjustment decision. The carbon neutrality processing process is improved and optimized by introducing a feedback mechanism to continuously improve the effect and efficiency of carbon neutrality processing.

[0091] Preferably, step S1 comprises the following steps:

[0092] Step S11: Real-time monitoring, collection and processing of carbon emission sources are performed through sensors to obtain measured carbon emission data;

[0093] Step S12: performing data cleaning processing on the carbon emission measured data to obtain carbon emission data to be denoised;

[0094] Step S13: using an intelligent noise reduction algorithm to perform noise reduction processing on the carbon emission noise reduction data to obtain carbon emission noise reduction data.

[0095] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in FIG. 1 , in this embodiment, step S1 includes the following steps:

[0096] Step S11: Real-time monitoring and collection of carbon emission sources are performed through sensors to obtain measured carbon emission data;

[0097] The embodiment of the present invention selects a suitable sensor type and calibrates and debugs the selected sensor, performs real-time monitoring and data collection processing on the carbon emission source through the sensor, and finally obtains the actual carbon emission data.

[0098] Step S12: performing data cleaning processing on the carbon emission measured data to obtain carbon emission data to be denoised;

[0099] The embodiment of the present invention improves the accuracy and completeness of the carbon emission measured data by performing data cleaning, removing duplicate data, abnormal data, invalid data, and filling missing values, and finally obtains the carbon emission data to be denoised.

[0100] Step S13: using an intelligent noise reduction algorithm to perform noise reduction processing on the carbon emission noise reduction data to obtain carbon emission noise reduction data.

[0101] The embodiment of the present invention constructs an appropriate intelligent denoising algorithm by setting appropriate parameters such as the weight of the Gaussian distribution, the standard deviation of the Gaussian distribution, the mean of the Gaussian distribution, the l2 difference norm, the Dirac pulse function, the noise weight function and the correction value. The constructed intelligent denoising algorithm is used to eliminate the influence of noise sources in the carbon emission data to be denoised, so as to improve the reliability and stability of the carbon emission data to be denoised, and finally obtain the carbon emission denoised data.

[0102] The present invention uses sensors to monitor and collect carbon emission sources in real time to obtain carbon emission measured data. The sensor can collect real-time monitored carbon emission measured data according to different measurement methods and detection principles. Then, the carbon emission measured data collected by the sensor is cleaned. On the one hand, the carbon emission measured data collected by the sensor may be subject to various interferences, such as sensor failure, signal noise, etc., so it is necessary to remove and filter the carbon emission measured data; on the other hand, the carbon emission measured data can also be detected and repaired for outliers to ensure the quality and accuracy of the carbon emission measured data. After data cleaning, cleaner and more accurate carbon emission noise reduction data can be obtained. Finally, the obtained carbon emission noise reduction data is subjected to noise reduction by setting a suitable intelligent noise reduction algorithm to accurately calculate the amplitude and spatial distribution characteristics of the noise. On this basis, the carbon emission noise reduction data is subjected to noise reduction by filtering and other methods to eliminate the influence of noise on the carbon emission noise reduction data, thereby improving the reliability and accuracy of the carbon emission noise reduction data. After intelligent noise reduction, high-quality carbon emission noise reduction data can be obtained.

[0103] Preferably, step S13 comprises the following steps:

[0104] Step S131: using an intelligent noise reduction algorithm to calculate the noise value of the carbon emission data to be denoised, to obtain the carbon emission noise value;

[0105]

[0106] In the formula, e(x) is the carbon emission noise value, x is the carbon emission data set to be denoised, n is the number of Gaussian distributions, and a is the i is the weight of the i-th Gaussian distribution, σ i is the standard deviation of the ith Gaussian distribution, μ i is the mean of the i-th Gaussian distribution, y is the noisy carbon emission data to be denoised, is the l2 difference norm between the noisy carbon emission denoising data y and the carbon emission denoising data set x, m is the number of noise pulse signals in the carbon emission denoising data, c j is the amplitude value of the jth noise pulse signal in the carbon emission data to be denoised, x j is the position of the jth noise pulse signal in the carbon emission data to be de-noised, δ(x―x j ) is the Dirac pulse function, w(t-x) is the noise weight function of the adjacent carbon emission data to be de-noised in the time domain, t is the time variable in the time domain, and ε is the correction value of the carbon emission noise value;

[0107] Step S132: judging the carbon emission noise value according to a preset carbon emission noise threshold, and when the carbon emission noise value is greater than or equal to the preset carbon emission noise threshold, removing the carbon emission noise reduction data corresponding to the carbon emission noise value to obtain the carbon emission noise reduction data;

[0108] Step S133: the carbon emission noise value is judged according to the preset carbon emission noise threshold. When the carbon emission noise value is less than the preset carbon emission noise threshold, the carbon emission noise reduction data corresponding to the carbon emission noise value is directly defined as the carbon emission noise reduction data.

[0109] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 2 FIG. 1 is a schematic diagram of a detailed step flow chart of step S13 in FIG. 1 . In this embodiment, step S13 includes the following steps:

[0110] Step S131: using an intelligent noise reduction algorithm to calculate the noise value of the carbon emission data to be denoised, to obtain the carbon emission noise value;

[0111] The embodiment of the present invention constructs an appropriate intelligent noise reduction algorithm by setting appropriate parameters such as the weight of the Gaussian distribution, the standard deviation of the Gaussian distribution, the mean of the Gaussian distribution, the l2 difference norm, the Dirac pulse function, the noise weight function and the correction value, and calculates the noise value of each carbon emission data to be denoised in the carbon emission data to be denoised according to the intelligent noise reduction algorithm that appears in the noise reduction process, and finally obtains the carbon emission noise value.

