Carbon emission prediction method and system based on big data analysis

Abstract

The invention discloses a carbon emission prediction method and system based on big data analysis, and relates to the technical field of data prediction. The method comprises the steps of firstly determining a carbon emission data vector of each carbon source based on carbon emission data of a target object in past preset days, then determining a carbon emission feature matrix by using an improved time convolution network, and then processing the carbon emission feature matrix by using a pre-trained attention mechanism network to obtain a carbon emission feature matrix; and finally, inputting the attention weighted feature matrix into a pre-trained long-short-term memory network to determine carbon emission prediction data of each carbon source in a preset number of days in the future. Therefore, the importance difference and the fluctuation characteristic difference of different carbon sources are fully considered, the carbon emission prediction precision is improved, and the technical problems that the carbon emission prediction method in the prior art does not fully consider the importance difference and the fluctuation characteristic difference of different carbon sources, and the prediction precision is limited due to simple data merging are solved.

Description

Technical Field

[0001] This application relates to the technical field of data prediction, and particularly to a carbon emission prediction method and system based on big data analysis. Background Art

[0002] With the increasingly severe global climate change problem, reducing greenhouse gas emissions has become the focus of global attention. As the main component of greenhouse gases, accurate prediction of carbon emissions is of great significance for formulating effective emission reduction policies and measures. In recent years, the rapid development of big data analysis and artificial intelligence technologies has provided new methods and tools for carbon emission prediction. Carbon emission prediction is the basis for evaluating and optimizing emission reduction strategies. By accurately predicting future carbon emission trends, enterprises and research institutions can better formulate emission reduction targets, optimize resource allocation, and evaluate the effectiveness of policies. For example, enterprises can optimize production processes and reduce their carbon footprint by predicting carbon emissions.

[0003] Currently, carbon emission prediction methods include correlation analysis algorithms and model estimation algorithms. The correlation analysis algorithm predicts the carbon emissions of the emission entity by analyzing the relationship between carbon emissions and strongly correlated factors (such as energy consumption, industrial production index, etc.); the model estimation algorithm predicts carbon emissions by constructing a mathematical model (such as a regression model). For example, the method disclosed in the patent with publication number CN113240155A and name "Method, Device and Terminal for Predicting Carbon Emissions" includes: obtaining historical carbon emission data of multiple emission entities, where the historical carbon emission data includes data on factors affecting carbon emissions and carbon emission result data for each emission entity; performing a correlation analysis on the data on factors affecting carbon emissions and the carbon emission result data to select at least two strongly correlated factors affecting the carbon emission result from the data on factors affecting carbon emissions; and predicting the carbon emissions of the emission entity based on the at least two strongly correlated factors.

[0004] Another example is the method disclosed in the patent with publication number CN113657661A and name "Enterprise Carbon Emission Prediction Method, Device, Computer Equipment and Storage Medium", which includes: obtaining data to be analyzed, where the data to be analyzed includes at least one of per capita gross domestic product, energy intensity, added value of the industry to which the target enterprise belongs, total energy consumption of the industry to which the target enterprise belongs, main business income of the target enterprise, operating profit of the target enterprise, and electricity consumption of the target enterprise; performing empirical mode decomposition on the data to be analyzed to obtain multiple subsequences with different frequencies and a residual component sequence; inputting each of the multiple sequences composed of the subsequences and the residual component sequence into a carbon emission prediction model to obtain sequence prediction results corresponding to each sequence respectively; performing sequence reconstruction on the obtained sequence prediction results to obtain a reconstruction result, and determining the enterprise carbon emission prediction result corresponding to the target enterprise based on the reconstruction result.

[0005] However, existing carbon emission prediction methods usually process the data of all carbon sources as a whole, ignoring the differences and interactions between different carbon sources, which seriously affects the accuracy of carbon emission prediction results. For example, the influencing factors of carbon sources such as industrial fuel combustion, electricity consumption, and transportation emissions are significantly different. If simply combined into a single sequence for modeling, it will lead to the loss of key features and seriously affect the accuracy of the prediction results. In addition, the carbon emissions of different carbon sources may have different dynamic fluctuation characteristics (such as high-frequency mutations and low-frequency periodicity), while existing methods fail to fully consider these fluctuation differences.

