Regional net carbon emission long-term prediction method and system based on multi-factor mixed judgment

Abstract

The invention belongs to the field of information processing, and provides a regional net carbon emission long-term prediction method and system based on multi-factor mixed judgment, and the method comprises the steps: obtaining carbon emission data, and carrying out the preprocessing; based on the preprocessed carbon emission data, analyzing the relationship between each influence factor and the carbon emission through a correlation analysis method, selecting factors with large influence as prediction input characteristics according to an analysis result, and carrying out standardization processing; on the basis of the standardized prediction input features, utilizing a pre-trained convolutional long-short-term memory network model to predict the regional net carbon emission; the improved particle swarm optimization algorithm is adopted to optimize the parameters of the convolutional long-short-term memory network model, the efficiency of the training process is improved, the complex relation of carbon emission can be fitted more accurately, and therefore the problem that the long-term prediction accuracy of the model is low is solved.

Description

Technical Field

[0001] The present invention belongs to the field of information processing technology, and in particular relates to a method and system for long-term prediction of regional net carbon emissions based on multi-factor mixed determination. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The development of human society, economic globalization and industrial civilization are inseparable from energy. In recent years, global energy consumption has continued to grow, and large amounts of energy use have led to rising carbon emissions, bringing severe environmental problems. Against the backdrop of global efforts to address climate warming, achieving a win-win situation between low-carbon emissions and economic development has become a common goal pursued by countries around the world. Against the backdrop of global energy conservation and emission reduction, long-term predictions of regional net carbon emissions are conducive to a correct understanding and grasp of the evolution trend of carbon emissions, and have important practical significance.

[0004] The inventors found in their research that existing carbon emission prediction methods include:

[0005] (a) Forecasting method based on statistical model: Use traditional time series modeling techniques, such as autoregressive moving average and exponential smoothing, to conduct statistical analysis on historical carbon emission data, and then use historical trends to predict future carbon emissions.

[0006] (b) Prediction method based on machine learning: A prediction model is trained using feature variables in historical data through methods such as support vector regression, random forest, and gradient boosting tree, and the model is used to predict future carbon emissions.

[0007] (c) Forecasting methods based on deep learning: Recurrent neural networks (RNN), long short-term memory networks (LSTM), convolutional neural networks (CNN) and other models are usually used to process time series data and make forecasts. These methods have the advantages of strong ability to model complex nonlinear relationships and automatic feature extraction, so they have potential application prospects in the field of carbon emission forecasting.

[0008] In order to formulate targeted emission reduction policies and achieve the expected results, it is necessary to understand the factors that lead to increased carbon emissions so that carbon emissions can be effectively predicted through these factors. Since there are many factors that affect carbon emissions, different angles and classifications may lead to different results. As a result, existing prediction methods often only consider one of many factors such as time, temperature, economic development and energy consumption to model and complete the prediction of carbon emissions. The prediction time span is mostly in months or weeks, and the effect of long-term prediction is not good. Summary of the invention

[0009] In order to solve the above problems, the present invention proposes a long-term prediction method and system for regional net carbon emissions based on multi-factor mixed judgment. The present invention can correctly model the factors with high impact on carbon emissions such as time, temperature, economic development and energy consumption through correlation analysis method, and realize long-term prediction of regional net carbon emissions through deep learning technology, so as to provide more comprehensive and accurate support for the targeted formulation of emission reduction policies.

[0010] According to some embodiments, the first solution of the present invention provides a method for long-term prediction of regional net carbon emissions by mixed determination of multiple factors, which adopts the following technical solution:

[0011] A long-term prediction method for regional net carbon emissions based on multi-factor mixed determination, including:

[0012] Obtain carbon emission data and pre-process it;

[0013] Based on the pre-processed carbon emission data, the relationship between each influencing factor and carbon emission is analyzed through correlation analysis method. According to the analysis results, the factors with the greatest influence are selected as the prediction input features and standardized.

[0014] Based on the standardized prediction input features, the regional net carbon emissions are predicted using a pre-trained convolutional long short-term memory network model.

[0015] Furthermore, the carbon emission data is obtained and preprocessed, specifically:

[0016] Identify abnormal data using vertical processing of data and horizontal processing of data;

[0017] The identified abnormal data is smoothed to complete the outlier correction and obtain the pre-processed carbon emission data.

