Energy conservation and emission reduction measure determination method based on carbon emission prediction

Abstract

The invention relates to the technical field of energy conservation and emission reduction, and discloses an energy conservation and emission reduction measure determination method based on carbon emission prediction, which comprises the following steps: dividing an emission data set to obtain a training set, a verification set and a test set, constructing a prediction model by using the training set and the test set to obtain a trained prediction model, and performing prediction based on the trained prediction model to obtain a prediction result. The method breaks through the limitation of a traditional static model, captures the causal relationship of multiple variables along with time evolution through the emission coefficient, and provides support for prediction; meanwhile, prediction and intervention form closed-loop cooperation, medium-and-long-term emission trend prediction is achieved based on a prediction model of emission coefficient weighting, intervention measures are designed in combination with a generation model embedded with energy conservation constraints, and it is ensured that the measure combination strictly follows the physical law. And the technology maturity, the cost benefit, the social acceptance and the policy compliance are evaluated through the multi-dimensional scoring matrix, and the whole-process scientific guidance from prediction to measure landing is formed.

Description

Technical Field

[0001] This invention relates to the field of energy conservation and emission reduction technology, and more specifically, to a method for determining energy conservation and emission reduction measures based on carbon emission prediction. Background Technology

[0002] Energy conservation and emission reduction refer to reducing energy consumption and harmful emissions through effective technologies and management measures to protect the environment, address climate change, and promote sustainable development. This concept encompasses improving energy efficiency and controlling pollutant emissions across multiple sectors, including industry, construction, and transportation. It includes promoting high-efficiency energy equipment, optimizing production processes, using renewable energy, and strengthening policy guidance. Energy conservation and emission reduction not only help businesses reduce operating costs but also improve air quality, protect the ecological environment, and drive the economy towards a low-carbon and green transformation, making it a crucial strategy for achieving sustainable development.

[0003] However, existing technologies still have several obvious drawbacks. First, the limitations of traditional static models make it difficult to effectively capture the evolution of dynamic causal relationships, resulting in a decrease in prediction accuracy when dealing with new environments and challenges. Second, existing prediction and intervention measures are often disconnected, lacking an effective closed-loop coordination mechanism. This means that after carbon emission prediction, relevant intervention measures may not follow up in a timely or effective manner, leading to a gap between policy implementation and actual conditions, and preventing the full realization of the effects of energy conservation and emission reduction measures.

[0004] Furthermore, traditional assessment methods often focus on a single dimension, such as only considering the amount of carbon emissions or economic costs, while neglecting the comprehensive impact of multiple factors. This one-sided assessment approach may lead to an underestimation or overestimation of the actual effects of the measures, thereby affecting the scientific nature and effectiveness of decision-making. Therefore, these shortcomings, to some extent, limit the rationality and enforceability of energy conservation and emission reduction measures, and there is an urgent need for more dynamic and comprehensive assessment methods to provide effective decision support. Summary of the Invention

[0005] To overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide a method for determining energy-saving and emission-reduction measures based on carbon emission prediction.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for determining energy conservation and emission reduction measures based on carbon emission prediction, the method comprising: S101: Acquire energy consumption data, industrial output data, and meteorological data; preprocess the energy consumption data and industrial output data to obtain energy consumption and industrial output respectively; standardize the energy consumption data, industrial output data, and meteorological data to obtain the emissions dataset. S102: Divide the emission dataset into training set, validation set and test set. Use the training set and test set to build a prediction model to obtain a trained prediction model. Make predictions based on the trained prediction model to obtain the prediction results. S103: Generate an initial combination of measures based on the prediction results, construct a multi-dimensional scoring function, score the initial combination of measures, obtain a scoring matrix and decision recommendations, and adjust the initial combination of measures based on the decision recommendations; S104: Verify the adjusted initial combination of measures and feed the verification results back to the prediction model.