[0006] Regarding the problem that existing carbon emission prediction methods do not fully consider the importance differences and fluctuation characteristic differences of different carbon sources, and the simple combination of data results in limited prediction accuracy, no effective solution has been proposed yet. Summary of the Invention

[0007] Embodiments of the present disclosure provide a carbon emission prediction method and system based on big data analysis, which can at least solve the technical problem that existing carbon emission prediction methods do not fully consider the importance differences and fluctuation characteristic differences of different carbon sources, and the simple combination of data results in limited prediction accuracy.

[0008] According to one aspect of the embodiments of the present disclosure, a carbon emission prediction method based on big data analysis is provided, including: based on the carbon emission data of a target object in the past preset number of days, determining the carbon emission data vector of each carbon source of the target object in the past preset number of days; wherein, the target object has at least two carbon sources; according to the carbon emission data vectors of all carbon sources in the past preset number of days, using an improved temporal convolutional network to determine the carbon emission feature matrix of the target object; wherein, the improved temporal convolutional network assigns independent convolutional kernels and dilation coefficients to each carbon source, and the dilation coefficients are determined in real time based on the carbon emission fluctuation conditions of the carbon sources; the dimension of the carbon emission feature matrix is n×m, where n is the number of carbon sources and m is the past preset number of days, and each element in the carbon emission feature matrix represents the carbon emission feature value of the corresponding carbon source on the jth day in the past; and using a pre-trained attention mechanism network to process the carbon emission feature matrix to obtain the attention-weighted feature matrix of the target object, and inputting the attention-weighted feature matrix into a pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source of the target object in the future preset number of days.

[0009] According to another aspect of the embodiments of the present disclosure, a storage medium is further provided, and the storage medium includes a stored program, wherein the above-mentioned method is executed by a processor when the program runs.

[0010] According to another aspect of the embodiments of the present disclosure, there is also provided a carbon emission prediction system based on big data analysis, including: a first determination module, configured to determine the carbon emission data vectors of each carbon source of the target object in the past preset number of days based on the carbon emission data of the target object in the past preset number of days; wherein, the target object has at least two carbon sources; a second determination module, configured to determine the carbon emission feature matrix of the target object by using an improved temporal convolutional network according to the carbon emission data vectors of all carbon sources in the past preset number of days; wherein, the improved temporal convolutional network assigns independent convolutional kernels and dilation coefficients to each carbon source, and the dilation coefficient is determined in real time based on the carbon emission fluctuation condition of the carbon source; the dimension of the carbon emission feature matrix is n×m, where n is the number of carbon sources and m is the past preset number of days, and each element in the carbon emission feature matrix represents the carbon emission feature value of the corresponding carbon source on the j-th day in the past; and a third determination module, configured to process the carbon emission feature matrix by using a pre-trained attention mechanism network to obtain the attention-weighted feature matrix of the target object, and input the attention-weighted feature matrix into a pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source of the target object in the future preset number of days.

[0011] According to another aspect of the embodiments of the present disclosure, there is also provided a carbon emission prediction system based on big data analysis, including a processor; and a memory, connected to the processor, configured to provide instructions for the processor to perform the following processing steps: determining the carbon emission data vectors of each carbon source of the target object in the past preset number of days based on the carbon emission data of the target object in the past preset number of days; wherein, the target object has at least two carbon sources; determining the carbon emission feature matrix of the target object by using an improved temporal convolutional network according to the carbon emission data vectors of all carbon sources in the past preset number of days; wherein, the improved temporal convolutional network assigns independent convolutional kernels and dilation coefficients to each carbon source, and the dilation coefficient is determined in real time based on the carbon emission fluctuation condition of the carbon source; the dimension of the carbon emission feature matrix is n×m, where n is the number of carbon sources and m is the past preset number of days, and each element in the carbon emission feature matrix represents the carbon emission feature value of the corresponding carbon source on the j-th day in the past; and processing the carbon emission feature matrix by using a pre-trained attention mechanism network to obtain the attention-weighted feature matrix of the target object, and inputting the attention-weighted feature matrix into a pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source of the target object in the future preset number of days.