[0018] Furthermore, the abnormal data identified is smoothed to complete abnormal value correction, and the pre-processed carbon emission data is obtained, specifically:

[0019] If the amount of missing data on a certain day is small, the linear interpolation method is used, and the fill-in value needs to be inserted at time t+j between time t and time t+n on day d;

[0020] When the amount of missing data on a certain day is large, the longitudinal interpolation method based on similar days is used. According to the periodic nature of data changes, the average value of multiple similar emission data is used to replace the missing data.

[0021] Furthermore, based on the pre-processed carbon emission data, the relationship between each influencing factor and carbon emission is analyzed by a correlation analysis method, and the factors with the greatest influence are selected as the prediction input features according to the analysis results, specifically:

[0022] Calculate the corresponding entropy based on the marginal probability density of influencing factors and carbon emissions respectively;

[0023] Calculate the corresponding joint entropy based on the joint probability density of influencing factors and carbon emissions;

[0024] The maximum value of the entropy of the influencing factors and carbon emissions is taken as the minimum value of the joint entropy, and the sum of the entropy of the influencing factors and carbon emissions is taken as the maximum value of the joint entropy to determine the value range of the joint entropy; that is, the maximum value is obtained when the influencing factors and carbon emissions are completely independent of each other, and the minimum value is obtained when they are dependent on each other;

[0025] A variety of influencing factors are screened as characteristic inputs of regional carbon emissions through the threshold of the maximum information coefficient.

[0026] Furthermore, the prediction of regional net carbon emissions based on the standardized prediction input features using a pre-trained convolutional long short-term memory network model includes:

[0027] Based on the prediction input features transformed from the table, the convolution kernel weights are used to perform convolution operations with the local sequence segments of carbon emission information to obtain a preliminary feature matrix;

[0028] The initial feature matrix is ​​weighted by attention as input, and a pooling window is used to slide on the matrix sequence. The maximum value of the window is taken for pooling each time, and the local feature matrix is ​​output.

[0029] Based on the local feature matrix, local features are modeled and integrated in the time series direction to extract time series feature representation;

[0030] These time series feature representations are mapped to a single output, which is the carbon emission prediction result.

[0031] Furthermore, the parameters of the convolutional long short-term memory network model are optimized by using an improved particle swarm optimization algorithm. The specific steps include:

[0032] Assume that in a D-dimensional search space, the number of particles is N, and a particle X in the particle swarm i Represented as X i =[x i1 x i2 …,x iN ], i = 1, 2, ..., N, a particle X in the particle swarm i The speed is represented by V i =[v i1 ,ν i2 ,…,ν iN ], i=1,2,…,N, then the calculation formula for adjusting the particle speed and position is as follows:

[0033]

[0034] Among them, w i is the inertia weight, t is the number of iterations, V id is the velocity of the ith particle in the dth dimension, X id is the position of the ith particle in the dth dimension, c i is the learning factor, the value is a non-negative constant, r i is a random number between [0,1]; the particle fitness function is:

[0035]

[0036] Where n is the number of predicted sample points, y and y′ are the expected output and actual output of the same sample point.

[0037] Furthermore, the prediction input features based on table conversion use convolution kernel weights and local sequence segments of carbon emission information to perform convolution operations to obtain a preliminary feature matrix, specifically:

[0038] Calculate the importance of channels by training the attention weight vector;

[0039] Then use the attention weight to weight the convolutional layer output to get the weighted output of the attention layer.

[0040] According to some embodiments, the second solution of the present invention provides a long-term prediction system for regional net carbon emissions with multi-factor mixed determination, which adopts the following technical solution:

[0041] The long-term prediction system of regional net carbon emissions based on multi-factor mixed determination includes:

[0042] A data processing module, configured to obtain carbon emission data and perform pre-processing;

[0043] The data analysis module is configured to analyze the relationship between various influencing factors and carbon emissions through a correlation analysis method based on the pre-processed carbon emission data, select factors with great influence as prediction input features according to the analysis results, and perform standardization processing;

[0044] The carbon emission prediction module is configured to predict regional net carbon emissions based on standardized prediction input features using a pre-trained convolutional long short-term memory network model.

[0045] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium.

[0046] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the method for long-term prediction of regional net carbon emissions based on multi-factor mixed determination as described in the first aspect above.

[0047] According to some embodiments, a fourth aspect of the present invention provides a computer device.