[0007] Energy consumption data can be obtained from the "China Industrial Park Energy Statistics Yearbook" Volumes 2019-2024, while industrial output data can be obtained from quarterly GDP and industry-specific output. The energy consumption data is preprocessed as follows: According to the "General Rules for Calculating Comprehensive Energy Consumption," energy conversion factors are used: Among them, the energy conversion factors are: electricity = 0.1229 kgce / kWh (kg standard coal / kWh), coal = 0.7143 kgce / kg, and natural gas = 1.3300 kgce / m³.

[0008] The industrial output data is preprocessed as follows: The Producer Price Index (PPI) is used for deflating, using the following formula: In the formula: These represent the ex-factory price index at different times.

[0009] Furthermore, obtaining the training set, validation set, and test set includes: The emissions dataset was divided into an initial training set, a validation set, and a test set in chronological order. For the training set: a sliding window is constructed to divide the training set, sliding in a quarterly increment to form the training set; For the validation set and the test set, the most recent historical data is obtained and populated respectively. Then, a sliding window is constructed to divide the populated validation set and the test set to obtain the validation set and the test set.

[0010] Furthermore, the prediction model constructed using the training and test sets includes: a prediction layer and a response layer; The specific steps for training the prediction layer are as follows: Input the training set, optimize the network weights through backpropagation, and use the loss function as the negative log-likelihood loss plus the KL divergence term, as shown in the formula: In the formula: - For variational posterior distribution The log-likelihood expectation of the observed data y. Describe the variational posterior distribution With prior distribution The dispersion, where L is the loss value of the current batch. y represents the network weights, y represents carbon emissions, and X represents the input features; Network weights The update is performed using the following formula: In the formula: The learning rate; Validation is performed using a validation set, and the layer with the smallest validation set loss is selected as the prediction layer.

[0011] Furthermore, the prediction based on the trained prediction model includes: The gradient method is used to calculate the emission coefficients, resulting in an emission coefficient matrix. The energy consumption and emission coefficient matrix are then input into the trained prediction model. The predicted emissions are obtained using the prediction formula: In the formula: Let i be the emission coefficient of variable i at time t. This represents the value of the input variable at time t. This is the random error term.

[0012] As a concrete example, the specific process of calculating the emission factor using the gradient method is as follows: Taking the partial derivative of the output variable, i.e., carbon emissions, with respect to the input variable, and combining it with the time-varying decay term, the formula is: In the formula: This represents the instantaneous gradient of the output variable with respect to the input variable. This is the attenuation coefficient, with a value of 0.03. Indicates the lag time.

[0013] Furthermore, the prediction layer includes an input layer, a hidden layer, and a first output layer; The hidden layers include a first hidden layer and a second hidden layer, both containing activation functions. Each layer is followed by a BatchNorm layer and a Dropout layer. Gradient optimization uses the AdamW optimizer with a learning rate of 0.001 and weight decay of 0.01. The specific calculations are as follows: First hidden layer: ,in The weight matrix is ​​a 128-dimensional vector with 20 feature variables. It is the bias vector; Second hidden layer: ,in The weight matrix is ​​a 128-dimensional vector with 20 feature variables. For the same bias vector; It's worth noting that the GroupNorm layer is a group normalization technique used for training deep neural networks. Its main purpose is to reduce internal covariate bias during training, thereby improving training stability and efficiency. In GroupNorm, feature channels are divided into several groups, and then normalization is performed on each group. It is suitable for mini-batch training because it calculates the standard deviation and mean of each group, rather than normalizing the entire batch, resulting in better training performance.

[0014] Dropout is a regularization technique designed to prevent overfitting in neural networks. During training, it randomly discards a portion of the neurons in the network, temporarily setting their outputs to zero to reduce dependence on specific neurons. This forces the network to learn more robust features and improves the model's generalization ability on unseen data.

[0015] The first output layer includes a carbon emission prediction branch and an emission coefficient matrix branch.

[0016] Furthermore, the measure layer includes an encoder, a decoder, and a second output layer; The encoder includes three fully connected layers, and the decoder includes an embedded energy conservation constraint layer and three fully connected layers.