[0012] This application first determines the carbon emission data vector of each carbon source based on the carbon emission data of the target object in the past preset number of days, so as to fully consider the differences and interactions between different carbon sources, effectively avoiding the problem of loss of key features caused by general processing of all carbon source data. Then, an improved temporal convolutional network is used to determine the carbon emission feature matrix. This network assigns independent convolutional kernels and dilation coefficients to each carbon source, enabling more accurate capture of the carbon emission characteristics of each carbon source, and the dilation coefficient is determined in real time based on the carbon emission fluctuation of the carbon source, further improving the prediction accuracy. Secondly, the pre-trained attention mechanism network is used to process the carbon emission feature matrix to obtain the attention-weighted feature matrix. The attention mechanism can automatically learn and emphasize the information that has an important impact on the prediction result, further improving the prediction accuracy. Finally, the attention-weighted feature matrix is input into the pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source in the future preset number of days. The long short-term memory network has a powerful time series processing ability, can capture the long-term dependence relationship in the carbon emission data, and achieve accurate prediction of future carbon emissions. Thus, through the combination of refined data prediction, the improved temporal convolutional network, the attention mechanism network and the long short-term memory network, this application fully considers the importance differences and fluctuation feature differences of different carbon sources, significantly improving the accuracy of carbon emission prediction. Furthermore, it solves the technical problem in the prior art that the carbon emission prediction method does not fully consider the importance differences and fluctuation feature differences of different carbon sources, and the simple combination of data results in limited prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are used to provide a further understanding of the present disclosure, form a part of this application, and the illustrative embodiments and descriptions thereof of the present disclosure are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure. In the drawings:

[0014] Figure 1 is a hardware structure block diagram of a computing device for implementing the method according to Embodiment 1 of this application;

[0015] Figure 2 is a schematic flowchart of the carbon emission prediction method based on big data analysis according to Embodiment 1 of this application;

[0016] Figure 3 is a schematic diagram of the carbon emission prediction system based on big data analysis according to Embodiment 2 of this application; and

[0017] Figure 4 is a schematic diagram of the carbon emission prediction system based on big data analysis according to Embodiment 3 of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] To enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present disclosure.

[0019] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order different from those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0020] Embodiment 1

[0021] According to this embodiment, a method embodiment of a carbon emission prediction method based on big data analysis is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0022] The method embodiment provided in this embodiment can be executed on a mobile terminal, a computer terminal, a server or a similar computing device. Figure 1 A hardware structure block diagram of a computing device for implementing a carbon emission prediction method based on big data analysis is shown. As Figure 1 shown, the computing device may include one or more processors (the processor may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory for storing data, a transmission device for communication functions, and an input / output interface. Among them, the memory, the transmission device, and the input / output interface are connected to the processor through a bus. In addition, it may further include: a display, a keyboard, and a cursor control device connected to the input / output interface. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the computing device may further include more Figure 1more or fewer components shown therein, or having a configuration different from that shown in Figure 1 that shown.

[0023] It should be noted that one or more of the above-mentioned processors and / or other data prediction circuits can generally be referred to as "data prediction circuits" herein. The data prediction circuit can be embodied in whole or in part as software, hardware, firmware, or any combination thereof. In addition, the data prediction circuit can be a single independent processing module, or be incorporated in whole or in part into any one of other elements in a computing device. As involved in the embodiments of the present disclosure, the data prediction circuit is a kind of processor control (for example, the selection of a variable resistance terminal path connected to an interface).

[0024] The memory can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the carbon emission prediction method based on big data analysis in the embodiments of the present disclosure. The processor executes various functional applications and data predictions by running the software programs and modules stored in the memory, that is, implements the carbon emission prediction method based on big data analysis of the above-mentioned application program. The memory can include high-speed random access memory, and can also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory can further include a memory remotely provided with respect to the processor, and these remote memories can be connected to the computing device through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0025] The transmission device is used to receive or send data via a network. Specific examples of the above-mentioned network can include a wireless network provided by a communication provider of a computing device. In one instance, the transmission device includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one instance, the transmission device can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0026] The display can be, for example, a touch-screen liquid crystal display (LCD), and the liquid crystal display enables a user to interact with the user interface of the computing device.

