Urban scale carbon emission prediction method and system based on small sample data enhancement

Abstract

The invention relates to an urban scale carbon emission prediction method and system based on small sample data enhancement. The method comprises the following steps: firstly, acquiring urban night light data, correcting the data by using a multiple linear regression model to obtain a fitting light ratio, and calculating the carbon dioxide emission of each city by combining an urban night light data inversion calculation formula to obtain a carbon emission data set; performing data enhancement on the carbon emission data set by using an improved time sequence enhanced generative adversarial network model TimeTGAN to obtain a carbon emission enhanced data set; then training the space-time sequence prediction network model to obtain a trained space-time sequence prediction model; and finally, collecting urban spatial geographic position data, and inputting the urban spatial geographic position data into the trained space-time sequence prediction model to obtain a carbon emission prediction result. Compared with the prior art, the method has the advantages that the problem of small samples in carbon emission prediction is effectively solved, and the accuracy of a prediction result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission prediction, and in particular to a method and system for predicting urban-scale carbon emissions based on small sample data enhancement. Background Art

[0002] Today, the world is actively responding to climate change and promoting carbon emission reduction actions. As a highly concentrated gathering place of population, industry and economic activities, cities have huge carbon emissions and complex sources. Predicting carbon emissions at the city scale can provide indispensable decision-making basis for city managers to formulate scientific and reasonable carbon emission reduction strategies and plan green and low-carbon development paths.

[0003] However, the openness and complexity of the current urban system make it difficult to accurately define the accounting boundaries of carbon emissions and difficult to collect carbon emissions data. Existing carbon emissions monitoring technologies have limitations. It is difficult to achieve comprehensive, real-time and accurate monitoring of some scattered small emission sources, which increases the difficulty of data collection and leads to repeated data collection and inconsistency.

[0004] Currently, the amount of urban-scale carbon emissions data available is relatively small. The small amount of data samples cannot fully reflect the complex dynamic changes in urban carbon emissions, and cannot capture the subtle correlations between various influencing factors and carbon emissions. As a result, it is impossible to provide sufficient material support for building high-precision and high-reliability prediction models, which ultimately forms a typical small sample problem and greatly hinders the effective prediction of future urban carbon emissions. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a city-scale carbon emission prediction method and system based on small sample data enhancement.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] According to one aspect of the present invention, a method for predicting urban scale carbon emissions based on small sample data enhancement is provided, comprising the following steps:

[0008] S1. Obtain the nighttime urban light data and use a multiple linear regression model to correct the data to obtain the fitted light proportion;

[0009] S2. Based on the fitted light ratio, the carbon dioxide emissions of each city are calculated using the inversion calculation formula of the city night light data to obtain a carbon emission data set;

[0010] S3. Use the improved time series enhanced generative adversarial network model TimeTGAN to enhance the carbon emission dataset and obtain a carbon emission enhanced dataset;

[0011] S4. Use the carbon emission enhanced data set to train the spatiotemporal series prediction network model to obtain a trained spatiotemporal series prediction model;

[0012] S5. Collect urban spatial geographic location data and input it into the trained spatiotemporal series prediction model to obtain carbon emission prediction results.

[0013] As a preferred technical solution, the multivariate linear regression model in S1 incorporates socioeconomic indicators closely related to carbon dioxide emissions, including the proportion of industrial power data to the province's industrial power data, the proportion of PM2.5 emissions to the province's emissions, the proportion of population data to the province's total population data, the proportion of private car ownership to the province's ownership, and the proportion of retail sales data to the province's sales data.

[0014] As a preferred technical solution, the inversion calculation formula of urban night light data in S2 is specifically:

[0015]

[0016] Where CE1 is the city-level carbon emission data; NTC i is the corrected night light ratio, CE P It is provincial carbon emission data.

[0017] As a preferred technical solution, the process of data enhancement of the carbon emission dataset in S3 introduces a bidirectional temporal convolutional network (BiTCN) network and dilated causal convolution. The specific enhancement process is: the carbon emission dataset is encoded and input into the embedding function, feature extraction and data compression are performed in the embedding function and passed to the latent space representation, random data is output, and the random data is used as the input of the generator to generate time series data, and the authenticity is judged by the time series data discriminator. If the identification is passed, it is converted into a usable carbon emission enhanced dataset through the recovery function. If the identification fails, it is fed back to the generator, and new data is continued to be generated in the generator until the required preset amount of data is generated and integrated into the carbon emission enhanced dataset. As a preferred technical solution, the embedding network function is applied in the embedding network, and its specific formula is:

[0018] h S_TCN =TCN(h S_F ),h X_TCN =TCN(h X_F )

[0019] h S_F =FC(S),h X_F =FC(X)

[0020] In the formula, h S_F Represents the static characteristic h S The result after processing by the fully connected layer; h X_F Represents the tense feature h X The result after processing by the fully connected layer; h S_TCN is the potential space; h X_TCN is the temporal feature; S is the embedded static feature; X is the embedded temporal feature.

