Enhanced data driving building carbon emission integrated prediction system and method

Abstract

The invention discloses an enhanced data-driven building carbon emission integrated prediction system and method, and the method specifically comprises the steps: carrying out the amplification of building carbon emission data in a data enhancement module through employing a Bootstrap resampling method, and combining the amplified data with original data to construct a new enhanced data set; the feature extraction module uses an auto-encoder to extract cycle features of the building carbon emission time sequence data, separates original features, and obtains random features to form building carbon emission residual dynamic data; in the model prediction module, adaptive prediction models are respectively trained by using the cyclic characteristic data set and the residual dynamic data set, and the prediction models of the two parts are fused for prediction of building carbon emission. According to the invention, the problems of difficulty and insufficiency in building carbon emission data acquisition are solved; according to the method, the random feature processing capability is enhanced, the prediction accuracy and the operation rate of the prediction model can be improved, and the method has high precision, strong generalization capability and good expandability.

Description

Technical Field

[0001] The present invention relates to the technical field of building carbon emission prediction in the energy industry, and in particular to an enhanced data-driven integrated building carbon emission prediction system and method. Background Art

[0002] As global climate change and environmental issues become increasingly severe, carbon emissions in the construction sector have become an important focus of carbon neutrality goals. The construction industry is one of the main contributors to global energy consumption and carbon dioxide emissions. According to statistics, it accounts for nearly 40% of global carbon emissions. Scientific and accurate prediction of building carbon emissions is an important basis for promoting low-carbon building design and green building policies. Therefore, achieving carbon emission reduction targets in the construction industry is of great significance to achieving global carbon neutrality.

[0003] Building carbon emissions mainly come from the following aspects: carbon emissions from the combustion of fuels (such as natural gas, coal, etc.) used in building operations; carbon emissions caused by electricity consumed during building operations; including carbon emissions from the production, transportation, construction, operation and demolition of building materials. Its emissions are long-term and complex, especially during the entire life cycle of the building. The emission characteristics of different regions, climate conditions, energy structures and building types vary significantly. Data collection and calculation are difficult, and traditional methods cannot fully cover the dynamic characteristics of building carbon emissions.

[0004] Traditional carbon emission prediction methods are mainly based on linear models and statistical analysis, relying on simple calculations of building area, energy consumption data and carbon emission factors. These methods have the following shortcomings: lack of dynamic response capabilities, unable to effectively capture changes in carbon emissions during the building operation phase; lack of comprehensive consideration of multiple variables (such as climate, usage patterns, energy prices, etc.), limited prediction accuracy; difficult to adapt to the needs of different regions, building types or policy change scenarios, poor model applicability; strong data dependence, and weak ability to analyze large-scale, real-time data. Summary of the invention

[0005] The present invention provides an enhanced data-driven integrated prediction system and method for building carbon emissions, aiming to improve the accuracy and applicability of the prediction model and provide support for the low-carbon development of the construction industry.