[0048] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for long-term prediction of regional net carbon emissions based on multi-factor mixed determination as described in the first aspect above are implemented.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention adopts a multi-factor mixed judgment method, which can more comprehensively understand the changing law of regional net carbon emissions by analyzing the relationship between each influencing factor and carbon emissions and then correctly modeling. Compared with the prediction method of a single factor, this comprehensive analysis can more accurately reflect the complex influencing mechanism of carbon emissions.

[0051] By constructing a CNN-LSTM model, the present invention gives full play to the advantages of convolutional neural network (CNN) in filtering noise information in time series data, and the advantages of long short-term memory network (LSTM) in extracting high-level feature representation. This combination helps to better capture patterns and trends in time series data.

[0052] The improved particle swarm optimization algorithm (IPSO) is used to optimize the neural network parameters, which improves the efficiency of the training process and can more accurately fit the complex relationship of carbon emissions, thereby solving the problem of low accuracy of the model for long-term predictions. The optimized algorithm can improve the generalization ability of the model, help the model capture the relationship between carbon emissions and various influencing factors, and improve the accuracy of model predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0054] Figure 1 This is a flow chart of a method for long-term prediction of regional net carbon emissions using multi-factor mixed determination in an embodiment of the present invention;

[0055] Figure 2 Schematic diagram of the CNN-LSTM model constructed in an embodiment of the present invention;

[0056] Figure 3It is a prediction flow chart of the IPSO optimization model in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0058] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0059] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0060] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0061] Embodiment 1

[0062] This embodiment provides a method for long-term prediction of regional net carbon emissions with multi-factor mixed determination. This embodiment uses the method applied to a server as an example for illustration. It is understandable that the method can also be applied to a terminal, and can also be applied to a terminal, a server, and a system, and is implemented through the interaction between the terminal and the server. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communications, middleware services, domain name services, security services CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, which is not limited in this application. In this embodiment, the method includes the following steps:

[0063] Obtain carbon emission data and pre-process it;

[0064] Based on the pre-processed carbon emission data, the relationship between each influencing factor and carbon emission is analyzed through correlation analysis method. According to the analysis results, the factors with the greatest influence are selected as the prediction input features and standardized.

[0065] Based on the standardized prediction input features, the regional net carbon emissions are predicted using a pre-trained convolutional long short-term memory network model.

[0066] like Figure 1 As shown, this embodiment provides a training process of a method for long-term prediction of regional net carbon emissions based on multi-factor mixed determination, including:

[0067] Step 1: Preprocess the collected data to correct outliers and fill in missing values;

[0068] Step 2: Analyze the relationship between each influencing factor and carbon emissions through correlation analysis method, select the factors with greater influence as input features according to the analysis results, and standardize the data;

[0069] Step 3: Build a CNN-LSTM model, use CNN to filter out noise information in the time series data that is not helpful for prediction, and then use LSTM to extract higher-level feature representations;

[0070] Step 4: Optimize the parameters of the neural network through the improved particle swarm optimization algorithm (IPSO) to accurately and efficiently complete the carbon emission prediction.

[0071] As one or more embodiments, the step 1: preprocessing the collected data to correct outliers and fill in missing values, specifically includes the following steps:

[0072] The carbon emission data is obtained and preprocessed, specifically:

[0073] Identify abnormal data using vertical processing of data and horizontal processing of data;

[0074] The abnormal data identified are smoothed to complete the abnormal value correction and obtain the pre-processed carbon emission data, which is as follows:

[0075] If the amount of missing data on a certain day is small, the linear interpolation method is used, and the fill-in value needs to be inserted at time t+j between time t and time t+n on day d;

[0076] When the amount of missing data on a certain day is large, the longitudinal interpolation method based on similar days is used. According to the periodic nature of data changes, the average value of multiple similar emission data is used to replace the missing data.

[0077] The method of identifying abnormal data is vertical processing of data and horizontal processing of data. The purpose of outlier correction is to smooth the abnormal data by identifying it. Specifically:

[0078] Step 101: Vertical processing of data

[0079] In view of the periodicity of carbon emission data, the change curves of similar days are similar, so we can judge whether there is abnormal data by judging the change range of emissions on similar days. After setting a certain error range, if there is a large difference between the data of two similar days, which obviously exceeds the error range, it is considered that there is abnormal data. The judgment method is as follows:

[0080]

[0081] The correction method is as follows:

[0082]

[0083] Among them, t represents the time when the abnormal value occurs, y(i,t) is the corrected value at that time, i represents the day, m(t) represents the average historical load data of t days, α(t), β(t) and ξ(t) all represent thresholds, which can be set according to actual conditions.