[0021] As a preferred technical solution, a recovery network function is applied in the recovery network, and its specific formula is:

[0022]

[0023] In the formula, For the static characteristics of the recovery; is the temporal features of the recovery, and is the intermediate step of the recovery network; is the static feature finally generated; is the final generated temporal feature; FC′ is the fully connected layer in the recovery network; TCN′ represents the temporal convolutional network in the recovery network.

[0024] As a preferred technical solution, the specific formula of the generator is:

[0025]

[0026] In the formula, TCN G is the temporal convolutional network in the generator network; FC G is the fully connected layer in the generator network; Z S is a static noise vector, and static features are generated through a fully connected layer network; For its potential representation; Z 1:T is the time series noise vector, and the entire time series is generated through the TCN network; For its potential representation.

[0027] As a preferred technical solution, the discriminator uses a bidirectional TCN network, and its specific formula is:

[0028]

[0029] u t =TCN forward (x t ,u t-1 ),v t =TCN backward (x t ,v t+1 )

[0030] h t=Combine(u t ,v t )

[0031] In the formula, is the evaluation result of the potential representation of static features; is the evaluation result of the potential representation of temporal features; u t For time series data x t At each time point t in , the recursive calculation formula under the forward hidden state represents the forward propagation result of the temporal convolutional network; v t For time series data x t At each time point t in the backward hidden state, the recursive calculation formula represents the backward propagation result of the temporal convolutional network; h t is the hidden state that combines the results of forward and backward propagation.

[0032] As a preferred technical solution, the spatiotemporal sequence prediction network model in S4 uses the convolutional long short-term memory network model (ConvLSTM), and its model formula is:

[0033] i t =σ(W xi *X t +W hi *H t-1 +W ci ⊙c t-1 +b i )

[0034] f t =σ(W xf *X t +W hf *H t-1 +W cf ⊙c t-1 +b f )

[0035] C t =f t ⊙C t-1 +i i ⊙tanh g (W xc *x t +W hc *H t-1 +b c )

[0036] o t =σ(W xo *X t +W h0 *H t-1 +W co ⊙C t +bo )

[0037] H t =o t ⊙tanh g (c t )

[0038] Where σ is the sigmoid activation function; tanh g is the hyperbolic tangent activation function; X t H is the city spatial location data of the current time step; t-1 is the prediction result of the previous time step; C t-1 is the memory unit of the previous time step; f t 、i t , C t and t They are forget gate, input gate, memory unit and output gate respectively; W xi , W xf , W xc , W xo 、b i 、b f 、b c and b o are all model parameters; H t Carbon emission prediction results.

[0039] According to another aspect of the present invention, a city-scale carbon emission prediction system based on small sample data enhancement is provided, the system is applied to the city-scale carbon emission prediction method based on small sample data enhancement as described above, and the system includes a carbon emission calculation module, a carbon emission data enhancement module, a model training module and a carbon emission prediction module;

[0040] Among them, the carbon emission calculation module is used to obtain the city's night light data, use the multivariate linear regression model to correct the data to obtain the fitted light ratio, and based on the fitted light ratio, use the city's night light data inversion calculation formula to calculate the carbon dioxide emissions of each city to obtain a carbon emission data set; the carbon emission data enhancement module is used to use the improved time series enhancement generative adversarial network model TimeTGAN to enhance the carbon emission data set to obtain a carbon emission enhanced data set; the model training module is used to use the carbon emission enhanced data set to train the spatiotemporal series prediction network model to obtain a trained spatiotemporal series prediction model; the carbon emission prediction module is used to collect urban spatial geographic location data, input it into the trained spatiotemporal series prediction model, and obtain carbon emission prediction results.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. In the present invention, the city night light data is first obtained, and the fitting light proportion is obtained by correcting the data using a multivariate linear regression model; based on the fitting light proportion, the carbon dioxide emissions of each city are calculated using the city night light data inversion calculation formula to obtain a carbon emission data set; the carbon emission data set is then enhanced using the improved time series enhanced generative adversarial network model TimeTGAN to obtain a carbon emission enhanced data set; the carbon emission enhanced data set is then used to train the spatiotemporal sequence prediction network model to obtain a trained spatiotemporal sequence prediction model; finally, the city spatial geographic location data is collected and input into the trained spatiotemporal sequence prediction model to obtain the carbon emission prediction result. The present invention generates carbon emission training data based on the improved time series enhanced generative adversarial network model TimeTGAN, expands and enriches the carbon emission training data, and effectively solves the small sample problem in carbon emission prediction.

[0043] 2. In the process of data enhancement of carbon emission data set, the present invention introduces BiTCN network and dilated causal convolution. The specific enhancement process is: after encoding the carbon emission data set, input it into the embedding function, perform feature extraction and data compression in the embedding function and pass it to the latent space representation, output random data, use the random data as the input of the generator to generate time series data, and use the time series data discriminator to identify the authenticity. If the identification is passed, it is converted into a usable carbon emission enhancement data set through the recovery function. If the identification is not passed, it is fed back to the generator, and new data is continuously generated in the generator until the required preset amount of data is generated and integrated into the carbon emission enhancement data set. Effectively capture the long-term dependencies in the time series data, thereby generating more accurate and coherent carbon emission enhancement data, and enriching the training resources for carbon emission prediction.