[0006] The present invention provides the following technical solution: an enhanced data-driven integrated prediction method for building carbon emissions, comprising the following steps: S1. Use the Bootstrap resampling method to amplify the original data of building carbon emissions, obtain virtual data, and combine the virtual data with the original data to construct a new enhanced data set; S2. Use the autoencoder to extract the cyclic features of the building carbon emission enhancement data samples and construct a stable training set for building carbon emissions; separate the original data features, obtain random features to form the dynamic data of building carbon emission residuals, and convert the dynamic data of building carbon emission residuals into a training set of building carbon emission residuals; the specific steps are as follows: The first step is to read in the building carbon emission enhancement data and perform data standardization; The second step is to train the autoencoder model, including forward propagation and back propagation and parameter update of the encoding and decoding parts. The forward propagation process of the encoding part is as follows: ; Where: It is the cycle feature extracted from the encoding part; and They are the weight and bias of the encoding part respectively; is the activation function; the forward propagation process of the decoding part is as follows: ; Where: The decoding part is for loop special The reconstruction characteristics of the levy; and They are the weight and bias of the decoding part respectively; is the activation function; The third step is to use autoencoders to extract the cyclic features of building carbon emission data Construct a stable training set of building carbon emissions. This cyclic feature is a stable component of the original time series data. The fourth step is to delete the stable part of the original data to extract the dynamic data of building carbon emission residuals, and convert the dynamic data of building carbon emission residuals into a training set of building carbon emission residuals; S3, using two data sets, the building carbon emission stable training set and the building carbon emission residual training set, to train the prediction model; Among them, the building carbon emission stability training set is used to train the cycle characteristic model, and the cycle characteristic model is constructed by the ELM model; The building carbon emission residual training set is used to train the dynamic residual model. The dynamic residual model is composed of a T-LSTM model. The specific training process of the dynamic residual model is as follows: The first step is to preprocess the input data. The input data of the dynamic residual model is first standardized through the Layer Norm layer, and then the processed data is first sent to the gated MLP for up-projection to expand the data dimension and obtain richer data features. The second step is to encode the input data. By using the self-attention mechanism in Transformer, the query, key, and value of the input data are calculated to capture the global dependency, identify the correlation between each position in the sequence, and perform nonlinear transformation on the features of each position to enhance the feature expression capability. The third step is to process the EN-LSTM unit. EN-LSTM controls the flow of information by introducing multiple gating mechanisms. The data input to the EN-LSTM unit passes through the forget gate and the input gate in turn to complete the update of the unit state in EN-LSTM, and finally obtains the input of the next time step or the output for the final result through the output gate. In addition, an exponential gate is added before the input gate and the forget gate to introduce a branch of the standard state in the unit processing process. The fourth step is to feed the time series data processed by the LSTM unit into the gated MLP for down-projection; The fifth step is to adjust the model parameters by calculating the error between the model output and the true target until the model converges; finally, the result is linearly transformed in the output stage to obtain the final output; S4, fusing the two prediction models of the cyclic feature model and the dynamic residual model to form an integrated prediction model; S5. Use the enhanced data as the prediction data input of the integrated prediction model, use the integrated prediction model to predict building carbon emissions, and output the prediction results.

[0007] Preferably, the cyclical characteristic model captures the periodic changes in the time series, and the dynamic residual model is used to adjust the model prediction error.

[0008] Preferably, the output prediction result is a combination of the prediction results of the cyclic characteristic model and the dynamic residual model. The output of the cyclic characteristic model and the output of the dynamic residual model are fused together by a weighted average method to construct a complete building carbon emission time series prediction data curve.

[0009] Preferably, the process of obtaining virtual data is as follows: The first step is to use the original sample Constructing subsamples , randomly draw samples from the original sample and generate multiple sub-samples with replacement, set to 1000 times, and the size of each sub-sample is equal to the original sample size; The second step is to calculate the target statistic or model parameter for each subsample. ; The third step is to construct an approximate Bootstrap distribution for the target statistic of each subsample; The fourth step is to estimate the uncertainty of the parameters or construct a confidence interval based on the Bootstrap distribution, and take the samples within the confidence interval as virtual data.

[0010] Preferably, the specific steps of the ELM model training are as follows: The first step is to perform random feature mapping from the input layer to the hidden layer, and transfer the model weights and biases. The update process is as follows: ; In the formula, is the output of each neuron in the hidden layer; , are the weight and bias of the model respectively; is the activation function; The second step is to solve the linear parameters from the hidden layer to the output layer by setting the hidden layer allocation weights To obtain the optimal solution, the specific process is as follows: ; In the formula, The final output is is the hidden layer assignment weight.

[0011] Preferably, back propagation and Adam optimizer are used in S3 to perform parameter optimization and adjust model parameters.

[0012] A prediction system for an enhanced data-driven integrated prediction method for building carbon emissions, comprising a data enhancement module, a feature extraction module and a model prediction module; The data enhancement module is used to solve the problem of insufficient acquisition of building carbon emission data. The prediction system uses the Bootstrap resampling method in the data enhancement module to amplify the original building carbon emission data, and combines the amplified virtual data with the original data to construct a new enhanced data set; The feature extraction module is used to extract cyclic features from the enhanced data samples. The feature extraction module uses an autoencoder network model to extract cyclic features of the building carbon emission time series data, construct a stable training set for building carbon emissions, separate the original data features, delete the stable part of the original data, and obtain random feature data to form the building carbon emission residual dynamic data; The model prediction module uses the building carbon emission stability training set and the building carbon emission residual training set generated by the feature extraction module to train the corresponding prediction model, fuses the prediction model into an integrated prediction model, and uses the integrated prediction model to predict the building carbon emission; The model prediction module includes a cyclic characteristic model and a dynamic residual model; the building carbon emission stability training set is used for training the cyclic characteristic model, and the cyclic characteristic model is constructed by the ELM model; the building carbon emission residual training set is used for training the dynamic residual model, and the dynamic residual model is composed of a T-LSTM model.