[0112]

[0113] In the formula, e(x) is the carbon emission noise value, x is the carbon emission data set to be denoised, n is the number of Gaussian distributions, and a is the i is the weight of the i-th Gaussian distribution, σ i is the standard deviation of the ith Gaussian distribution, μ i is the mean of the i-th Gaussian distribution, y is the noisy carbon emission data to be denoised, is the l2 difference norm between the noisy carbon emission denoising data y and the carbon emission denoising data set x, m is the number of noise pulse signals in the carbon emission denoising data, x j is the amplitude value of the jth noise pulse signal in the carbon emission data to be denoised, x j is the position of the jth noise pulse signal in the carbon emission data to be de-noised, δ(x―x j ) is the Dirac pulse function, w(t-x) is the noise weight function of the adjacent carbon emission data to be de-noised in the time domain, t is the time variable in the time domain, and ε is the correction value of the carbon emission noise value;

[0114] The present invention constructs a formula of an intelligent denoising algorithm function, which is used to calculate the noise value of the carbon emission data to be denoised. In order to eliminate the influence of the noise source in the carbon emission data to be denoised on the subsequent construction process of the multi-algorithm prediction model, it is necessary to perform denoising on the carbon emission data to be denoised to obtain cleaner and more accurate carbon emission data to be denoised. The intelligent denoising algorithm can effectively remove the noise and interference data in the carbon emission data to be denoised, thereby improving the accuracy and reliability of the carbon emission data to be denoised. The algorithm function formula fully considers the carbon emission data set x to be denoised, the number of Gaussian distributions n, the weight a of the i-th Gaussian distribution i , the standard deviation σ of the i-th Gaussian distribution i , the mean μ of the i-th Gaussian distribution i , the noisy carbon emission denoising data y, the l2 difference norm between the noisy carbon emission denoising data y and the carbon emission denoising data set x The number m of noise pulse signals in the carbon emission data to be de-noised, the amplitude value c of the jth noise pulse signal in the carbon emission data to be de-noised j , the position x of the jth noise pulse signal in the carbon emission data to be denoised j , Dirac Delta function δ(x―x j ), the noise weight function w(t-x) of the adjacent carbon emission data to be de-noised in the time domain, the time variable t in the time domain, and the correlation between the carbon emission noise value e(x) and the above parameters form a functional relationship The algorithm function formula realizes the calculation of the noise value of the carbon emission data to be denoised. At the same time, the correction value ε of the carbon emission noise value in the algorithm function formula can be adjusted according to actual conditions, thereby improving the accuracy and applicability of the intelligent denoising algorithm.

[0115] Step S132: judging the carbon emission noise value according to a preset carbon emission noise threshold, and when the carbon emission noise value is greater than or equal to the preset carbon emission noise threshold, removing the carbon emission noise reduction data corresponding to the carbon emission noise value to obtain the carbon emission noise reduction data;

[0116] The embodiment of the present invention determines whether the calculated carbon emission noise value exceeds the preset carbon emission noise threshold according to the preset carbon emission noise threshold. When the carbon emission noise value is greater than or equal to the preset carbon emission noise threshold, it means that the interference effect of the noise source in the carbon emission noise reduction data corresponding to the carbon emission noise value is relatively large. In this case, the carbon emission noise reduction data corresponding to the carbon emission noise value is eliminated to finally obtain the carbon emission noise reduction data.

[0117] Step S133: the carbon emission noise value is judged according to the preset carbon emission noise threshold. When the carbon emission noise value is less than the preset carbon emission noise threshold, the carbon emission noise reduction data corresponding to the carbon emission noise value is directly defined as the carbon emission noise reduction data.

[0118] The embodiment of the present invention determines whether the calculated carbon emission noise value exceeds the preset carbon emission noise threshold according to the preset carbon emission noise threshold. When the carbon emission noise value is less than the preset carbon emission noise threshold, it means that the interference effect of the noise source in the carbon emission noise reduction data corresponding to the carbon emission noise value is small, and the carbon emission noise reduction data corresponding to the carbon emission noise value is directly defined as the carbon emission noise reduction data.

[0119] Since the present invention may have noise interference and outliers in the acquired carbon emission data to be de-noised, which may have an adverse effect on the accuracy and reliability of the subsequent model building work, a suitable intelligent noise reduction algorithm is set to calculate the noise value of the carbon emission data to be de-noised, so that the noise and interference signals existing in the carbon emission data to be de-noised can be identified and measured, and the noise signal can be removed from the source, and the signal-to-noise ratio of the carbon emission data to be de-noised can be enhanced, thereby improving the accuracy and reliability of the carbon emission data to be de-noised. The intelligent noise reduction algorithm optimizes the noise reduction process by combining the weight of the Gaussian distribution, the standard deviation of the Gaussian distribution, the mean of the Gaussian distribution, the l2 difference norm, the Dirac pulse function, and the noise weight function, and adjusts and optimizes the intelligent noise reduction algorithm by setting appropriate noise pulse signal parameters and related noise reduction function parameters to obtain the best noise reduction effect and calculation results, thereby more accurately calculating the carbon emission noise value. Then, according to the specific data processing requirements and quality standards, an appropriate carbon emission noise threshold is set to judge the calculated carbon emission noise value, which can effectively eliminate the carbon emission noise reduction data with large carbon emission noise values, avoid the impact of these carbon emission noise reduction data with large noise values ​​on the overall data, and help to further improve the quality of the data, reduce unnecessary interference and errors, thereby ensuring the accuracy and reliability of the carbon emission noise reduction data. Finally, the carbon emission noise value is judged using the preset carbon emission noise threshold, and the carbon emission noise reduction data with small carbon emission noise values ​​are defined as carbon emission noise reduction data, which can obtain more accurate and reliable carbon emission noise reduction data, which are less affected by noise and can provide a more stable data basis for subsequent multi-algorithm prediction models, thereby improving the availability and effectiveness of carbon emission noise reduction data.

[0120] Preferably, step S2 comprises the following steps:

[0121] Step S21: tracing the carbon emission noise reduction data historically using a historical data tracing algorithm to obtain carbon emission historical data;

[0122] Step S22: preprocessing the carbon emission historical data using feature extraction and conversion technology to obtain a carbon emission historical feature data set;

[0123] Step S23: Predictive analysis is performed on the historical characteristic data set of carbon emissions using a preset artificial intelligence-based multi-algorithm prediction model to obtain an initial result of carbon emission factor prediction.

[0124] As an embodiment of the present invention, refer to Figure 4 As shown, Figure 1 Detailed step flow diagram of step S2 in FIG. 1 , in this embodiment, step S2 includes the following steps:

[0125] Step S21: tracing the carbon emission noise reduction data historically using a historical data tracing algorithm to obtain carbon emission historical data;

[0126] The embodiment of the present invention constructs an appropriate historical data tracing algorithm by selecting appropriate tracing exponential decay function, attenuation oscillation amplitude value, attenuation rate and tracing time control parameters, tracing discontinuity function, time step decay function and related parameters to perform historical tracing processing on the denoised carbon emission denoising data, and finally obtains carbon emission historical data.

[0127] Step S22: preprocessing the carbon emission historical data using feature extraction and conversion technology to obtain a carbon emission historical feature data set;

[0128] The embodiment of the present invention uses feature extraction and conversion technology to perform feature collection, feature extraction and feature conversion on the carbon emission noise reduction data according to the characteristics and distribution of the carbon emission noise reduction data, so as to generate a set of feature data sets with special significance, and finally obtain a carbon emission historical feature data set.