[0044] 3. The present invention incorporates social indicators closely related to carbon dioxide emissions into the multivariate linear regression model, including the proportion of industrial power data to the province's industrial power data, the proportion of PM2.5 emissions to the province's emissions, the proportion of population data to the province's total population data, the proportion of private car ownership to the province's ownership, and the proportion of retail sales data to the province's sales data. The results more accurately reflect how the socioeconomic characteristics of each city affect the relationship between NTL data and actual carbon dioxide emissions, and improve the accuracy of carbon dioxide emission estimates.

[0045] 4. In the present invention, a convolutional long short-term memory network model is used as the spatiotemporal series prediction network model, and it is trained based on the data generated by TimeTGAN, thereby improving the accuracy and reliability of carbon emission prediction at the urban scale. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the steps of the method for predicting carbon emissions at the city scale in the present invention;

[0047] Figure 2 It is a schematic diagram of the structure of the city-scale carbon emission prediction system in the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0049] Example 1

[0050] In this embodiment, a city-scale carbon emission prediction method based on small sample data enhancement is applied. The method steps are as follows: Figure 1 As shown, the following steps are included:

[0051] S1. Obtain the nighttime urban light data and use a multiple linear regression model to correct the data to obtain the fitted light proportion;

[0052] S2. Based on the fitted light ratio, the carbon dioxide emissions of each city are calculated using the inversion calculation formula of the city night light data to obtain a carbon emission data set;

[0053] S3. Use the improved time series enhanced generative adversarial network model TimeTGAN to enhance the carbon emission dataset and obtain a carbon emission enhanced dataset;

[0054] S4. Use the carbon emission enhanced data set to train the spatiotemporal series prediction network model to obtain a trained spatiotemporal series prediction model;

[0055] S5. Collect urban spatial geographic location data and input it into the trained spatiotemporal series prediction model to obtain carbon emission prediction results.

[0056] In this embodiment, first, the urban night light data within the range time is selected, and the public data set is calculated according to the IPCC carbon emission factor method. A cluster analysis is performed on the city scale according to the carbon emission calculation index to obtain the construction orientation type of each city, and this is used as the basis for the matching coefficient of the night lighting data (NTL) of each city to calibrate. Finally, the fitting coefficient obtained by the multivariate linear regression model and the inversion calculation formula of the night light data are used to estimate the carbon dioxide emissions of each city.

[0057] In this embodiment, socioeconomic indicators closely related to carbon dioxide emissions are included in the multivariate linear regression model, and the socioeconomic indicators include the proportion of industrial power data to the province's industrial power data, the proportion of PM2.5 emissions to the province's emissions, the proportion of population data to the province's total population data, the proportion of private car ownership to the province's ownership, and the proportion of retail sales data to the province's sales data.

[0058] In this embodiment, the calculation formula for inverting the city night light data is specifically:

[0059]

[0060] Where CE1 is the city-level carbon emission data; NTC i is the corrected night light ratio, CE P It is provincial carbon emission data.

[0061] In this embodiment, the carbon emission enhancement module uses the improved time series enhancement generative adversarial network model TimeTGAN to achieve carbon emission data enhancement.

[0062] By using the embedding network and the recovery network, the mapping between the carbon emission time series data features and the latent space is obtained, and the dimension of the carbon emission data features is reduced.

[0063] The static and temporal features of the carbon emission data are reconstructed using the restoration network, which is responsible for converting the latent representation produced by the generator back to the original data space.

[0064] The introduction of the bidirectional temporal convolutional network BiTCN (Bidirectional Temporal Convolutional Network) network and dilated causal convolution can effectively capture the long-term dependencies in time series data, thereby generating more accurate and coherent carbon emission enhancement data and enriching the training resources for carbon emission prediction.

[0065] In this embodiment, the process of data enhancement for the carbon emission dataset is as follows: the carbon emission dataset is encoded and input into an embedding function, feature extraction and data compression are performed in the embedding function and the data is passed to the latent space representation, random data is output, and the random data is used as the input of the generator to generate time series data, and the authenticity is judged by the time series data discriminator. If the identification passes, it is converted into a usable carbon emission enhanced dataset through a recovery function. If the identification fails, it is fed back to the generator, and new data is continued to be generated in the generator until the required preset amount of data is generated and integrated into the carbon emission enhanced dataset.

[0066] In this embodiment, an embedded network function is applied in the embedded network, and its specific formula is:

[0067] h S_TCN=TCN(h S_F ),h X_TCN =TCN(h X_F )

[0068] h S_F =FC(S),h X_F =FC(X)

[0069] In the formula, h S_F Represents the static characteristic h S The result after processing by the fully connected layer; h X_F Represents the tense feature h X The result after processing by the fully connected layer; h S_TCN is the potential space; h X_TCN is the temporal feature; S is the embedded static feature; X is the embedded temporal feature.