[0017] Furthermore, the construction of the multi-dimensional scoring function includes: In the formula: Indicates the success rate of similar projects, Indicates the investment payback period and net present value assessment, Indicating social satisfaction For compliance scoring, a, b, c, and d represent the corresponding weights.

[0018] Furthermore, the verification of the adjusted initial combination of measures includes: The impact of the adjusted initial combination of measures is calculated using the finite difference method, and the formula is as follows: In the formula: The predicted emissions before the measures are implemented. The emission coefficient corresponding to the measures, Indicates the magnitude of the adjustment after the measures have been adjusted; Then, the energy conversion efficiency is verified by the first law of thermodynamics to check the physical consistency of the measures parameters.

[0019] An electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the above-described method for determining energy-saving and emission-reduction measures based on carbon emission prediction.

[0020] A computer-readable storage medium storing a computer program, which, when executed, implements the aforementioned method for determining energy-saving and emission-reduction measures based on carbon emission prediction.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: This application discloses a method for determining energy conservation and emission reduction measures based on carbon emission prediction, including: dividing the emission dataset to obtain a training set, a validation set, and a test set; using the training set and the test set to construct a prediction model to obtain a trained prediction model; and making predictions based on the trained prediction model to obtain prediction results. This invention breaks through the limitations of traditional static models by capturing the causal relationships among multiple variables over time through emission coefficients, providing support for prediction. Simultaneously, prediction and intervention form a closed-loop synergy. A prediction model based on emission coefficient weighting achieves medium- and long-term emission trend prediction, while a generative model embedded with energy conservation constraints designs intervention measures, ensuring that the combination of measures strictly follows physical laws. Furthermore, a multi-dimensional scoring matrix assesses technological maturity, cost-effectiveness, social acceptance, and policy compliance, forming a scientific guide for the entire process from prediction to measure implementation. Finally, a closed-loop verification and dynamic optimization mechanism ensures continuous evolution of the solution. Counterfactual reasoning verifies the actual effects of the measures, physical consistency verification ensures that energy conversion efficiency conforms to thermodynamic laws, and error backpropagation dynamically updates model parameters, forming a continuous iterative closed loop. This provides adaptive adjustment capabilities in the face of uncertainty, ultimately achieving full controllability in emission reduction target setting, measure generation, and effect verification, comprehensively enhancing the innovation of energy conservation and emission reduction measures. Attached Figure Description

[0022] Figure 1 A flowchart illustrating a method for determining energy conservation and emission reduction measures based on carbon emission prediction, provided by the present invention; Figure 2 A schematic diagram of the structure of an electronic device provided by the present invention; Figure 3 A schematic diagram of the structure of a computer-readable storage medium provided by the present invention; Figure 4 This invention provides an architecture diagram of a method for determining energy conservation and emission reduction measures based on carbon emission prediction. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1 Please see Figure 1 and Figure 4 As shown in the figure, this embodiment discloses a method for determining energy conservation and emission reduction measures based on carbon emission prediction. The method includes: S101: Acquire energy consumption data, industrial output data, and meteorological data; preprocess the energy consumption data and industrial output data to obtain energy consumption and industrial output respectively; standardize the energy consumption data, industrial output data, and meteorological data to obtain the emissions dataset. In this embodiment, energy consumption data can be obtained from the "China Industrial Park Energy Statistics Yearbook" Volumes 2019-2024, and industrial output data can be obtained from quarterly GDP and industry-specific output. The energy consumption data is preprocessed as follows: According to the "General Rules for Calculating Comprehensive Energy Consumption," energy conversion factors are used: Among them, the energy conversion factors are: electricity = 0.1229 kgce / kWh (kg standard coal / kWh), coal = 0.7143 kgce / kg, and natural gas = 1.3300 kgce / m³.

[0025] The industrial output data is preprocessed as follows: The Producer Price Index (PPI) is used for deflating, using the following formula: In the formula: These represent the ex-factory price index at different times.