[0129] Step S23: Predictive analysis is performed on the historical characteristic data set of carbon emissions using a preset artificial intelligence-based multi-algorithm prediction model to obtain an initial result of carbon emission factor prediction.

[0130] The embodiment of the present invention selects a suitable artificial intelligence-based multi-algorithm prediction model to perform predictive analysis on the carbon emission historical characteristic data set according to actual needs, and combines the carbon emission historical data and real-time data for training and tuning to generate high-quality prediction results, and finally obtain the initial results of carbon emission factor prediction.

[0131] The present invention uses an appropriate historical data tracing algorithm to trace the carbon emission noise reduction data to obtain accurate carbon emission historical data, which can ensure that subsequent predictions and analyses are based on real historical data. Then, the carbon emission historical data is processed by using feature extraction and conversion technology to obtain a carbon emission historical feature data set. Feature extraction can convert carbon emission historical data into feature information data that can be used for modeling and prediction, so as to facilitate subsequent predictions and analysis. Finally, by using a preset multi-algorithm prediction model based on artificial intelligence, the carbon emission historical feature data set is predictively analyzed and processed to obtain the initial results of carbon emission factor prediction. Through the combined use of multi-algorithm prediction models, the influence of multiple factors can be comprehensively considered to obtain more accurate initial results of carbon emission factor prediction, which can provide important reference and data support for the implementation of subsequent carbon neutralization processing optimization processes.

[0132] Preferably, the historical data tracing algorithm function formula in step S21 is specifically:

[0133]

[0134]

[0135] Where D(T) is the historical data of carbon emissions at the historical tracing time T, N is the number of tracing exponential decay functions, τ is the integral time component, and α k is the attenuation oscillation amplitude of the kth traceable exponential decay function, β k is the decay rate of the kth traceable exponential decay function, γ k is the tracing time control parameter of the kth tracing exponential decay function, exp is the exponential function, M is the number of tracing discontinuous functions, and f r (u r (T―τ)) is the rth traceable discontinuous function, u r (T―τ) is the time step decay function corresponding to the rth traceable discontinuous function, is the tracing time control function of the historical tracing time T, R is the number of historical tracing in the carbon emission noise reduction data, p l is the lth carbon emission noise reduction data, ξ is the traceability ratio of carbon emission noise reduction data per unit time, T0 is the initial time of historical traceability, T R is the end time of historical tracing, and η is the correction value of historical carbon emission data.

[0136] The present invention constructs a formula of a historical data tracing algorithm function, which is used for historical tracing of carbon emission noise reduction data. The historical data tracing algorithm analyzes and processes the historical tracing information of carbon emission noise reduction data, extracts the implicit laws and trends behind the carbon emission noise reduction data, and thus provides strong data support for the prediction of future trends. The core of the historical data tracing algorithm function is the tracing exponential decay function, which describes the decay law of historical data. The tracing exponential decay function is composed of the superposition of several exponential functions and the setting of appropriate decay oscillation amplitude value, decay rate and tracing time control parameters. In addition to the tracing exponential decay function, the historical data tracing algorithm function also includes a discontinuous function. Compared with the tracing exponential decay function, the discontinuous function is more in line with the actual situation. In the actual application process, historical data is often not completely smooth, and some discontinuous situations are likely to occur. Through the introduction of the discontinuous function, the marker function can be used to describe the mutation points in the data, so as to more accurately describe the evolution law of historical data. In addition, in the historical data traceability algorithm, the traceability time of historical data can also be modeled through the traceability time control function, which is related to the number of samples of historical data and the frequency of traceability. The historical data traceability algorithm can extract the inherent regularity information of the data through in-depth mining of historical data, providing an important reference for future predictions. The algorithm function formula fully considers the number N of traceability exponential decay functions, the integral time component τ, and the attenuation oscillation amplitude value α of the kth traceability exponential decay function. k , the decay rate β of the kth traceable exponential decay function k , the tracing time control parameter γ of the kth tracing exponential decay function k , the number of traceable discontinuous functions M, the rth traceable discontinuous function f r (u r (T―τ)), the time step decay function u corresponding to the rth traceable discontinuous function r (T―τ), the traceability time control function of the historical traceability time T The number of historical traceability in carbon emission noise reduction data R, the lth carbon emission noise reduction data p l , the carbon emission noise reduction data traceability ratio value ξ per unit time, the initial time of historical traceability T0, the number of historical traceability through carbon emission noise reduction data R, the lth carbon emission noise reduction data p l , the carbon emission noise reduction data traceability ratio value ξ per unit time, the historical traceability initial time T0 and the historical traceability time T constitute the traceability time control function relation According to the relationship between the historical carbon emission data D(T) at the historical traceability time T and the above parameters, a functional relationship is formed: The algorithm function formula realizes the historical traceability of carbon emission noise reduction data. At the same time, by introducing the correction value η of carbon emission historical data, it can be adjusted according to actual conditions, thereby improving the accuracy and applicability of the historical data traceability algorithm.

[0137] Preferably, step S23 includes the following steps:

[0138] Step S231: construct a multi-algorithm prediction model based on an artificial intelligence algorithm, wherein the multi-algorithm prediction model includes a neural network prediction model, a support vector machine classification model, a logistic regression prediction model, and a decision tree prediction model;

[0139] Step S232: using a neural network prediction model to perform autonomous learning prediction processing on the carbon emission historical characteristic data set to obtain a carbon emission factor prediction data set;

[0140] Step S233: using a support vector machine classification model to perform factor classification processing on the carbon emission factor prediction data set to obtain a carbon emission factor type data set;

[0141] Step S234: using a logistic regression prediction model to perform factor weight prediction on the carbon emission factor type data set to obtain a carbon emission factor weight data set;

[0142] Step S235: Use the decision tree prediction model to perform weighted tree classification processing on the carbon emission factor weight data set to obtain the initial results of carbon emission factor prediction.

[0143] As an embodiment of the present invention, refer to Figure 5 As shown, Figure 4 FIG. 2 is a schematic diagram of a detailed step flow chart of step S23 in FIG. 2 . In this embodiment, step S23 includes the following steps:

[0144] Step S231: construct a multi-algorithm prediction model based on an artificial intelligence algorithm, wherein the multi-algorithm prediction model includes a neural network prediction model, a support vector machine classification model, a logistic regression prediction model, and a decision tree prediction model;

[0145] The embodiment of the present invention constructs a suitable multi-algorithm prediction model by using multiple artificial intelligence algorithms. The multi-algorithm prediction model includes multiple prediction models such as a neural network prediction model, a support vector machine classification model, a logistic regression prediction model, and a decision tree prediction model.