[0027] It should be noted here that in some alternative embodiments, the above-mentioned Figure 1 computing device shown can include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware elements and software elements. It should be pointed out that Figure 1This is just an example of a specific concrete instance and is intended to illustrate the types of components that may exist in the above computing device.

[0028] Under the above operating environment, according to the first aspect of this embodiment, a carbon emission prediction method based on big data analysis is provided. Figure 2 The flowchart of this method is shown. Refer to Figure 2 As shown, this method includes:

[0029] S102: Based on the carbon emission data of the target object in the past preset number of days, determine the carbon emission data vector of each carbon source of the target object in the past preset number of days; wherein, the target object has at least two carbon sources;

[0030] S104: According to the carbon emission data vectors of all carbon sources in the past preset number of days, use the improved temporal convolutional network to determine the carbon emission feature matrix of the target object; wherein, the improved temporal convolutional network assigns independent convolutional kernels and dilation coefficients to each carbon source, and the dilation coefficient is determined in real time based on the carbon emission fluctuation of the carbon source; the dimension of the carbon emission feature matrix is n×m, where n is the number of carbon sources and m is the past preset number of days, and each element in the carbon emission feature matrix represents the carbon emission feature value of the corresponding carbon source on the jth day in the past; and

[0031] S106: Use the pre-trained attention mechanism network to process the carbon emission feature matrix to obtain the attention-weighted feature matrix of the target object, and input the attention-weighted feature matrix into the pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source of the target object in the future preset number of days.

[0032] In this embodiment, through research, it is found that carbon emission data has characteristics such as multi-source heterogeneity (such as industrial fuel combustion, electricity consumption, traffic emissions, etc.), strong time dependence (such as seasonal fluctuations, production cycle impacts), and significant dynamic fluctuations (such as emission mutations caused by emergencies). For example, an enterprise's carbon emissions may simultaneously include direct fuel combustion and indirect electricity consumption, and the fluctuation patterns of different carbon sources (such as the suddenness of production process emissions and the periodicity of electricity consumption) vary significantly. If multi-carbon source data is simply combined into a single sequence for modeling, key features will be lost. In addition, when existing deep learning models (such as standard TCN, LSTM) process multi-carbon source data, they usually adopt unified network parameters (such as convolution kernel size, dilation coefficient), making it difficult to dynamically adapt to the fluctuation characteristics of different carbon sources. For example, carbon emissions from industrial equipment may exhibit high-frequency mutations, while carbon emissions from building energy consumption have low-frequency periodicity. If a TCN with a fixed dilation coefficient is used, it is impossible to effectively capture the time series features of different frequencies. Therefore, based on the characteristics of multi-source heterogeneity, strong time dependence, and significant dynamic fluctuations of carbon emission data, this application designs independent feature extraction modules for each carbon source to dynamically adapt to its fluctuation characteristics, and adjusts the convolution kernel parameters in real time based on carbon source fluctuations to enhance the separation ability of high-frequency and low-frequency features.

[0033] Specifically, in order to fully consider the differences and interactions between different carbon sources, first, based on the carbon emission data of the target object in the past preset number of days, a carbon emission data vector for each carbon source is determined (corresponding to step S102). Here, the target object refers to the entity that needs to monitor and analyze carbon emissions, such as an enterprise, a factory, a city, etc. A carbon source refers to the source that generates carbon emissions, such as burning fossil fuels, industrial processes, transportation, etc. The target object usually has multiple carbon sources. The past preset number of days refers to the historical time range used for analysis, and the selection of this range is usually set as needed based on data availability and the purpose of analysis. For each carbon source, its carbon emission data within the past preset number of days is organized into a carbon emission data vector, providing a basis for subsequent analysis and prediction. In this way, the problem of key feature loss caused by simply processing all carbon source data in a general way is effectively avoided.