[0013] Preferably, the T-LSTM model integrates gated MLP and EN-LSTM, and combines an attention mechanism.

[0014] Preferably, the prediction data input of the integrated prediction model is enhanced data.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This enhanced data-driven integrated prediction system and method for building carbon emissions has high accuracy, strong generalization ability and good scalability, and can be widely used in green building design and energy management and other fields.

[0016] 2. The enhanced data-driven integrated prediction system and method for building carbon emissions uses the Bootstrap data resampling method to amplify the original data, and combines the original data and the expanded data into a new data set, which greatly solves the problem of difficulty and insufficiency in obtaining building carbon emission data.

[0017] 3. The enhanced data-driven integrated prediction system and method for building carbon emissions divides the building carbon emission time series data into cyclic feature data and random feature data through the autoencoder network model, uses the ELM model and T-LSTM model to train the cyclic feature data and dynamic residual data in the original data, and combines the prediction results of the two models to form the final prediction curve, which not only helps to improve the prediction accuracy of the model, but also helps to improve the model's computing speed.

[0018] 4. The enhanced data-driven integrated prediction system and method for building carbon emissions designed a dynamic residual prediction model T-LSTM, which combines gated MLP and EN-LSTM. By combining the attention mechanism, the processing ability of random features is enhanced to accurately predict the dynamic residual time series data of building carbon emissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic diagram of the structure of a prediction method of an enhanced data-driven building carbon emission integrated prediction system provided by the present invention; Figure 2 This is the ELM backbone network architecture diagram of the present invention; Figure 3 This is the EN-LSTM network structure diagram of the present invention. DETAILED DESCRIPTION

[0020] 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 only 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 are within the scope of protection of the present invention.

[0021] See also Figures 1 to 3 The present invention proposes an enhanced data-driven integrated prediction system and method for building carbon emissions, aiming to improve the accuracy and applicability of the prediction model and provide support for the low-carbon development of the construction industry. The present invention proposes an enhanced data-driven integrated prediction method for building carbon emissions, comprising the following steps: S1. Use the Bootstrap resampling method to amplify the original data of building carbon emissions to obtain virtual data, and combine the virtual data with the original data to construct a new enhanced data set to form samples for subsequent prediction model training; the process of obtaining virtual data is as follows: The first step is to use the original sample Constructing subsamples , randomly draw samples from the original sample and generate multiple sub-samples with replacement, set to 1000 times, and the size of each sub-sample is equal to the original sample size; The second step is to calculate the target statistic or model parameter for each subsample. ; The third step is to construct an approximate Bootstrap distribution for the target statistic of each subsample; The fourth step is to estimate the uncertainty of the parameters or construct a confidence interval based on the Bootstrap distribution, and take the samples within the confidence interval as virtual data.

[0022] S2. Use the autoencoder to extract the cyclic features of the building carbon emission enhancement data samples and construct a stable training set for building carbon emissions; separate the original data features, obtain random features to form the dynamic data of building carbon emission residuals, and convert the dynamic data of building carbon emission residuals into a training set of building carbon emission residuals; the specific steps are as follows: The first step is to read in the building carbon emission enhancement data and perform data standardization; The second step is to train the autoencoder model, including forward propagation and back propagation and parameter update of the encoding and decoding parts. The forward propagation process of the encoding part is as follows: ; Where: It is the cycle feature extracted from the encoding part; and They are the weight and bias of the encoding part respectively; is the activation function; the forward propagation process of the decoding part is as follows: ; Where: The decoding part is for loop special The reconstruction characteristics of the levy; and They are the weight and bias of the decoding part respectively; is the activation function; The third step is to use autoencoders to extract the cyclic features of building carbon emission data Construct a stable training set of building carbon emissions. This cyclic feature is a stable component of the original time series data. The fourth step is to delete the stable part of the original data to extract the dynamic data of building carbon emission residuals, and convert the dynamic data of building carbon emission residuals into a building carbon emission residual training set.