[0070] In this embodiment, a network recovery function is applied in the network recovery, and its specific formula is:

[0071]

[0072] In the formula, For the static characteristics of the recovery; is the temporal features of the recovery, and is the intermediate step of the recovery network; is the static feature finally generated; is the final generated temporal feature; FC′ is the fully connected layer in the recovery network; TCN′ represents the temporal convolutional network in the recovery network.

[0073] In this embodiment, the specific formula of the generator is:

[0074]

[0075] In the formula, TCN G is the temporal convolutional network in the generator network; FC G is the fully connected layer in the generator network; Z S is a static noise vector, and static features are generated through a fully connected layer network; For its potential representation; Z 1:T is the time series noise vector, and the entire time series is generated through the TCN network; For its potential representation.

[0076] In this embodiment, the discriminator uses a bidirectional TCN network, and its specific formula is:

[0077]

[0078] u t =TCN forward (xt ,u t-1 ),v t =TCN backward (x t ,v t+1 )

[0079] h t =Combine(u t ,v t )

[0080] In the formula, is the evaluation result of the potential representation of static features; is the evaluation result of the potential representation of temporal features; u t For time series data x t At each time point t in , the recursive calculation formula under the forward hidden state represents the forward propagation result of the temporal convolutional network; v t For time series data x t At each time point t in the backward hidden state, the recursive calculation formula represents the backward propagation result of the temporal convolutional network; h t is the hidden state that combines the results of forward and backward propagation.

[0081] In this embodiment, we first collect data on city-scale influencing factors, select 18 carbon emission influencing factors such as industrial structure, energy structure, and energy emissions related to carbon emissions, and perform Pearson analysis to select the 10 most correlated influencing factors as eigenvalues.

[0082] Furthermore, a spatial correlation analysis of urban-scale carbon emissions was conducted, and it was found that urban-scale carbon emissions have the characteristics of spatial positive correlation.

[0083] Since the urban carbon emission problem has the characteristic of spatial positive correlation, the urban spatial geographic location data is introduced, and the carbon emission spatiotemporal series prediction model ConvLSTM is selected to predict urban-scale carbon emissions.

[0084] In this embodiment, the spatiotemporal sequence prediction network model uses a convolutional long short-term memory network model, and its model formula is:

[0085] i t =σ(W xi *X t +W hi *H t-1 +W ci ⊙c t-1 +b i )

[0086] f t =σ(W xf *X t +Whf *H t-1 +W cf ⊙c t-1 +b f )

[0087] C t =f t ⊙C t-1 +i i ⊙tanh g (W xc *x t +W hc *H t-1 +b c )

[0088] o t =σ(W xo *X t +W h0 *H t-1 +W co ⊙C t +b o )

[0089] H t =o t ⊙tanh g (C t )

[0090] Where σ is the sigmoid activation function; tanh g is the hyperbolic tangent activation function; X t H is the city spatial location data of the current time step; t-1 is the prediction result of the previous time step; C t-1 is the memory unit of the previous time step; f t 、i t , C t and t They are forget gate, input gate, memory unit and output gate respectively; W xi , W xf , W xc , W xo 、b i 、b f 、b c and b o are all model parameters; H t Carbon emission prediction results.

[0091] In summary, this method proposes to generate carbon emission training data based on the improved time series enhanced generative adversarial network model TimeTGAN, which expands and enriches the carbon emission training data and effectively solves the small sample problem in carbon emission prediction.

[0092] Example 2

[0093] In this embodiment, a city-scale carbon emission prediction system based on small sample data enhancement is applied. Figure 2 As shown, it includes a carbon emission calculation module, a carbon emission data enhancement module, a model training module and a carbon emission prediction module;

[0094] Among them, the carbon emission calculation module is used to obtain the city night light data, use the multivariate linear regression model to correct the data to obtain the fitted light ratio, and based on the fitted light ratio, use the city night light data inversion calculation formula to calculate the carbon dioxide emissions of each city to obtain the carbon emission data set;

[0095] The carbon emission data enhancement module is used to enhance the carbon emission dataset using the improved time series enhancement generative adversarial network model TimeTGAN to obtain a carbon emission enhanced dataset;

[0096] The model training module is used to train the spatiotemporal series prediction network model using the carbon emission enhanced data set to obtain a trained spatiotemporal series prediction model;

[0097] The carbon emission prediction module is used to collect urban spatial geographic location data and input it into the trained spatiotemporal series prediction model to obtain carbon emission prediction results.

[0098] In this embodiment, the urban night light data within the range time is first selected, and the public data set is calculated according to the IPCC carbon emission factor method.

[0099] In this embodiment, based on data such as urban employment and industrial structure, cities are clustered by k-means clustering algorithm (K-meansk) according to carbon emission calculation indicators, and cities are divided into resource-supported, private economy-supported, industrial support-supported, high-tech and other types. The construction orientation type of each city is obtained, and based on this, the matching coefficient of the night lighting data (NTL) of each city is corrected.