[0146] Step S232: using a neural network prediction model to perform autonomous learning prediction processing on the carbon emission historical characteristic data set to obtain a carbon emission factor prediction data set;

[0147] The embodiment of the present invention utilizes a neural network prediction model to autonomously learn a carbon emission historical characteristic data set, explores potential laws and patterns from the carbon emission historical characteristic data set, processes a multi-dimensional carbon emission historical characteristic data set through a neural network prediction model, and finally obtains a carbon emission factor prediction data set.

[0148] Step S233: using a support vector machine classification model to perform factor classification processing on the carbon emission factor prediction data set to obtain a carbon emission factor type data set;

[0149] The embodiment of the present invention performs model training on a carbon emission factor prediction data set by utilizing a support vector machine classification model, divides the data set into different categories according to the characteristics of each factor in the data set, and performs classification processing on the data set by selecting a support vector machine classification model with appropriate kernel functions and parameters, and finally obtains a carbon emission factor type data set.

[0150] Step S234: using a logistic regression prediction model to perform factor weight prediction on the carbon emission factor type data set to obtain a carbon emission factor weight data set;

[0151] The embodiment of the present invention uses a logistic regression prediction model to predict factor weights for a carbon emission factor type data set. By analyzing the carbon emission factor type data set using the logistic regression prediction model, the influence of different factors on carbon emissions is predicted, thereby determining the weight of each factor and ultimately obtaining a carbon emission factor weight data set.

[0152] Step S235: Use the decision tree prediction model to perform weighted tree classification processing on the carbon emission factor weight data set to obtain the initial results of carbon emission factor prediction.

[0153] The embodiment of the present invention establishes a decision tree prediction model through a carbon emission factor weight data set, classifies and processes the carbon emission factor weight data set in more detail according to the weights, and finally obtains an initial result of carbon emission factor prediction.

[0154] The present invention constructs a multi-algorithm prediction model by utilizing a variety of artificial intelligence algorithms, and the multi-algorithm prediction model includes a neural network prediction model, a support vector machine classification model, a logistic regression prediction model, and a decision tree prediction model. The multi-algorithm prediction model comprehensively considers the impact of multiple factors on carbon emissions, and can predict and optimize according to the historical characteristic data of carbon emissions. The combined use of these artificial intelligence algorithm models can improve the accuracy and reliability of the prediction. The carbon emission historical characteristic data set is processed by using a neural network prediction model to obtain a carbon emission factor prediction data set. The neural network can explore potential laws and patterns from historical data through autonomous learning, and can process multi-dimensional characteristic data to obtain more accurate prediction results. Then, the carbon emission factor prediction data set is subjected to factor classification processing using a support vector machine classification model to obtain a carbon emission factor type data set. The support vector machine can classify the data into different categories according to the characteristics of the data attributes, so that the factors can be further classified and processed to provide more information for subsequent prediction and optimization. The carbon emission factor type data set is subjected to factor weight prediction using a logistic regression prediction model to obtain a carbon emission factor weight data set. The logistic regression model can predict the impact of different factors on carbon emissions by analyzing the factor type data set, so as to determine the weight of each factor and provide more accurate information for subsequent optimization. Finally, the decision tree prediction model is used to perform weight tree classification on the carbon emission factor weight data set to obtain the initial results of carbon emission factor prediction. The decision tree model can classify and process the weight data set in more detail to obtain more accurate prediction results. Accurate carbon emission factor prediction results are obtained through the multi-algorithm prediction model, providing accurate data support for subsequent carbon neutralization processing.

[0155] Preferably, the weight balancing algorithm function formula in step S3 is specifically:

[0156]

[0157]

[0158] Where V(θ) is the weight balancing algorithm function, θ is the time variable of the model fusion process, ω q is the weight coefficient of the qth algorithm model in the multi-algorithm prediction model, is the control granularity parameter of the time exponential function, b is the time variable harmonic smoothing parameter, F q is the prediction function of the qth algorithm model in the multi-algorithm prediction model, λ q is the harmonic smoothing parameter of the prediction function of the qth algorithm model in the multi-algorithm prediction model, J is the number of parameters of the prediction function, φ q,g is the gth parameter value of the prediction function of the qth algorithm model in the multi-algorithm prediction model, Xθ―g is the model input data of the multi-algorithm prediction model at time θ―g, exp is the exponential function, X q is the model input data of the qth algorithm model in the multi-algorithm prediction model, X′ q is the model output data of the qth algorithm model in the multi-algorithm prediction model, d q (X q ,X′ q ) is the model input data X of the qth algorithm model in the multi-algorithm prediction model q With the model output data X′ q The distance function between is the weight controlling smoothing parameter, It is the correction value of the weight balancing algorithm function.

[0159] The present invention constructs a formula of a weighted balance algorithm function for model fusion processing of multiple algorithm prediction models. The weighted balance algorithm is a fusion technology for multiple algorithm prediction models. By assigning different weight coefficients to each algorithm model, it is fused into a more accurate and robust fusion prediction model, thereby improving the accuracy and stability of the prediction results. The weighted balance algorithm function contains several important parameters, among which the weight coefficient determines the contribution of each algorithm model to the final result. The weight coefficient of each algorithm model is determined according to its prediction performance and error index. The better the performance of the algorithm model, the larger the corresponding weight coefficient. The weighted balance algorithm function is a function that integrates the prediction results of multiple algorithm models. It can control the process and results of model fusion through multiple factors such as weight coefficients, prediction functions, and harmonic smoothing parameters. In the application of carbon emission factor prediction, the function can be used to fuse the results of multiple prediction models, thereby improving the prediction accuracy and reliability. The algorithm function formula fully considers the model fusion process time variable θ, the weight coefficient ω of the qth algorithm model in the multi-algorithm prediction model q , the time exponential function controls the granularity parameter The time variable harmonic smoothing parameter b, the prediction function F of the qth algorithm model in the multi-algorithm prediction model q , the harmonic smoothing parameter λ of the prediction function of the qth algorithm model in the multi-algorithm prediction model q , the number of parameters J of the prediction function, the gth parameter value φ of the prediction function of the qth algorithm model in the multi-algorithm prediction model q,g , the model input data X of the multi-algorithm prediction model at time θ―g θ―g , the model input data X of the qth algorithm model in the multi-algorithm prediction model q , the model output data X′ of the qth algorithm model in the multi-algorithm prediction model q , the model input data X of the qth algorithm model in the multi-algorithm prediction modelq With the model output data X′ q The distance function d q (X q ,X′ q ), weight controls the smoothing parameter The correction value ζ of the weighted balance algorithm function, where the weight coefficient is the core of the weighted balance algorithm function, is obtained by using the model input data X of the qth algorithm model in the multi-algorithm prediction model. q , the model output data X′ of the qth algorithm model in the multi-algorithm prediction model q , the model input data X of the qth algorithm model in the multi-algorithm prediction model q With the model output data X′ q The distance function d q (X q ,X′ q ) and weights to control the smoothing parameter A functional relationship According to the relationship between the weight balance algorithm function V(θ) and the above parameters, a functional relationship is formed: The algorithm function formula realizes the model fusion of multiple algorithm prediction models. At the same time, the correction value of the algorithm function is balanced by weight. The introduction of can be used to adjust special situations that occur during the model fusion process, further improving the applicability and stability of the weight balancing algorithm function, thereby improving the generalization ability and robustness of the fusion prediction model.