[0084] Step 102: Horizontal processing of data

[0085] Under normal circumstances, historical emission data is smooth and continuous, that is, the values ​​of two adjacent points are relatively close. If, during the data preprocessing process, it is found that the load value difference between two adjacent points is too large and exceeds the set error threshold range, it is considered to be abnormal load data. The judgment method is as follows:

[0086]

[0087] The correction method is as follows:

[0088]

[0089] Where t represents the time when the abnormal value appears, y(i,t) is the corrected value at that time, i represents the day, α(t) and β(t) both represent thresholds, which can be set according to actual conditions.

[0090] Step 103: Supplementation of missing data

[0091] If the amount of missing data on a certain day is small, linear interpolation is used. If the value L needs to be inserted at time t+j between time t and time t+n on day d, (d,t+j) , the calculation formula is as follows:

[0092]

[0093] Among them, L (d,t+j) is the emission value that needs to be filled at time t+j, L (d,t) is the emission value at time t on day d, L (d,t+b) is the emission value at time t+n on day d.

[0094] However, when the amount of missing data on a certain day is large, the use of linear interpolation will have a major drawback, resulting in insufficient sensitivity to the law of periodic changes, which will cause serious distortion of the data. At this time, it is more reasonable to use the longitudinal interpolation method based on similar days. The longitudinal interpolation method of similar days is based on the periodic characteristics of data changes, and uses the average value of multiple similar emission data to replace the missing data. The calculation formula is as follows:

[0095]

[0096] Among them, L (d,t+j) is the emission value that needs to be filled at time t+j, M is the number of days to select similar days, is the emission value at time t on a similar day.

[0097] As one or more embodiments, the step 2: analyzing the relationship between each influencing factor and carbon emissions by a correlation analysis method, selecting factors with greater influence as input features and performing a normalization operation according to the analysis results, specifically includes the following steps:

[0098] Step 201: Correlation Analysis

[0099] Based on the pre-processed carbon emission data, the relationship between each influencing factor and carbon emissions is analyzed through correlation analysis method. According to the analysis results, the factors with the greatest influence are selected as the prediction input features, which are as follows:

[0100] Calculate the corresponding entropy based on the marginal probability density of influencing factors and carbon emissions respectively;

[0101] Calculate the corresponding joint entropy based on the joint probability density of influencing factors and carbon emissions;

[0102] The maximum value of the entropy of the influencing factors and carbon emissions is taken as the minimum value of the joint entropy, and the sum of the entropy of the influencing factors and carbon emissions is taken as the maximum value of the joint entropy to determine the value range of the joint entropy; that is, the maximum value is obtained when the influencing factors and carbon emissions are completely independent of each other, and the minimum value is obtained when they are dependent on each other;

[0103] A variety of influencing factors are screened as characteristic inputs of regional carbon emissions through the threshold of the maximum information coefficient.

[0104] Correlation analysis aims to analyze the relationship between various influencing factors and carbon emissions. The maximum information coefficient method can simultaneously reflect the linear correlation and nonlinear correlation between variables. Assume that there are two random variables W = {w i , i=1,2,…,n} and Z={z i, i=1,2,…,n}, p(w) and p(z) are the marginal probability densities of variables W and Z respectively; p(w,z) is the joint probability density of the two variables, then variable W={w i ,i=1,2,…,n}, the entropy H(W) is:

[0105] H(W)=-∑ w∈W p(w)log 2 p(w);

[0106] The joint entropy H(W,Z) of variables W and Z is:

[0107] H(W,Z)=-∑ w∈W ∑ z∈Z p(w,z)log 2 p(w,z);

[0108] The value range of joint entropy H(W,Z) is:

[0109] max{H(W),H(Z)}≤H(W,Z)≤H(W)+H(Z);

[0110] The maximum value is achieved when the variables W and Z are completely independent of each other, and the minimum value is achieved when they are dependent on each other.