[0100] Then, a multivariate linear regression model was constructed to calculate the correction model for cities of different categories. Combined with the provincial carbon emission data, the urban carbon emissions were estimated based on the correction of the proportion of night lights in the area of ​​each city.

[0101]

[0102] PCE n,m =a0+a n x n,m +ε i

[0103] PCEn+1,m =a0+a n x n,m +a n+1 x n+1,m +ε i

[0104] PCE n+2,m =a0+a n x n,m +a n+1 x n+1,m +a n+2 x n+2,m +ε i

[0105] Where PCE i is the ratio of carbon emissions of city i to provincial carbon dioxide emissions calculated by the multivariate linear regression model, and is the dependent variable. This ratio is calculated by the linear regression model; a0 is the intercept term, and a n Indicates that this is the independent variable x added in step n n,m The regression coefficient of x n,m For every unit change in the dependent variable PCE i The number of units of expected change. ,β j is the independent variable X j The regression coefficient of , j is the number of independent variables. i is the error term. n and m represent the number of regression steps and city type in the hierarchical regression model, respectively. As the step increases by 1, an additional variable is added in step n+i.

[0106] In this embodiment, a series of multi-source and multi-dimensional socio-economic indicators that are closely related to carbon dioxide emissions are included in the multivariate linear regression model, including the proportion of industrial power data to the province's industrial power data (IEC), the proportion of PM2.5 emissions to the province's emissions (PM), the proportion of population data to the province's total population data (POP), and the proportion of private car ownership to the province's ownership (PCO), etc., and the proportion of retail sales data to the province's sales data (RC), etc., and the matching coefficient of the night lighting data (NTL) of each city is corrected. In this paper, POP is added to the first layer, and on the basis of the first layer, IEC, PM, PCO, and RC are added to the second, third, fourth, and fifth layers respectively. Variables with significance and r-squared changes are selected as independent variables to construct a specific regression model for this type of city. The results more accurately reflect how the socio-economic characteristics of each city affect the relationship between NTL data and actual carbon dioxide emissions, and improve the accuracy of carbon dioxide emission estimation. The multivariate hierarchical regression results are shown in Table 1.

[0107] Table 1 Multivariate stratified regression results

[0108]

[0109] Note: ***p<0.01; **p<0.05; *p<0.1

[0110] In this embodiment, R2 is the coefficient of determination, which indicates the proportion of the independent variables in the model to explain the variation of the dependent variable. Its value is between 0 and 1. The closer the value is to 1, the better the fitting effect of the model is, and the higher the proportion of the independent variables that can explain the variation of the dependent variable is. Adjusting R2 is a correction to R2 so that it can more accurately reflect the explanatory power of the model. The F value is a statistic used to test the overall significance of the multivariate regression model. It tests whether the explanatory power of all independent variables in the model for the dependent variable is significant. If the F value is large, it means that the model as a whole has a good explanatory power for the dependent variable. When performing stepwise regression analysis, the △F value indicates the change in the F statistic after adding or deleting a variable in the model. If the △F value is large after adding a variable, and the corresponding p value is less than the significance level (such as 0.05), it means that this variable has a significant contribution to the model and should be retained in the model.

[0111] In this embodiment, the above regression results show that the selected variables have different effects on the five types of cities. Therefore, according to the multivariate regression model to which the city belongs, the new proportion of urban-scale carbon emissions to provincial-level emissions is obtained. Taking 2017-2020 as an example, the correction formula for fitting the lighting proportion based on socio-economic factors is obtained as shown in Table 2.

[0112] Table 2 Lighting proportion correction formula based on socio-economic factors

[0113]

[0114] In this embodiment, the light proportion based on socioeconomic factors obtained by the multivariate linear regression model is used to estimate the carbon dioxide emissions of each city, and the carbon dioxide emission characteristics at the city level are analyzed accordingly. The night light data inversion model formula is as follows:

[0115]

[0116] Where NTC i is the corrected night light ratio, CE P It is the provincial carbon emission data. Based on the rough calculation, the city-level carbon emission data CE1 can be obtained.

[0117] Finally, the fitting coefficients obtained from the multivariate linear regression model and the inversion calculation formula of night light data were used to estimate the carbon dioxide emissions of each city.

[0118] In this embodiment, the carbon emission enhancement module uses the improved time series enhancement generative adversarial network model TimeTGAN to achieve carbon emission data enhancement, which includes an embedding network, a recovery network, a generator and a discriminator.

[0119] In this embodiment, the real data of carbon emissions is encoded, and through the embedding function, feature extraction, data compression and other operations are performed, and then passed to the latent space representation. At this time, the real data is converted into random data, and the random data is used as the input of the generator to generate new time series data. The time series data discriminator is used to identify and judge the authenticity. If the identification passes, it is converted into a usable carbon emission enhanced data set through the recovery function. If the identification fails, it is fed back to the sequence generator to continue to generate new data in the sequence generator until a sufficient amount of usable carbon emission enhanced data set is generated.