[0160] Preferably, step S4 comprises the following steps:

[0161] Step S41: formulating a corresponding carbon neutralization processing scheme according to the carbon emission factor prediction result, and using the carbon neutralization processing scheme to perform a corresponding carbon neutralization processing technology on the carbon emission noise reduction data to obtain carbon neutralization processing data;

[0162] The embodiment of the present invention analyzes and classifies carbon emission factors through the carbon emission factor prediction results, selects corresponding carbon neutralization treatment schemes for different carbon emission sources, performs corresponding carbon neutralization treatment technologies on the carbon emission noise reduction data, and adopts carbon capture, carbon storage, carbon utilization and other technologies and measures, such as adding pollutant treatment equipment, optimizing production processes and raw material usage, etc., to perform corresponding carbon neutralization treatment measures on the carbon emission noise reduction data, and finally obtains carbon neutralization treatment data.

[0163] Step S42: performing data preprocessing on the carbon neutrality processing data to obtain a carbon neutrality processing data set;

[0164] The embodiment of the present invention performs pre-processing such as data cleaning, data integration, feature selection and feature conversion on the carbon neutrality processing data, and finally obtains a carbon neutrality processing data set.

[0165] Step S43: Use the carbon neutrality monitoring model based on reinforcement learning to perform real-time monitoring and optimization processing on the carbon neutrality processing data set to obtain the carbon neutrality processing optimization results.

[0166] The embodiment of the present invention utilizes a carbon neutrality monitoring model based on reinforcement learning to perform real-time monitoring and optimization processing of the carbon neutrality processing data set, and sets an incentive reward mechanism in combination with reinforcement learning to reward the carbon emission reduction contribution of the carbon emission sources in the carbon neutrality processing data set, so as to promote the optimization of the carbon neutrality processing process and ultimately obtain a carbon neutrality processing optimization result.

[0167] The present invention formulates a corresponding carbon neutralization treatment scheme according to the prediction results of carbon emission factors, and performs corresponding carbon neutralization treatment technology on carbon emission noise reduction data by using the formulated carbon neutralization treatment scheme. Through the application of carbon neutralization treatment technology, carbon emissions can be reduced to a certain extent, thereby obtaining carbon neutralization treatment data. Then, the obtained carbon neutralization treatment data is preprocessed so that it can be subsequently input into the carbon neutralization monitoring model for analysis and optimization. The purpose of data preprocessing is to remove duplicate invalid data, make the carbon neutralization treatment data cleaner and easier to analyze, and at the same time, feature selection and conversion operations can be performed to improve the accuracy and generalization ability of the carbon neutralization monitoring model. After data preprocessing, a carbon neutralization treatment data set suitable for monitoring and analysis can be obtained. Finally, by using a carbon neutralization monitoring model based on reinforcement learning to perform real-time monitoring and optimization of the carbon neutralization treatment data set, the optimization result of the carbon neutralization treatment process can be obtained. By real-time updating and adjusting the carbon neutralization treatment process of the carbon neutralization monitoring model, the carbon neutralization treatment process can be optimized and controlled in real time, thereby obtaining a better carbon neutralization treatment effect, and having higher real-time performance and carbon neutralization treatment efficiency.

[0168] Preferably, step S43 includes the following steps:

[0169] Step S431: dividing the carbon neutrality processing data set into a carbon neutrality training data set, a carbon neutrality verification data set, and a carbon neutrality test data set according to a preset division rule;

[0170] The embodiment of the present invention divides the carbon neutrality processing dataset into a carbon neutrality training dataset, a carbon neutrality verification dataset and a carbon neutrality test dataset according to a certain division ratio, and divides the carbon neutrality processing dataset into 70% of the carbon neutrality training dataset, 20% of the carbon neutrality verification dataset and 10% of the carbon neutrality test dataset according to a preset division ratio of 7:2:1.

[0171] Step S432: constructing a carbon neutrality monitoring model based on reinforcement learning, wherein the carbon neutrality monitoring model includes model training, model verification and model evaluation;

[0172] An embodiment of the present invention constructs an appropriate carbon neutrality monitoring model by utilizing reinforcement learning. The carbon neutrality monitoring model includes model training, model verification and model evaluation, wherein the carbon neutrality monitoring model is trained by a carbon neutrality training data set, and the carbon neutrality monitoring model is verified by a carbon neutrality verification data set. At the same time, the carbon neutrality monitoring model is evaluated by a carbon neutrality test data set, so as to improve the generalization performance and robustness of the carbon neutrality monitoring model.

[0173] Step S433: inputting the carbon neutrality training data set into the carbon neutrality monitoring model based on reinforcement learning for model training, and tuning the model parameters by monitoring the loss function to generate a verification model; inputting the carbon neutrality verification data set into the verification model for model verification to generate a test model;

[0174] The embodiment of the present invention performs model training by inputting the divided carbon neutrality training data set into the constructed carbon neutrality monitoring model based on reinforcement learning, and during the training process, an appropriate monitoring loss function is set through the time weight coefficient, CO2 concentration, CO2 concentration disturbance control space, disturbance space variable and probability density function to tune the carbon neutrality monitoring model parameters to generate a verification model. Then, the divided carbon neutrality verification data set is input into the verification model after parameter optimization for model verification, the optimal carbon neutrality monitoring model parameter combination is determined, and finally a test model is generated.