[0026] S102: Divide the emission dataset into training set, validation set and test set. Use the training set and test set to build a prediction model to obtain a trained prediction model. Make predictions based on the trained prediction model to obtain the prediction results. As a specific example, obtaining the training set, validation set, and test set includes: The emission dataset is divided into an initial training set, a validation set, and a test set in chronological order. In this embodiment, the initial training set, validation set, and test set are divided in a ratio of 7:2:1. For the training set: Construct a sliding window to divide the training set, sliding it in a quarterly increment to form a training set. This step allows each window to be used as an independent sample input value in the prediction model, preserving the continuous characteristics of the time series. For the validation set and the test set, the most recent historical data is obtained and populated respectively. Then, a sliding window is constructed to divide the populated validation set and the test set to obtain the validation set and the test set.

[0027] The training set is used for model training, the validation set is used for hyperparameter tuning, and the test set is used for final evaluation.

[0028] The prediction model constructed using the training set and the test set includes: a prediction layer and a response layer; The specific steps for training the prediction layer are as follows: Input the training set, optimize the network weights through backpropagation, and use the loss function as the negative log-likelihood loss plus the KL divergence term, as shown in the formula: In the formula: - For variational posterior distribution The log-likelihood expectation of the observed data y. Describe the variational posterior distribution With prior distribution The dispersion, where L is the loss value of the current batch. y represents the network weights, y represents carbon emissions, and X represents the input features; In this embodiment, the training process uses the Adam optimizer with a learning rate of 0.001 and 1000 iterations. Network weights The update is performed using the following formula: In the formula: The learning rate; Validation is performed using a validation set, and the layer with the smallest validation set loss is selected as the prediction layer.

[0029] For example, the prediction based on the trained prediction model includes: The gradient method is used to calculate the emission coefficients, resulting in an emission coefficient matrix. The energy consumption and emission coefficient matrix are then input into the trained prediction model. The predicted emissions are obtained using the prediction formula: In the formula: Let i be the emission coefficient of variable i at time t. This represents the value of the input variable at time t. This is the random error term.

[0030] As a concrete example, the specific process of calculating the emission factor using the gradient method is as follows: Taking the partial derivative of the output variable, i.e., carbon emissions, with respect to the input variable, and combining it with the time-varying decay term, the formula is: In the formula: This represents the instantaneous gradient of the output variable with respect to the input variable. This is the attenuation coefficient, with a value of 0.03. Indicates the lag time.

[0031] The prediction layer includes an input layer, a hidden layer, and a first output layer; The hidden layers include a first hidden layer and a second hidden layer, both containing activation functions. Each layer is followed by a BatchNorm layer and a Dropout layer. Gradient optimization uses the AdamW optimizer with a learning rate of 0.001 and weight decay of 0.01. The specific calculations are as follows: First hidden layer: ,in The weight matrix is ​​a 128-dimensional vector with 20 feature variables. It is the bias vector; Second hidden layer: ,in The weight matrix is ​​a 128-dimensional vector with 20 feature variables. For the same bias vector; It's worth noting that the GroupNorm layer is a group normalization technique used for training deep neural networks. Its main purpose is to reduce internal covariate bias during training, thereby improving training stability and efficiency. In GroupNorm, feature channels are divided into several groups, and then normalization is performed on each group. It is suitable for mini-batch training because it calculates the standard deviation and mean of each group, rather than normalizing the entire batch, resulting in better training performance.

[0032] Dropout is a regularization technique designed to prevent overfitting in neural networks. During training, it randomly discards a portion of the neurons in the network, temporarily setting their outputs to zero to reduce dependence on specific neurons. This forces the network to learn more robust features and improves the model's generalization ability on unseen data.

[0033] The first output layer includes a carbon emission prediction branch and an emission coefficient matrix branch. The carbon emission prediction branch is a fully connected layer that maps 128-dimensional hidden states to 1-dimensional prediction results, with a linear activation function. Causality coefficient matrix branch: The causal strength between variables is calculated using a bilinear pooling layer, as shown in the following formula: In the formula: This is the emission coefficient weight matrix.

[0034] It should be noted that the first output layer is used to receive the emissions dataset; The measure layer includes an encoder, a decoder, and a second output layer; The encoder includes three fully connected layers, and the decoder includes an embedded energy conservation constraint layer and three fully connected layers.