[0120] In this embodiment, the embedding network is constructed by combining the fully connected layer FC and the temporal convolutional network TCN. The fully connected layer FC first maps the static features to a latent space h S_TCN Then the temporal convolutional network TCN further processes the temporal features h based on this intermediate latent space X_TCN , to capture the dynamic changes and time dependencies of timing features.

[0121] The embedding network function is as follows:

[0122] h S_TCN =TCN(h S_F ),h X_TCN =TCN(h X_F )

[0123] h S_F =FC(S),h X_F =FC(X)

[0124] In the formula, h S_F Represents the static characteristic h S The result after being processed by the fully connected layer (FC); h X_F Represents the tense feature h X The result after processing by the fully connected layer; h S_TCN is the potential space; h X_TCN is the temporal feature; S is the embedded static feature; X is the embedded temporal feature.

[0125] In this embodiment, the recovery network reconstructs the static and temporal features of the carbon emission data and is responsible for converting the potential representation generated by the generator back to the original data space. The recovery network function is as follows:

[0126]

[0127] In the formula, For the static characteristics of the recovery; is the temporal feature to be restored, is the intermediate step of restoring the network, is the static feature finally generated; is the final generated temporal feature, FC′ is the fully connected layer in the recovery network, and TCN′ represents the temporal convolutional network in the recovery network.

[0128] In this embodiment, a generator is constructed using a fully connected layer and a bidirectional temporal convolutional network Bi-TCN. The generator network consists of two parts. S and Z 1:T is a vector space, generating H S and H X , the generator function uses static and temporal random vector groups to generate potential codes. The generator function is as follows:

[0129]

[0130] Where TCNG represents the temporal convolutional network in the generator network, FCG represents the fully connected layer in the generator network, and the static noise vector Z S The static features are generated by the fully connected layer network, and the potential representation is Time series noise vector Z 1:T The entire time series is generated by the TCN network, and the potential representation is

[0131] In this embodiment, the discriminator uses a bidirectional TCN to analyze the potential representation, and the discriminator function is as follows:

[0132]

[0133] u t =TCN forward (x t ,u t-1 ),v t =TCN backward (x t ,v t+1 )

[0134] h t =Combine(u t ,v t )

[0135] In the formula, represents the evaluation result of the potential representation of static features, Represents the evaluation result of the latent representation of temporal features. t For time series data x tAt each time point t in the forward hidden state, the recursive calculation formula represents the forward propagation result of the temporal convolutional network, v t For time series data x t At each time point t in , the recursive calculation formula in the backward hidden state represents the backward propagation result of the temporal convolutional network. In the bidirectional TCN, each time point fully represents h t represents the hidden state that combines the results of forward and backward propagation.

[0136] In this embodiment, a time series classification model BiTCN is introduced to distinguish the original and generated datasets. TimeTGAN is tested and compared with different enhancement methods based on the discrimination score. The comparison results are shown in Table 3. TimeTGAN always generates higher quality synthetic data. The generated samples reach 0.101, which is 23% lower than the second best baseline (TimeGAN, 0.171), indicating that the model performance of TimeTGAN is significantly better than other models.

[0137] Table 3 Identification test results

[0138]

[0139] In this embodiment, during the carbon emission prediction process, the carbon emission is first analyzed for spatial correlation:

[0140] Moran's I analysis was used to analyze the spatial correlation of carbon emissions in the Yangtze River Delta region and explore the spatial agglomeration of emissions in the Yangtze River Delta urban agglomeration. Moran's I analysis is mainly used to evaluate spatial distribution characteristics and their differences, and is divided into global Moran's I index and local Moran's I index (LISA). The global Moran's I index mainly reflects the agglomeration nature of the space as a whole, and the value range of the analysis result is -1 to 1. When the Moran's I index is greater than 0, it means there is spatial correlation, and the closer the analysis value is to 1, the higher the correlation is; when the Moran's I index is less than 0, it means there is spatial discreteness, and the closer the analysis value is to -1, the higher the discreteness is; when the Moran's I index is 0, it means there is no correlation between spatial units.

[0141] In this embodiment, taking the carbon emissions of the Yangtze River Delta urban agglomeration in a certain year as an example, ArcGIS software is used to perform a spatial correlation analysis of carbon emissions. The analysis results are shown in Table 4.

[0142] Table 4 Spatial correlation analysis of urban agglomerations in a certain year

[0143]

[0144] The Moran I index is 0.447169, indicating that the spatial distribution of the Yangtze River Delta urban agglomeration is positively correlated with carbon emissions. The Z (standard deviation multiple) score is 4.052541, which is greater than 2.58, indicating that the confidence interval of the data is greater than 99%. Therefore, it can be concluded that the carbon emissions of the Yangtze River Delta urban agglomeration have the characteristics of spatial agglomeration. Therefore, when predicting the carbon emissions of the Yangtze River Delta urban agglomeration, the urban spatial geographic location data should be introduced, and an algorithm model suitable for spatiotemporal series prediction should be selected to predict carbon emissions in the Yangtze River Delta region.