[0023] S3. Use the two data sets of building carbon emission stable training set and building carbon emission residual training set to train the prediction model; among them, the building carbon emission stable training set is used for EML model training, and the building carbon emission residual training set is more random than the stable training set, so it is used for a more robust Transformer model; specifically, the following steps are included: The building carbon emission stability training set is used to train the cycle characteristic model, which is constructed by the ELM model; Figure 2 As shown, the specific steps of ELM model training are as follows: The first step is to perform random feature mapping from the input layer to the hidden layer, and transfer the model weights and biases. The update process is as follows: ; In the formula, is the output of each neuron in the hidden layer; , are the weight and bias of the model respectively; is the activation function; The second step is to solve the linear parameters from the hidden layer to the output layer by setting the hidden layer allocation weights To obtain the optimal solution, the specific process is as follows: ; In the formula, The final output is is the hidden layer allocation weight; The building carbon emission residual training set is used to train the dynamic residual model. The dynamic residual model is composed of a T-LSTM model. The specific training process of the dynamic residual model is as follows: The first step is to preprocess the input data. The input data of the dynamic residual model is first standardized through the Layer Norm layer, and then the processed data is first sent to the gated MLP for up-projection to expand the data dimension and obtain richer data features. The second step is to encode the input data. By using the self-attention mechanism in Transformer, the query, key, and value of the input data are calculated to capture the global dependency, identify the correlation between each position in the sequence, and perform nonlinear transformation on the features of each position to enhance the feature expression capability. The third step is to process the EN-LSTM unit. EN-LSTM controls the flow of information by introducing multiple gating mechanisms, so that it can effectively capture the long-term dependency of data. The data input to the EN-LSTM unit passes through the forget gate and the input gate in turn to complete the update of the unit state in EN-LSTM, and finally obtains the input of the next time step or the output for the final result through the output gate. In order to solve the strong randomness in the residual dynamic data, an exponential gate is added before the input gate and the forget gate respectively. In order to enhance the stability of the model, a standard state branch is introduced in the unit processing process. In the fourth step, the time series data processed by the LSTM unit is sent to the gated MLP for down-projection. While aligning the data format, the noise and interference in the output data are also filtered; The fifth step is to adjust the model parameters by calculating the error between the model output and the true target. The parameter optimization is completed using back propagation and Adam optimizer until the model converges. Finally, the result is linearly transformed in the output stage to obtain the final output; Figure 3 shown.

[0024] The cyclical characteristic model captures the cyclical changes in the time series, while the dynamic residual model is used to adjust the model prediction error, thereby improving the accuracy and stability of the overall prediction.

[0025] S4, fusing the two prediction models of the cyclic feature model and the dynamic residual model to form an integrated prediction model; S5. Use the enhanced data as the prediction data input of the integrated prediction model, use the integrated prediction model to predict the building carbon emissions, and output the prediction results. The output prediction results are a combination of the prediction results of the cyclic characteristic model and the dynamic residual model. The output of the cyclic characteristic model and the output of the dynamic residual model are fused together by the weighted average method to construct a complete building carbon emission time series prediction data curve.

[0026] From the above description, it can be seen that the present invention utilizes the Bootstrap data resampling method to amplify the original data when in use, and combines the original data and the expanded data into a new data set, which greatly solves the problem of difficulty and insufficiency in obtaining building carbon emission data.

[0027] The present invention combines the prediction results of the two prediction models to form a final prediction curve, which not only helps to improve the prediction accuracy of the model, but also helps to improve the calculation speed of the model.

[0028] The present invention designs a dynamic residual prediction model T-LSTM, which integrates gated MLP and EN-LSTM. By combining the attention mechanism, the processing ability of random features is enhanced to accurately predict the dynamic residual time series data of building carbon emissions.

[0029] The above prediction method is implemented through a prediction system, such as Figure 1 As shown, a prediction system used in an enhanced data-driven integrated prediction method for building carbon emissions adopts a modular architecture, including a data enhancement module, a feature extraction module and a model prediction module, and the data enhancement module, the feature extraction module and the model prediction module operate independently, so that the prediction system has good scalability, thereby improving the adaptability of the prediction system.