[0035] It should be noted that the encoder consists of three fully connected layers with LeakyReLU activation function and a negative slope of 0.01. The input is a concatenated vector of the difference between the current state matrix and the target. Each layer is followed by a GroupNorm layer and a Dropou layer.

[0036] The constraint formulas included in the energy conservation constraint layer are as follows: In the formula: This indicates the change in carbon emissions. For energy input power, To ensure effective output power, a penalty term is added to the loss function when constraints are violated. ,and The value is 5.

[0037] Furthermore, in this embodiment, after obtaining the prediction results, the prediction results can be verified and rolled over for further prediction: The logic of rolling forecasting is to start with a certain quarter of a past year and predict subsequent carbon emissions sequentially. For each quarter, the forecast results are used as input features for the next quarter. Confidence intervals are calculated based on these forecast results, using the following formula: In the formula: This is the standard deviation of the historical prediction error, which can be calculated using data from the validation set. 1.96 is the critical value for the standard normal distribution at the 95% confidence level. S103: Generate an initial combination of measures based on the prediction results, construct a multi-dimensional scoring function, score the initial combination of measures, obtain a scoring matrix and decision recommendations, and adjust the initial combination of measures based on the decision recommendations; The construction of the multi-dimensional scoring function includes: In the formula: Indicates the success rate of similar projects, Indicates the investment payback period and net present value assessment, Indicating social satisfaction For compliance scoring, a, b, c, and d represent the corresponding weights.

[0038] In this embodiment, a, b, c, and d are respectively set to 0.35, 0.25, 0.2, and 0.2.

[0039] It should be noted that the generation logic for generating the initial combination of measures based on the prediction results is as follows: The predictive model searches the space of feasible measures under constraints based on the difference between the current state and the target. For example, when the target emission reduction is 200,000 tons, the predictive model adjusts the photovoltaic coverage and waste heat recovery efficiency to meet the energy conservation constraint while ensuring that the total emission reduction reaches 200,000 tons.

[0040] S104: Verify the adjusted initial combination of measures and feed the verification results back to the prediction model.

[0041] The verification of the adjusted initial measure combination includes: The impact of the adjusted initial combination of measures is calculated using the finite difference method, and the formula is as follows: In the formula: The predicted emissions before the measures are implemented. The emission coefficient corresponding to the measures, Indicates the magnitude of the adjustment after the measures have been adjusted; Then, the energy conversion efficiency is verified by the first law of thermodynamics to check the physical consistency of the measures parameters.

[0042] For example, the specific steps for feeding the verification results back into the prediction model are as follows: The network weights are adjusted through error backpropagation, and the calculation formula is as follows: In the formula: Loss is the mean square error between the predicted value and the actual value. The learning rate is 0.01.

[0043] Example 2 Please see Figure 2 As shown, this embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the energy-saving and emission-reduction measure determination method based on carbon emission prediction provided by the above methods.

[0044] Since the electronic device described in this embodiment is the electronic device used to implement the method for determining energy-saving and emission-reduction measures based on carbon emission prediction in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method for determining energy-saving and emission-reduction measures based on carbon emission prediction described in the embodiments of this application. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. Any electronic device used by those skilled in the art to implement the method for determining energy-saving and emission-reduction measures based on carbon emission prediction in the embodiments of this application falls within the scope of protection of this application.

[0045] Example 3 Please see Figure 3 As shown, this embodiment discloses a computer-readable storage medium, including a memory, a processor, and a computer program stored on the memory and running on the processor. When the processor executes the computer program, it implements the method for determining energy-saving and emission-reduction measures based on carbon emission prediction provided by the above methods.