[0111] Mutual information is a quantitative indicator used to measure whether two random variables are related. If the two variables have no relationship and are independent of each other, the mutual information is zero. If the two variables are related to each other, the mutual information is a positive value representing the strength of the association. i , i=1,2,…,n} and Z={z i ,i=1,2,…,z} is expressed as:

[0112]

[0113] In the carbon emission prediction task, the variables W and Z represent the carbon emissions at each moment and other influencing factors respectively. From the above formula, we can see that the mutual information value MI is correlated with the entropy of each variable, and the relationship can be expressed as:

[0114] MI(W,Z)=H(W)+H(Z)-H(W,Z);

[0115] The maximum information coefficient MIC is based on mutual information and also combines the grid partitioning method. Assuming that the variables W and Z constitute a set of n ordered element pairs D = {(w i ,z i), i=1,2…,n}, divide the grid G ​​with a self-defined size of a*b, that is, divide W into a parts and Z into b parts, calculate MI in each grid, divide the set in multiple ways under the same grid size, and take the maximum MI as the MI value of the grid G, then:

[0116] MI * (D,w,z)=maxMI(D|G);

[0117] Where D|G refers to dividing the set D into a grid G, and then taking the maximum MI value obtained under multiple partitioning methods to obtain the normalized feature matrix M(D) w,z , as shown below:

[0118]

[0119] Then the MIC calculation formula for given variables W and Z is:

[0120]

[0121] Among them, B(n) is the upper limit of the maximum area for searching grids of different areas. The value of MIC is strictly controlled in the interval [0,1]. The stronger the correlation, the closer the value is to 1. The threshold of MIC can be used to screen multiple factors as characteristic inputs of regional carbon emissions.

[0122] A MIC threshold is set according to specific application requirements. Only when the MIC value of a factor is greater than this threshold, it is considered to be a feature that has a significant correlation with carbon emissions. All factors greater than the threshold are used as feature inputs, namely, predicted input features, for subsequent modeling processes.

[0123] Step 202: Data Standardization

[0124] For predicting input features, this paper uses the min-max normalization method to limit the value of the data sample to the interval [0,1]. The specific formula is as follows:

[0125]

[0126] Where X is the data vector to be normalized; X′ is the data vector after normalization; max(X) and min(X) are the maximum and minimum values ​​in the original samples, respectively.

[0127] Normalized data is used to input the model for training and learning, and the output is also normalized data. The final prediction result needs to be denormalized. The calculation formula is as follows:

[0128] X=X′×(max(X)-min(X))+min(X);

[0129] As one or more embodiments, step 3: constructing a CNN-LSTM model, filtering out noise information in the time series data that is not helpful for prediction through CNN, and then extracting a higher-level feature representation through LSTM, specifically includes the following steps:

[0130] The one-dimensional CNN is integrated with LSTM to construct a CNN-LSTM model, which fully utilizes the feature information extraction capability of CNN and the sensitivity of LSTM to time series data to improve the effect of carbon emission prediction. Specifically:

[0131] The number of network layers of the CNN-LSTM fusion model constructed in this specification can be adjusted according to the complexity of regional carbon emission data and influencing factors. The model diagram is shown in Figure 2 As shown in Figure 1. The CNN part of the model is composed of a one-dimensional convolution layer and a maximum pooling layer. The one-dimensional convolution layer traverses the input carbon emission information, and uses the convolution kernel weight to perform convolution operations with the local sequence segment of the carbon emission information to obtain a preliminary feature matrix, which has stronger expressive power than the time series matrix of the original carbon emission information. The maximum pooling layer takes the feature matrix calculated by the previous convolution layer as input after attention weighting, slides the pooling window on the matrix sequence, takes the maximum value of the window for pooling each time it slides, and outputs a more expressive feature matrix. The carbon emission information is refined by the CNN part into the input of the LSTM part, which is more sensitive to time series information. The LSTM layer can model and integrate local features in the time series direction to extract higher-level feature representations. Finally, the fully connected layer maps these feature representations to a single output and outputs the carbon emission prediction results.

[0132] set up is the input vector of the carbon emission dataset, and n represents the number of variables. The output vector of the convolutional layer is The calculation formula is as follows:

[0133]

[0134] in, is the output vector based on the previous convolutional input layer Calculated, represents the offset of the j-th mapping feature, w is the weight of the convolution kernel, m is the filter index value, and σ is the ReLU activation function.