[0175] Among them, the monitoring loss function is as follows:

[0176]

[0177] Where L(θ) is the monitoring loss function, θ is the carbon neutrality monitoring model parameter, H is the number of training rounds for model training, and s h is the hth model training round, ρ(s h ) is the time weight coefficient of the hth model training round, z is the CO2 concentration, z max is the maximum CO2 concentration in the carbon neutrality training dataset, is the CO2 concentration disturbance control space, is the disturbance space variable, is the h-th model training round s h , CO2 concentration z and disturbance space The predicted output result of the model parameter θ under the condition of is the actual output of the model, is the number of rounds s of model training at a given hth time h , CO2 concentration z and disturbance space The probability density function of the actual output result of the model under the condition of , ∈ is the correction value of the monitoring loss function;

[0178] The present invention constructs a formula of a monitoring loss function for tuning the parameters of a carbon neutrality monitoring model. When training a carbon neutrality training data set by using a carbon neutrality monitoring model, in order to help the carbon neutrality monitoring model fit the data as much as possible, it is necessary to use a suitable monitoring loss function as an indicator for model parameter optimization. The main function of the monitoring loss function is to monitor the output results of the carbon neutrality monitoring model, and then adjust the model parameters according to the monitoring results to improve the prediction accuracy and stability of the carbon neutrality monitoring model. The monitoring loss function is an integral form of formula, which requires comparing the prediction results and actual output results of the carbon neutrality monitoring model, and taking into account the disturbance factor of CO2 concentration. In the monitoring loss function, the time weight coefficient is used to weight the model training effect of the historical training rounds, so that the more recent training results have a greater impact on the carbon neutrality monitoring model. At the same time, the introduction of disturbance space variables can increase the adaptability of the carbon neutrality monitoring model to abnormal situations, thereby improving the robustness of the carbon neutrality monitoring model. When the predicted results of the carbon neutrality monitoring model are inconsistent with the actual output results, the value of the monitoring loss function will increase, thereby guiding the carbon neutrality monitoring model parameters to be adjusted in a more optimized direction. Therefore, the monitoring loss function improves the training effect and stability of the carbon neutrality monitoring model from multiple aspects by introducing the CO2 concentration disturbance factor, weighting the historical training effect, and taking into account the actual output results, so that it can adapt to the input data in different scenarios and accurately monitor and effectively control the carbon neutrality processing data set to be monitored. The algorithm function formula fully considers the carbon neutrality monitoring model parameters θ, the number of training rounds H for model training, and the hth model training round s h , the time weight coefficient ρ(s h ), CO2 concentration z, the maximum CO2 concentration z in the carbon neutral training dataset max , CO2 concentration disturbance control space Perturbation space variables In the hth model training round s h , CO2 concentration z and disturbance space The predicted output of the model parameter θ under the condition The actual output of the model At a given h-th model training round s h , CO2 concentration z and disturbance space The probability density function of the actual output of the model under the condition According to the monitoring loss function L(θ), a functional relationship is formed The algorithm function formula realizes the tuning of the parameters of the carbon neutrality monitoring model. At the same time, by introducing the correction value ∈ of the monitoring loss function, adjustments can be made to special situations that arise during model training, further improving the applicability and stability of the monitoring loss function, thereby improving the generalization ability and robustness of the carbon neutrality monitoring model.

[0179] Step S434: Input the carbon neutrality test data set into the test model for model evaluation to obtain an optimized carbon neutrality monitoring model; re-input the carbon neutrality treatment data set into the optimized carbon neutrality monitoring model for real-time monitoring, and optimize the carbon emission sources in the carbon neutrality treatment data set by setting up an incentive mechanism to obtain carbon neutrality treatment optimization results.

[0180] The embodiment of the present invention inputs the divided carbon neutrality test data set into the test model for model evaluation, and further checks and optimizes the parameters of the carbon neutrality monitoring model by calculating the accuracy, recall rate, F1 value and other indicators of the carbon neutrality monitoring model to obtain a more efficient and accurate optimized carbon neutrality monitoring model. At the same time, the carbon neutrality processing data set is re-input into the optimized carbon neutrality monitoring model for real-time monitoring, and the carbon emission reduction contribution of the carbon emission sources in the carbon neutrality processing data set is evaluated by setting a reward mechanism combined with reinforcement learning, and an incentive reward mechanism is set to reward emission reduction behavior and emission reduction effects, thereby promoting the optimization of carbon neutrality processing and finally obtaining carbon neutrality processing optimization results.

[0181] The present invention divides the carbon neutrality processing data set by a preset division rule, and can better utilize the carbon neutrality processing data set to train, verify and evaluate the carbon neutrality monitoring model. At the same time, by separating the carbon neutrality processing data set, overfitting of the carbon neutrality monitoring model and improving the generalization ability of the carbon neutrality monitoring model can be avoided. By using a method based on reinforcement learning, a carbon neutrality monitoring model suitable for carbon neutrality processing optimization is constructed. By performing model training, model verification and model evaluation on the carbon neutrality monitoring model, all-round monitoring and optimization of the carbon neutrality monitoring model can be achieved. Then, by inputting the carbon neutrality training data set and the carbon neutrality verification data set, and adaptively adjusting the model parameters through the calculation of the appropriate monitoring loss function, the carbon neutrality monitoring model can more accurately predict the carbon neutrality effect. Through cyclic iterative training, verification and tuning, a more optimized test model can be generated for subsequent monitoring and optimization of the carbon neutrality processing process. Finally, by inputting the carbon neutrality test data set, the generated test model is evaluated to obtain an optimized carbon neutrality monitoring model. The carbon neutrality monitoring model can be used to monitor the carbon neutrality treatment process in real time, and optimize the carbon emission source by using the set reward mechanism to obtain a better carbon neutrality treatment effect, which can further improve the real-time and efficiency of the carbon neutrality treatment process. Using the carbon neutrality monitoring model based on reinforcement learning, through the steps of carbon neutrality treatment data set division, model training, model verification and model evaluation, the carbon neutrality treatment process can be more effectively monitored and optimized, thereby achieving a more complete carbon neutrality treatment effect.

[0182] Preferably, step S5 comprises the following steps:

[0183] Step S51: using adaptive decision-making technology to intelligently adjust and evaluate the carbon neutrality processing optimization results to obtain a carbon neutrality adjustment decision;

[0184] The embodiment of the present invention utilizes adaptive decision-making technology that introduces reinforcement learning to intelligently evaluate and predict carbon emission status based on environmental data and combined with carbon neutrality processing optimization results, and quickly make the best carbon neutrality adjustment method to ultimately obtain a carbon neutrality adjustment decision.