[0046] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters, weights, and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0047] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless network. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0048] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0049] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0050] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0051] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0052] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0053] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0054] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for determining energy conservation and emission reduction measures based on carbon emission prediction, characterized in that, The method includes: S101: Acquire energy consumption data, industrial output data, and meteorological data; preprocess the energy consumption data and industrial output data to obtain energy consumption and industrial output respectively; standardize the energy consumption data, industrial output data, and meteorological data to obtain the emissions dataset. S102: Divide the emission dataset into training set, validation set and test set. Use the training set and test set to build a prediction model to obtain a trained prediction model. Make predictions based on the trained prediction model to obtain the prediction results. S103: Generate an initial combination of measures based on the prediction results, construct a multi-dimensional scoring function, score the initial combination of measures, obtain a scoring matrix and decision recommendations, and adjust the initial combination of measures based on the decision recommendations; S104: Verify the adjusted initial combination of measures and feed the verification results back to the prediction model.

2. The method for determining energy conservation and emission reduction measures based on carbon emission prediction according to claim 1, characterized in that, The process of obtaining the training set, validation set, and test set includes: The emissions dataset was divided into an initial training set, a validation set, and a test set in chronological order. For the training set: a sliding window is constructed to divide the training set, sliding in a quarterly increment to form the training set; For the validation set and the test set, the most recent historical data is obtained and populated respectively. Then, a sliding window is constructed to divide the populated validation set and the test set to obtain the validation set and the test set.

3. The method for determining energy conservation and emission reduction measures based on carbon emission prediction according to claim 1, characterized in that, The prediction model constructed using training and testing sets includes: a prediction layer and a response layer; The specific steps for training the prediction layer are as follows: Input the training set, optimize the network weights through backpropagation, and use the loss function as the negative log-likelihood loss plus the KL divergence term, as shown in the formula: In the formula: - For variational posterior distribution The log-likelihood expectation of the observed data y. Describe the variational posterior distribution With prior distribution The dispersion, where L is the loss value of the current batch. y represents the network weights, y represents carbon emissions, and X represents the input features; Network weights The update is performed using the following formula: In the formula: The learning rate; Validation is performed using a validation set, and the layer with the smallest validation set loss is selected as the prediction layer.

4. The method for determining energy conservation and emission reduction measures based on carbon emission prediction according to claim 3, characterized in that, The prediction based on the trained prediction model includes: The gradient method is used to calculate the emission coefficients, resulting in an emission coefficient matrix. The energy consumption and emission coefficient matrix are then input into the trained prediction model. The predicted emissions are obtained using the prediction formula: In the formula: Let i be the emission coefficient of variable i at time t. This represents the value of the input variable at time t. This is the random error term.

5. The method for determining energy conservation and emission reduction measures based on carbon emission prediction according to claim 3, characterized in that: The prediction layer includes an input layer, a hidden layer, and a first output layer; The hidden layer includes a first hidden layer and a second hidden layer, both of which contain an activation function; The first output layer includes a carbon emission prediction branch and an emission coefficient matrix branch.

6. The method for determining energy conservation and emission reduction measures based on carbon emission prediction according to claim 3, characterized in that, The measure layer includes an encoder, a decoder, and a second output layer; The encoder includes three fully connected layers, and the decoder includes an embedded energy conservation constraint layer and three fully connected layers.

7. The method for determining energy conservation and emission reduction measures based on carbon emission prediction according to claim 1, characterized in that, The construction of the multi-dimensional scoring function includes: In the formula: Indicates the success rate of similar projects, Indicates the investment payback period and net present value assessment, Indicating social satisfaction For compliance scoring, a, b, c, and d represent the corresponding weights.

8. The method for determining energy conservation and emission reduction measures based on carbon emission prediction according to claim 1, characterized in that, The verification of the adjusted initial measure combination includes: The impact of the adjusted initial combination of measures is calculated using the finite difference method, and the formula is as follows: In the formula: The predicted emissions before the measures are implemented. This indicates the emission factor corresponding to the measure. Indicates the magnitude of the adjustment after the measures have been adjusted; Then, the energy conversion efficiency is verified by the first law of thermodynamics to check the physical consistency of the measures parameters.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the method for determining energy-saving and emission-reduction measures based on carbon emission prediction as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the method for determining energy-saving and emission-reduction measures based on carbon emission prediction as described in any one of claims 1 to 8.

Images