[0135] For each output feature channel j of the convolutional layer, the attention weight vector α can be trained j To calculate the importance of the channel, the model can learn the importance of different features, thereby improving the performance and generalization ability of the model. Channel attention weight α ij The calculation method is as follows:

[0136]

[0137]

[0138] Among them, e ij is the unnormalized attention score, α ij is the normalized attention weight. ij Output of the convolutional layer The weighted method is as follows:

[0139]

[0140] The weighted output of the attention layer is then connected to the next level through the pooling layer after the convolution layer. The pooling layer reduces the number of parameters and networks by reducing the spatial dimension, thereby reducing the computational cost. As input to the pooling layer, the max pooling layer operates as follows:

[0141]

[0142] The output after convolution and pooling operations As the input of LSTM. Assume that the input gate of LSTM unit at stage t is i t , the forget gate is f t , the output gate is o t , the hidden layer state is h t , then the relevant calculation formula for the unit at stage t is as follows:

[0143]

[0144] Among them, c t represents the unit state at stage t, σ represents the activation function, W is the weight matrix of each gate unit, b is the corresponding offset vector, and p t It represents the information output of CNN containing the key features of carbon emission prediction at time t.

[0145] In addition, the LSTM cell state c t and the hidden layer h t The calculation formula is as follows:

[0146]

[0147] The last layer of the CNN-LSTM model consists of a fully connected layer, which is the feature vector output of the LSTM unit being "flattened". Let h l ={h 1 ,h 2 ,…,hl}, where l is the number of LSTM units, and the output of LSTM is used as the input of the fully connected layer. The corresponding calculation formula of this layer is as follows:

[0148]

[0149] like Figure 3 As shown, as one or more embodiments, the step 4 optimizes the parameters of the neural network through an improved particle swarm optimization algorithm (IPSO) to accurately and efficiently complete the carbon emission prediction, and the specific steps include:

[0150] Assume that in a D-dimensional search space, the number of particles is N, and a particle X in the particle swarm i It can be expressed as X i =[x i1 x i2 …,x iN ], i = 1, 2, ..., N, a particle X in the particle swarm i The speed can be expressed as V i =[v i1 ,ν i2 ,…,ν iN ], i=1,2,…,N, then the calculation formula for adjusting the particle speed and position is as follows:

[0151]

[0152] Among them, w i is the inertia weight, t is the number of iterations, V id is the velocity of the ith particle in the dth dimension, X id is the position of the ith particle in the dth dimension, c i is the learning factor, the value is a non-negative constant, r i is a random number between [0,1]. The particle fitness function is:

[0153]

[0154] Where n is the number of predicted sample points, y and y′ are the expected output and actual output of the same sample point.

[0155] Embodiment 2

[0156] This embodiment provides a long-term prediction system for regional net carbon emissions based on multi-factor mixed determination, including:

[0157] A data processing module, configured to obtain carbon emission data and perform pre-processing;

[0158] The data analysis module is configured to analyze the relationship between various influencing factors and carbon emissions through a correlation analysis method based on the pre-processed carbon emission data, select factors with great influence as prediction input features according to the analysis results, and perform standardization processing;

[0159] The carbon emission prediction module is configured to predict regional net carbon emissions based on standardized prediction input features using a pre-trained convolutional long short-term memory network model.

[0160] The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment 1. It should be noted that the above modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0161] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0162] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0163] Embodiment 3

[0164] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps in the method for long-term prediction of regional net carbon emissions based on multi-factor mixed determination as described in the first embodiment above are implemented.

[0165] Embodiment 4

[0166] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for long-term prediction of regional net carbon emissions based on multi-factor mixed determination as described in the first embodiment above are implemented.

[0167] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0168] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0169] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0170] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0171] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0172] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A long-term prediction method for regional net carbon emissions based on multi-factor mixed determination, characterized in that: include: Obtain carbon emission data and pre-process it; Based on the pre-processed carbon emission data, the relationship between each influencing factor and carbon emission is analyzed through correlation analysis method. According to the analysis results, the factors with the greatest influence are selected as the prediction input features and standardized. Based on the standardized prediction input features, the regional net carbon emissions are predicted using a pre-trained convolutional long short-term memory network model.

2. The long-term prediction method for regional net carbon emissions based on multi-factor mixed determination as claimed in claim 1 is characterized in that: The carbon emission data is obtained and preprocessed, specifically: Identify abnormal data using vertical processing of data and horizontal processing of data; The identified abnormal data is smoothed to complete the outlier correction and obtain the pre-processed carbon emission data.