[0185] Step S52: Formulate a corresponding adjustment processing strategy according to the carbon neutrality adjustment decision, and use the adjustment processing strategy to dynamically adjust and optimize the carbon neutrality processing optimization result accordingly;

[0186] The embodiment of the present invention analyzes the problems that may be encountered in the carbon neutrality processing process, such as which variables are easily affected, which factors will increase carbon emissions, and which factors can reduce carbon emissions. According to the carbon neutrality adjustment decision, the corresponding adjustment processing strategy is selected, and the carbon neutrality processing optimization results are adjusted according to different adjustment processing strategies to achieve the optimal carbon neutrality processing effect, and finally the corresponding dynamic adjustment and optimization measures can be executed.

[0187] Step S53: Improve and optimize the carbon neutrality processing process by introducing a feedback mechanism into the carbon neutrality adjustment decision and utilizing the adjustment processing strategy.

[0188] The embodiment of the present invention introduces a feedback mechanism to monitor various indicators in the carbon neutrality processing process, transmits the real-time monitored data to the adaptive decision-making technology to re-formulate the optimal carbon neutrality adjustment decision, and optimizes and improves the adjustment processing strategy according to the optimal carbon neutrality adjustment decision, ultimately realizing the optimization of the carbon neutrality processing process.

[0189] The present invention uses adaptive decision-making technology to intelligently adjust and evaluate the optimization results of carbon neutralization treatment to obtain a more optimized carbon neutralization effect. The adaptive decision-making technology introduces the technology of reinforcement learning to adaptively evaluate and adjust the carbon neutralization treatment process. By drawing experience from historical carbon neutralization treatment data, different decisions are constantly tried, and the decision-making strategy is self-adjusted and improved through continuous experiments and feedback to achieve the optimal carbon neutralization adjustment decision. Then, after the carbon neutralization adjustment decision is obtained, a corresponding adjustment processing strategy is formulated according to the carbon neutralization adjustment decision and applied to the carbon neutralization treatment process. Through the adjustment processing strategy, some key parameters, equipment, processes and other steps in the carbon neutralization treatment process are dynamically adjusted and optimized. These dynamic adjustments and optimization measures are based on real-time data analysis and decision feedback, and can adaptively change the carbon neutralization treatment process to further improve carbon neutralization efficiency and reduce costs. Finally, a feedback mechanism is established to improve the carbon neutrality adjustment decision and optimize the carbon neutrality treatment process. After applying the carbon neutrality adjustment decision to the carbon neutrality treatment optimization results, the carbon neutrality treatment optimization results will be continuously monitored, analyzed and feedback will be utilized to evaluate the effectiveness of the determined carbon neutrality adjustment decision. The carbon neutrality treatment process can be adjusted in real time to achieve the optimal carbon neutrality treatment effect. At the same time, improving and optimizing the carbon neutrality treatment process based on the feedback mechanism can further improve the carbon neutrality treatment efficiency and ultimately achieve the optimization of the carbon neutrality treatment process.

[0190] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A carbon neutrality data processing method based on artificial intelligence, characterized in that: The following steps are involved: Step S1: obtaining carbon emission related data, and performing data preprocessing on the carbon emission related data to obtain carbon emission noise reduction data; performing noise reduction processing on the carbon emission noise reduction data using an intelligent noise reduction algorithm to obtain carbon emission noise reduction data; Step S2: Use the historical data tracing algorithm to perform historical tracing processing on the carbon emission noise reduction data to obtain the carbon emission historical data; use the preset artificial intelligence-based multi-algorithm prediction model to perform prediction analysis on the carbon emission historical data to obtain the initial results of carbon emission factor prediction. Step S2 includes the following steps: Step S21: Use the historical data tracing algorithm to trace the carbon emission noise reduction data to obtain the carbon emission historical data. The historical data tracing algorithm function formula in step S21 is specifically: Where D(T) is the historical data of carbon emissions at the historical tracing time T, N is the number of tracing exponential decay functions, τ is the integral time component, and α k is the attenuation oscillation amplitude of the kth traceable exponential decay function, β k is the decay rate of the kth traceable exponential decay function, γ k is the tracing time control parameter of the kth tracing exponential decay function, exp is the exponential function, M is the number of tracing discontinuous functions, and f r (u r (T-τ)) is the rth traceable discontinuous function, u r (T-τ) is the time step decay function corresponding to the rth traceable discontinuous function, is the tracing time control function of the historical tracing time T, R is the number of historical tracing in the carbon emission noise reduction data, p l is the lth carbon emission noise reduction data, ξ is the traceability ratio of carbon emission noise reduction data per unit time, T0 is the initial time of historical traceability, T R is the end time of historical tracing, and η is the correction value of historical carbon emission data; Step S22: preprocessing the carbon emission historical data using feature extraction and conversion technology to obtain a carbon emission historical feature data set; Step S23: Predict and analyze the historical characteristic data set of carbon emissions using a preset artificial intelligence-based multi-algorithm prediction model to obtain an initial result of carbon emission factor prediction. Step S23 includes the following steps: Step S231: construct a multi-algorithm prediction model based on an artificial intelligence algorithm, wherein the multi-algorithm prediction model includes a neural network prediction model, a support vector machine classification model, a logistic regression prediction model, and a decision tree prediction model; Step S232: using a neural network prediction model to perform autonomous learning prediction processing on the carbon emission historical characteristic data set to obtain a carbon emission factor prediction data set; Step S233: using a support vector machine classification model to perform factor classification processing on the carbon emission factor prediction data set to obtain a carbon emission factor type data set; Step S234: using a logistic regression prediction model to perform factor weight prediction on the carbon emission factor type data set to obtain a carbon emission factor weight data set; Step S235: using a decision tree prediction model to perform weight tree classification processing on the carbon emission factor weight data set to obtain an initial result of carbon emission factor prediction; Step S3: Based on the initial prediction results of carbon emission factors, the weight balance algorithm is used to perform model fusion processing on the multi-algorithm prediction model to generate a fusion prediction model; and the carbon emission historical data is re-predicted and analyzed through the fusion prediction model to obtain the carbon emission factor prediction results. The weight balance algorithm function formula in step S3 is specifically as follows: In the formula, is the weight balancing algorithm function, is the model fusion process time variable, ω q is the weight coefficient of the qth algorithm model in the multi-algorithm prediction model, is the control granularity parameter of the time exponential function, b is the time variable harmonic smoothing parameter, F q is the prediction function of the qth algorithm model in the multi-algorithm prediction model, λ q is the harmonic smoothing parameter of the prediction function of the qth algorithm model in the multi-algorithm prediction model, J is the number of parameters of the prediction function, φ q,g is the gth parameter value of the prediction function of the qth algorithm model in the multi-algorithm prediction model, For multi-algorithm prediction models in time The model input data at , exp is the exponential function, X q is the model input data of the qth algorithm model in the multi-algorithm prediction model, X′ q is the model output data of the qth algorithm model in the multi-algorithm prediction model, d q (X q ,X′ q ) is the model input data X of the qth algorithm model in the multi-algorithm prediction model q With the model output data X′ q The distance function between is the weight controlling smoothing parameter, is the correction value of the weight balancing algorithm function; Step S4: Formulate a corresponding carbon neutrality treatment plan based on the prediction results of carbon emission factors, and use the carbon neutrality treatment plan to execute the corresponding carbon neutrality treatment technology to obtain carbon neutrality treatment data; use the carbon neutrality monitoring model based on reinforcement learning to perform real-time monitoring and optimization processing on the carbon neutrality treatment data to obtain carbon neutrality treatment optimization results; Step S5: Use adaptive decision-making technology to intelligently adjust the carbon neutrality processing optimization results to obtain a carbon neutrality adjustment decision; formulate an adjustment processing strategy based on the carbon neutrality adjustment decision, introduce a feedback mechanism, and use the adjustment processing strategy to improve and optimize the carbon neutrality processing process.