[0145] In this embodiment, the data of urban-scale influencing factors are then collected, and 18 carbon emission influencing factors such as industrial structure, energy structure, and energy emissions related to carbon emissions are selected. A Pearson correlation analysis is performed, and 10 influencing factors such as population (10,000 people), urbanization rate, per capita GDP (yuan / person), and total energy consumption (10,000 tons of standard coal) with P values ​​greater than 0.5 are extracted as characteristic values.

[0146] In this embodiment, since carbon emissions of urban agglomerations have spatial positive correlation, urban spatial geographic location data is introduced, and a spatiotemporal sequence prediction model is selected: Convolutional Long-Short Term Memory (ConvLSTM) network is used to predict urban scale carbon emissions. ConvLSTM is a spatiotemporal sequence prediction model that combines convolutional neural network CNN and long short-term memory network LSTM. The ConvLSTM model formula is as follows:

[0147] i t =σ(W xi *X t +W hi *H t-1 +W ci ⊙c t-1 +b i )

[0148] f t =σ(W xf *X t +W hf *H t-1 +W cf ⊙c t-1 +b f )

[0149] C t =f t ⊙C t-1 +i i ⊙tanh g (W xc *x t +W hc *H t-1 +b c)

[0150] o t =σ(W xo *X t +W h0 *H t-1 +W co ⊙C t +b o )

[0151] H t =o t ⊙tanh g (C t )

[0152] Where σ is the sigmoid activation function, tanh g represents the hyperbolic tangent activation function, X t is the input of the current time step, H t-1 is the hidden state of the previous time step, C t-1 is the memory unit of the previous time step, f t ,i t , C t , o t Respectively represent the forget gate, input gate, memory unit and output gate, W and b are model parameters. t The value of the input gate is calculated, which determines how much new information will be added to the cell state; f t Calculate the value of the forget gate, which determines how much old information will be forgotten from the cell state; C t What is calculated is the cell state at the current time step. The cell state is the channel for information flow in the network, which transmits information through time steps; t The output gate determines how much information about the cell state will be output to the hidden state. The final prediction result to be output is the hidden state H t The hidden state is the output of the ConvLSTM network at each time step, which contains the information learned from the input sequence and can be used for subsequent carbon emission prediction tasks.

[0153] In this embodiment, in the prediction network, the parameters in the ConvLSTM model are particularly considered. In the experimental setting, the time step of all models is uniformly set to 16. In the prediction network, the number of layers is also uniformly set to 2, and the number of hidden layer units is set to 12, 16, 12, and 10 according to the dimensions of different data features, and the number of iterations (epoch) is uniformly set to 5000. To ensure the accuracy and reliability of the evaluation, the present invention uses three different evaluation indicators to compare the prediction results, including root mean square error RMSE, mean absolute error MAE, and mean absolute percentage error MAPE to ensure the accuracy of the results.

[0154] Table 5 Comparison of prediction models

[0155]

[0156] The comparison of prediction results is shown in Table 5. The comparison of prediction results shows that the prediction model trained with the data generated by TimeTGAN is significantly better than other models, indicating that the data generated by the TimeTGAN model is closer to the original distribution of carbon emission data, effectively expanding the scale of urban-scale carbon emission data sets, making it possible to perform effective data analysis and model training even with small samples. This not only solves the small sample problem, but also helps to improve the accuracy and reliability of urban-scale carbon emission predictions, which is of great significance to the formulation of urban-scale carbon emission reduction policies.

[0157] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A city-scale carbon emission prediction method based on small sample data enhancement, characterized in that: The following steps are involved: S1. Obtain the nighttime urban light data and use a multiple linear regression model to correct the data to obtain the fitted light proportion; S2. Based on the fitted light ratio, the carbon dioxide emissions of each city are calculated using the inversion calculation formula of the city night light data to obtain a carbon emission data set; S3. Use the improved time series enhanced generative adversarial network model TimeTGAN to enhance the carbon emission dataset and obtain a carbon emission enhanced dataset; S4. Use the carbon emission enhanced data set to train the spatiotemporal series prediction network model to obtain a trained spatiotemporal series prediction model; S5. Collect urban spatial geographic location data and input it into the trained spatiotemporal series prediction model to obtain carbon emission prediction results.

2. The urban scale carbon emission prediction method based on small sample data enhancement according to claim 1 is characterized in that: The multivariate linear regression model in S1 incorporates social indicators that are closely related to carbon dioxide emissions, including the proportion of industrial power data to the province's industrial power data, the proportion of PM2.5 emissions to the province's emissions, the proportion of population data to the province's total population data, the proportion of private car ownership to the province's ownership, and the proportion of retail sales data to the province's sales data.

3. The urban scale carbon emission prediction method based on small sample data enhancement according to claim 1 is characterized in that: The inversion calculation formula of the city night light data in S2 is specifically: Where CE1 is the city-level carbon emission data; NTC i is the corrected night light ratio, CE P It is provincial carbon emission data.