[0030] The data enhancement module is used to solve the problem of insufficient acquisition of building carbon emission data. The prediction system uses the Bootstrap resampling method in the data enhancement module to amplify the original building carbon emission data, and combines the amplified virtual data with the original data to construct a new enhanced data set for subsequent prediction model training; The feature extraction module is used to extract cyclic features from the enhanced data samples. The feature extraction module uses an autoencoder network model to extract the cyclic features of the building carbon emission time series data, construct a stable training set for building carbon emissions, and decompose the original data features, delete the stable part of the original data, and obtain random feature data to form the residual dynamic data of building carbon emissions; the feature extraction module includes the time series cyclic feature extraction of the autoencoder and the acquisition of residual dynamic data, and these two parts of data are constructed into a data set that can be used for subsequent training.

[0031] The model prediction module uses the building carbon emission stability training set and the building carbon emission residual training set generated by the feature extraction module to train the corresponding prediction model. Due to the different randomness of the two data features, different prediction models are selected, and the prediction models are integrated into a prediction model. The integrated prediction model is used to predict building carbon emissions, and the prediction data input of the integrated prediction model is enhanced data; The model prediction module includes a cyclic feature model and a dynamic residual model; the building carbon emission stability training set is used for training the cyclic feature model, and the cyclic feature model is constructed by the ELM model; the building carbon emission residual training set is used for training the dynamic residual model, and the dynamic residual model is composed of a T-LSTM model. The T-LSTM model integrates gated MLP and EN-LSTM, and combines the attention mechanism to enhance the processing ability of random features, so as to accurately predict the dynamic residual time series data of building carbon emissions.

[0032] In summary, this method has high accuracy, strong generalization ability and good scalability, and can be widely used in fields such as green building design and energy management.

[0033] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in the field. Although the embodiments of the present invention have been shown and described, it is understood by ordinary technicians in the field that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the attached claims and their equivalents.

Claims

1. An enhanced data-driven integrated prediction method for building carbon emissions, characterized in that: The following steps are involved: S1. Use the Bootstrap resampling method to amplify the original data of building carbon emissions, obtain virtual data, and combine the virtual data with the original data to construct a new enhanced data set; S2. Use the autoencoder to extract the cyclic features of the building carbon emission enhancement data samples and construct a stable training set for building carbon emissions; separate the original data features, obtain random features to form the dynamic data of building carbon emission residuals, and convert the dynamic data of building carbon emission residuals into a training set of building carbon emission residuals; the specific steps are as follows: The first step is to read in the building carbon emission enhancement data and perform data standardization; The second step is to train the autoencoder model, including forward propagation and back propagation and parameter update of the encoding and decoding parts. The forward propagation process of the encoding part is as follows: ； Where: It is the cycle feature extracted from the encoding part; and They are the weight and bias of the encoding part respectively; is the activation function; the forward propagation process of the decoding part is as follows: ； Where: The decoding part is for loop special The reconstruction characteristics of the levy; and They are the weight and bias of the decoding part respectively; is the activation function; The third step is to use autoencoders to extract the cyclic features of building carbon emission data Construct a stable training set of building carbon emissions. This cyclic feature is a stable component of the original time series data. The fourth step is to delete the stable part of the original data to extract the dynamic data of building carbon emission residuals, and convert the dynamic data of building carbon emission residuals into a training set of building carbon emission residuals; S3, using two data sets, the building carbon emission stable training set and the building carbon emission residual training set, to train the prediction model; Among them, the building carbon emission stability training set is used to train the cycle characteristic model, and the cycle characteristic model is constructed by the ELM model; The building carbon emission residual training set is used to train the dynamic residual model. The dynamic residual model is composed of a T-LSTM model. The specific training process of the dynamic residual model is as follows: The first step is to preprocess the input data. The input data of the dynamic residual model is first standardized through the Layer Norm layer, and then the processed data is first sent to the gated MLP for up-projection to expand the data dimension and obtain richer data features. The second step is to encode the input data. By using the self-attention mechanism in Transformer, the query, key, and value of the input data are calculated to capture the global dependency, identify the correlation between each position in the sequence, and perform nonlinear transformation on the features of each position to enhance the feature expression capability. The third step is to process the EN-LSTM unit. EN-LSTM controls the flow of information by introducing multiple gating mechanisms. The data input to the EN-LSTM unit passes through the forget gate and the input gate in turn to complete the update of the unit state in EN-LSTM, and finally obtains the input of the next time step or the output for the final result through the output gate. In addition, an exponential gate is added before the input gate and the forget gate to introduce a branch of the standard state in the unit processing process. The fourth step is to feed the time series data processed by the LSTM unit into the gated MLP for down-projection; The fifth step is to adjust the model parameters by calculating the error between the model output and the true target until the model converges; finally, the result is linearly transformed in the output stage to obtain the final output; S4, fusing the two prediction models of the cyclic feature model and the dynamic residual model to form an integrated prediction model; S5. Use the enhanced data as the prediction data input of the integrated prediction model, use the integrated prediction model to predict building carbon emissions, and output the prediction results.