3. The long-term prediction method for regional net carbon emissions based on multi-factor mixed determination as claimed in claim 2 is characterized in that: The abnormal data identified is smoothed to complete abnormal value correction, and the pre-processed carbon emission data is obtained, specifically: If the amount of missing data on a certain day is small, the linear interpolation method is used, and the fill-in value needs to be inserted at time t+j between time t and time t+n on day d; When the amount of missing data on a certain day is large, the longitudinal interpolation method based on similar days is used. According to the periodic nature of data changes, the average value of multiple similar emission data is used to replace the missing data.

4. The long-term prediction method for regional net carbon emissions based on multi-factor mixed determination as claimed in claim 1 is characterized in that: Based on the pre-processed carbon emission data, the relationship between each influencing factor and carbon emission is analyzed by correlation analysis method, and the factors with the greatest influence are selected as the prediction input features according to the analysis results, specifically: Calculate the corresponding entropy based on the marginal probability density of influencing factors and carbon emissions respectively; Calculate the corresponding joint entropy based on the joint probability density of influencing factors and carbon emissions; The maximum value of the entropy of the influencing factors and carbon emissions is taken as the minimum value of the joint entropy, and the sum of the entropy of the influencing factors and carbon emissions is taken as the maximum value of the joint entropy to determine the value range of the joint entropy; that is, the maximum value is obtained when the influencing factors and carbon emissions are completely independent of each other, and the minimum value is obtained when they are dependent on each other; A variety of influencing factors are screened as characteristic inputs of regional carbon emissions through the threshold of the maximum information coefficient.

5. The long-term prediction method for regional net carbon emissions based on multi-factor mixed determination as claimed in claim 1, characterized in that: The method of predicting regional net carbon emissions based on standardized prediction input features and using a pre-trained convolutional long short-term memory network model includes: Based on the prediction input features transformed from the table, the convolution kernel weights are used to perform convolution operations with the local sequence segments of carbon emission information to obtain a preliminary feature matrix; The initial feature matrix is ​​weighted by attention as input, and a pooling window is used to slide on the matrix sequence. The maximum value of the window is taken for pooling each time, and the local feature matrix is ​​output. Based on the local feature matrix, local features are modeled and integrated in the time series direction to extract time series feature representation; These time series feature representations are mapped to a single output, which is the carbon emission prediction result.

6. The long-term prediction method for regional net carbon emissions based on multi-factor mixed determination as claimed in claim 5, characterized in that: The parameters of the convolutional long short-term memory network model are optimized by using an improved particle swarm optimization algorithm. The specific steps include: Assume that in a D-dimensional search space, the number of particles is N, and a particle X in the particle swarm i Represented as X i =[x i1 x i2 …, x iN ], i = 1, 2, ..., N, a particle X in the particle group i The speed is represented by V i =[v i1 , v i2 , …, v iN ], i = 1, 2, ..., N, then the calculation formula for adjusting the particle speed and position is as follows: Among them, w i is the inertia weight, t is the number of iterations, V id is the velocity of the ith particle in the dth dimension, X id is the position of the ith particle in the dth dimension, c i is the learning factor, the value is a non-negative constant, r i is a random number between [0, 1]; the particle fitness function is: Where n is the number of predicted sample points, y and y′ are the expected output and actual output of the same sample point.

7. The long-term prediction method for regional net carbon emissions based on multi-factor mixed determination as claimed in claim 5, characterized in that: The prediction input features based on table conversion are convolved with the local sequence segments of carbon emission information using convolution kernel weights to obtain a preliminary feature matrix, specifically: Calculate the importance of channels by training the attention weight vector; Then use the attention weight to weight the convolutional layer output to get the weighted output of the attention layer.

8. A long-term prediction system for regional net carbon emissions based on multi-factor mixed judgment, characterized by: include: A data processing module, configured to obtain carbon emission data and perform pre-processing; The data analysis module is configured to analyze the relationship between various influencing factors and carbon emissions through a correlation analysis method based on the pre-processed carbon emission data, select factors with great influence as prediction input features according to the analysis results, and perform standardization processing; The carbon emission prediction module is configured to predict regional net carbon emissions based on standardized prediction input features using a pre-trained convolutional long short-term memory network model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for long-term prediction of regional net carbon emissions based on multi-factor mixed determination according to any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the long-term prediction method for regional net carbon emissions based on multi-factor mixed determination according to any one of claims 1 to 4 are implemented.

Images