2. The carbon neutrality data processing method based on artificial intelligence according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: Real-time monitoring and collection of carbon emission sources are performed through sensors to obtain measured carbon emission data; Step S12: performing data cleaning processing on the carbon emission measured data to obtain carbon emission data to be denoised; Step S13: using an intelligent noise reduction algorithm to perform noise reduction processing on the carbon emission noise reduction data to obtain carbon emission noise reduction data.

3. The carbon neutrality data processing method based on artificial intelligence according to claim 2 is characterized in that: Step S13 includes the following steps: Step S131: using an intelligent noise reduction algorithm to calculate the noise value of the carbon emission data to be denoised, to obtain the carbon emission noise value; In the formula, e(x) is the carbon emission noise value, x is the carbon emission data set to be denoised, n is the number of Gaussian distributions, and a is the i is the weight of the i-th Gaussian distribution, σ i is the standard deviation of the ith Gaussian distribution, μ i is the mean of the i-th Gaussian distribution, y is the noisy carbon emission data to be denoised, is the l2 difference norm between the noisy carbon emission denoising data y and the carbon emission denoising data set x, m is the number of noise pulse signals in the carbon emission denoising data, c j is the amplitude value of the jth noise pulse signal in the carbon emission data to be denoised, x j is the position of the jth noise pulse signal in the carbon emission data to be de-noised, δ(xx j ) is the Dirac pulse function, w(tx) is the noise weight function of the adjacent carbon emission data to be de-noised in the time domain, t is the time variable in the time domain, and ε is the correction value of the carbon emission noise value; Step S132: judging the carbon emission noise value according to a preset carbon emission noise threshold, and when the carbon emission noise value is greater than or equal to the preset carbon emission noise threshold, removing the carbon emission noise reduction data corresponding to the carbon emission noise value to obtain the carbon emission noise reduction data; Step S133: the carbon emission noise value is judged according to the preset carbon emission noise threshold. When the carbon emission noise value is less than the preset carbon emission noise threshold, the carbon emission noise reduction data corresponding to the carbon emission noise value is directly defined as the carbon emission noise reduction data.

4. The carbon neutrality data processing method based on artificial intelligence according to claim 1 is characterized in that: Step S4 includes the following steps: Step S41: formulating a corresponding carbon neutralization processing scheme according to the carbon emission factor prediction result, and using the carbon neutralization processing scheme to perform a corresponding carbon neutralization processing technology on the carbon emission noise reduction data to obtain carbon neutralization processing data; Step S42: performing data preprocessing on the carbon neutrality processing data to obtain a carbon neutrality processing data set; Step S43: Use the carbon neutrality monitoring model based on reinforcement learning to perform real-time monitoring and optimization processing on the carbon neutrality processing data set to obtain the carbon neutrality processing optimization results.

5. The carbon neutrality data processing method based on artificial intelligence according to claim 4 is characterized in that: Step S43 includes the following steps: Step S431: dividing the carbon neutrality processing data set into a carbon neutrality training data set, a carbon neutrality verification data set, and a carbon neutrality test data set according to a preset division rule; Step S432: constructing a carbon neutrality monitoring model based on reinforcement learning, wherein the carbon neutrality monitoring model includes model training, model verification and model evaluation; Step S433: inputting the carbon neutrality training data set into the carbon neutrality monitoring model based on reinforcement learning for model training, and tuning the model parameters by monitoring the loss function to generate a verification model; inputting the carbon neutrality verification data set into the verification model for model verification to generate a test model; Among them, the monitoring loss function is as follows: Where L(θ) is the monitoring loss function, θ is the carbon neutrality monitoring model parameter, H is the number of training rounds for model training, and s h is the hth model training round, ρ(s h ) is the time weight coefficient of the hth model training round, z is the CO2 concentration, z max is the maximum Co2 concentration in the carbon neutrality training dataset, is the Co2 concentration disturbance control space, is the disturbance space variable, is the h-th model training round s h , CO2 concentration z and disturbance space The predicted output result of the model parameter θ under the condition of is the actual output of the model, is the number of rounds s of model training at a given hth time h , CO2 concentration z and disturbance space The probability density function of the actual output result of the model under the condition of , ∈ is the correction value of the monitoring loss function; Step S434: Input the carbon neutrality test data set into the test model for model evaluation to obtain an optimized carbon neutrality monitoring model; re-input the carbon neutrality treatment data set into the optimized carbon neutrality monitoring model for real-time monitoring, and optimize the carbon emission sources in the carbon neutrality treatment data set by setting up an incentive mechanism to obtain carbon neutrality treatment optimization results.

6. The carbon neutrality data processing method based on artificial intelligence according to claim 1 is characterized in that: Step S5 includes the following steps: Step S51: using adaptive decision-making technology to intelligently adjust and evaluate the carbon neutrality processing optimization results to obtain a carbon neutrality adjustment decision; Step S52: Formulate a corresponding adjustment processing strategy according to the carbon neutrality adjustment decision, and use the adjustment processing strategy to dynamically adjust and optimize the carbon neutrality processing optimization result accordingly; Step S53: Improve and optimize the carbon neutrality processing process by introducing a feedback mechanism into the carbon neutrality adjustment decision and utilizing the adjustment processing strategy.

Images