2. The enhanced data-driven integrated prediction method for building carbon emissions according to claim 1, characterized in that: The cyclical characteristic model captures the periodic changes in the time series, and the dynamic residual model is used to adjust the model prediction error.

3. The enhanced data-driven integrated prediction method for building carbon emissions according to claim 2 is characterized by: The output prediction result is a combination of the prediction results of the cyclic characteristic model and the dynamic residual model. The output of the cyclic characteristic model and the output of the dynamic residual model are fused together through the weighted average method to construct a complete building carbon emission time series prediction data curve.

4. The enhanced data-driven integrated prediction method for building carbon emissions according to claim 1 is characterized by: The process of obtaining virtual data is as follows: The first step is to use the original sample Constructing subsamples , randomly draw samples from the original sample and generate multiple sub-samples with replacement, set to 1000 times, and the size of each sub-sample is equal to the original sample size; The second step is to calculate the target statistic or model parameter for each subsample. ; The third step is to construct an approximate Bootstrap distribution for the target statistic of each subsample; The fourth step is to estimate the uncertainty of the parameters or construct a confidence interval based on the Bootstrap distribution, and take the samples within the confidence interval as virtual data.

5. The enhanced data-driven integrated prediction method for building carbon emissions according to claim 4 is characterized by: The specific steps of ELM model training are as follows: The first step is to perform random feature mapping from the input layer to the hidden layer, and transfer the model weights and biases. The update process is as follows: ； In the formula, is the output of each neuron in the hidden layer; , are the weight and bias of the model respectively; is the activation function; The second step is to solve the linear parameters from the hidden layer to the output layer by setting the hidden layer allocation weights To obtain the optimal solution, the specific process is as follows: ； In the formula, The final output is is the hidden layer assignment weight.

6. The enhanced data-driven integrated prediction method for building carbon emissions according to claim 1, characterized in that: In the S3, back propagation and Adam optimizer are used to perform parameter optimization and adjust the model parameters.

7. A prediction system for an enhanced data-driven integrated prediction method for building carbon emissions according to any one of claims 1 to 6, characterized in that: Includes data enhancement module, feature extraction module and model prediction module; The data enhancement module is used to solve the problem of insufficient acquisition of building carbon emission data. The prediction system uses the Bootstrap resampling method in the data enhancement module to amplify the original building carbon emission data, and combines the amplified virtual data with the original data to construct a new enhanced data set; The feature extraction module is used to extract cyclic features from the enhanced data samples. The feature extraction module uses an autoencoder network model to extract cyclic features of the building carbon emission time series data, construct a stable training set for building carbon emissions, separate the original data features, delete the stable part of the original data, and obtain random feature data to form the building carbon emission residual dynamic data; The model prediction module uses the building carbon emission stability training set and the building carbon emission residual training set generated by the feature extraction module to train the corresponding prediction model, fuses the prediction model into an integrated prediction model, and uses the integrated prediction model to predict the building carbon emission; The model prediction module includes a cyclic characteristic model and a dynamic residual model; the building carbon emission stability training set is used for training the cyclic characteristic model, and the cyclic characteristic model is constructed by the ELM model; the building carbon emission residual training set is used for training the dynamic residual model, and the dynamic residual model is composed of a T-LSTM model.

8. The prediction system used in the enhanced data-driven integrated prediction method for building carbon emissions according to claim 7, characterized in that: The T-LSTM model combines gated MLP and EN-LSTM and incorporates an attention mechanism.

9. The prediction system used in the enhanced data-driven integrated prediction method for building carbon emissions according to claim 7, characterized in that: The prediction data input of the integrated prediction model is the enhanced data.

Images