4. The urban scale carbon emission prediction method based on small sample data enhancement according to claim 1 is characterized in that: The process of data enhancement for the carbon emission dataset in S3 introduces the BiTCN network and dilated causal convolution, and the specific enhancement process is: the carbon emission dataset is encoded and input into the embedding function, feature extraction and data compression are performed in the embedding function and passed to the latent space representation, random data is output, and the random data is used as the input of the generator to generate time series data, and the authenticity is judged by the time series data discriminator. If the identification is passed, it is converted into a usable carbon emission enhanced dataset through the recovery function. If the identification fails, it is fed back to the generator, and new data is continued to be generated in the generator until the required preset amount of data is generated and integrated into the carbon emission enhanced dataset.

5. The urban scale carbon emission prediction method based on small sample data enhancement according to claim 4 is characterized in that: The embedding network function is applied in the embedding network, and its specific formula is: h S_TCN =BC(h) S_F ),h X_TCN =BC(h) X_F ) h S_F =FC(S),h X_F =FC(X) In the formula, h S_F Represents the static characteristic h S The result after processing by the fully connected layer; h X_F Represents the tense feature h X The result after processing by the fully connected layer; h S_TCN is the potential space; h X_TCN is the temporal feature; S is the embedded static feature; X is the embedded temporal feature.

6. The urban scale carbon emission prediction method based on small sample data enhancement according to claim 4 is characterized in that: The recovery network function is applied in the recovery network, and its specific formula is: In the formula, For the static characteristics of the recovery; is the temporal features of the recovery, and is the intermediate step of the recovery network; is the static feature finally generated; is the final generated temporal feature; FC′ is the fully connected layer in the recovery network; TCN′ represents the temporal convolutional network in the recovery network.

7. The urban scale carbon emission prediction method based on small sample data enhancement according to claim 4 is characterized in that: The specific formula of the generator is: In the formula, TCN G is the temporal convolutional network in the generator network; FC G is the fully connected layer in the generator network; Z S is a static noise vector, and a static feature is generated through a fully connected layer network; For its potential representation; Z 1:T is the time series noise vector, and the entire time series is generated through the TCN network; For its potential representation.

8. The urban scale carbon emission prediction method based on small sample data enhancement according to claim 4 is characterized in that: The discriminator uses a bidirectional TCN network, and its specific formula is: u t =BC forward (x t ,u t-1 ),v t =BC backward (x t ,v t+1 ) h t =Combine(u t ,v t ) In the formula, is the evaluation result of the potential representation of static features; is the evaluation result of the potential representation of temporal features; u t For time series data x t At each time point t in , the recursive calculation formula under the forward hidden state represents the forward propagation result of the temporal convolutional network; v t For time series data x t At each time point t in the backward hidden state, the recursive calculation formula represents the backward propagation result of the temporal convolutional network; h t is the hidden state that combines the results of forward and backward propagation.

9. The urban scale carbon emission prediction method based on small sample data enhancement according to claim 1 is characterized in that: The spatiotemporal sequence prediction network model in S4 uses a convolutional long short-term memory network model, and its model formula is: i t =σ(W xi *X t +W hi *H t-1 +W ci ⊙c t-1 +b i ) f t =σ(W xf *X t +W hf *H t-1 +W cf ⊙c t-1 +b f ) C t =f t ⊙C t-1 +i i ⊙tanh g (W xc *x t +W hc *H t-1 +b c ) o t =σ(W xo *X t +W h0 *H t-1 +W co ⊙C t +b o ) H t =o t ⊙tanh g (C t ) Where σ is the sigmoid activation function; tanh g is the hyperbolic tangent activation function; X t H is the city spatial location data of the current time step; t-1 is the prediction result of the previous time step; C t-1 is the memory unit of the previous time step; f t 、i t , C t and t They are forget gate, input gate, memory unit and output gate respectively; W xi , W xf , W xc , W xo , b i , b f , b c and b o are all model parameters; H t Carbon emission prediction results.

10. A city-scale carbon emission prediction system based on small sample data enhancement, characterized in that: The system is applied to work by a city-scale carbon emission prediction method based on small sample data enhancement as described in any one of claims 1 to 9, and the system includes a carbon emission calculation module, a carbon emission data enhancement module, a model training module and a carbon emission prediction module; The carbon emission calculation module is used to obtain the city night light data, use the multivariate linear regression model to correct the data to obtain the fitted light ratio, and based on the fitted light ratio, use the city night light data inversion calculation formula to calculate the carbon dioxide emissions of each city to obtain a carbon emission data set; The carbon emission data enhancement module is used to enhance the carbon emission data set by using the improved time series enhancement generative adversarial network model TimeTGAN to obtain a carbon emission enhanced data set; The model training module is used to train the spatiotemporal sequence prediction network model using the carbon emission enhanced data set to obtain a trained spatiotemporal sequence prediction model; The carbon emission prediction module is used to collect urban spatial geographic location data, input it into the trained spatiotemporal sequence prediction model, and obtain carbon emission prediction results.

Images