[0034] The Temporal Convolutional Network (TCN) is a deep learning model for processing time series data. Although the standard TCN expands the receptive field through dilated convolutions, it does not optimize the convolutional kernel design by combining the dynamic fluctuation characteristics of carbon sources, resulting in poor separation and extraction effects for high-frequency fluctuations and low-frequency trends. Based on this, this embodiment improves the temporal convolutional network to adapt to the dynamic fluctuation characteristics of multi-source carbon emission data. Specifically, according to the carbon emission data vectors of all carbon sources in the past preset number of days, the improved temporal convolutional network is used to determine the carbon emission feature matrix of the target object (corresponding to step S104). Among them, the improved temporal convolutional network assigns independent convolutional kernels to each carbon source to capture the unique carbon emission patterns of each carbon source, and determines the corresponding dilation coefficient in real time according to the carbon emission fluctuations of the carbon source to better capture the short-term and long-term changes in the carbon emission situation of each carbon source. The dimension of the carbon emission feature matrix is n×m, where n is the number of carbon sources and m is the past preset number of days. Each element in the matrix represents the carbon emission feature value of the corresponding carbon source on the jth day in the past. These feature values are obtained by processing the original carbon emission data through the temporal convolutional network and can reflect the complexity and dynamics of carbon emissions. In this way, the carbon emission characteristics of each carbon source can be captured more accurately, providing more useful information for subsequent predictions.

[0035] Furthermore, the pre-trained attention mechanism network is used to process the carbon emission feature matrix to obtain the attention-weighted feature matrix of the target object, and the attention-weighted feature matrix is input into the pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source of the target object in the future preset number of days (corresponding to step S106). Specifically, the pre-trained attention mechanism network is used to process the carbon emission feature matrix to assign different weights to the features of each carbon source to highlight important carbon emission features. After the carbon emission feature matrix is processed by the attention mechanism network, the corresponding attention-weighted feature matrix is obtained, where the weight of each element reflects its importance in the prediction. The pre-trained LSTM network is used to predict the carbon emission data of each carbon source of the target object in the future preset number of days according to the attention-weighted feature matrix. Thus, important carbon emission features can be highlighted through the attention mechanism, and the long-term dependence relationship of time series data can be captured by using the LSTM network to achieve accurate prediction of the carbon emissions of each carbon source of the target object in the future preset number of days.

[0036] As described in the background art, existing carbon emission prediction methods usually process the data of all carbon sources as a whole, ignoring the differences and interactions between different carbon sources, which seriously affects the accuracy of carbon emission prediction results. For example, the influencing factors of carbon sources such as industrial fuel combustion, electricity consumption, and transportation emissions are significantly different. If simply combined into a single sequence for modeling, key features will be lost, seriously affecting the accuracy of prediction results. In addition, the carbon emissions of different carbon sources may have different dynamic fluctuation characteristics (such as high-frequency mutations, low-frequency periodicity), and existing methods fail to fully consider these fluctuation differences.

[0037] In view of this, the technical solution of this application first determines the carbon emission data vector of each carbon source based on the carbon emission data of the target object in the past preset number of days, so as to fully consider the differences and interactions between different carbon sources, and effectively avoid the problem of key feature loss caused by general processing of all carbon source data. Then, an improved temporal convolutional network is used to determine the carbon emission feature matrix. This network assigns independent convolutional kernels and dilation coefficients to each carbon source, which can more accurately capture the carbon emission characteristics of each carbon source, and the dilation coefficients are determined in real time based on the carbon emission fluctuation conditions of the carbon sources, further improving the accuracy of prediction. Secondly, a pre-trained attention mechanism network is used to process the carbon emission feature matrix to obtain an attention-weighted feature matrix. The attention mechanism can automatically learn and emphasize the information that has an important impact on the prediction result, further improving the accuracy of prediction. Finally, the attention-weighted feature matrix is input into a pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source in the future preset number of days. The long short-term memory network has a strong ability to process time series, can capture the long-term dependence relationship in the carbon emission data, and achieve accurate prediction of future carbon emissions. Thus, through the combination of refined data prediction, the improved temporal convolutional network, the attention mechanism network, and the long short-term memory network, this application fully considers the importance differences and fluctuation feature differences of different carbon sources, significantly improving the accuracy of carbon emission prediction. Furthermore, it solves the technical problem in the prior art that the existing carbon emission prediction method does not fully consider the importance differences and fluctuation feature differences of different carbon sources, and the simple combination of data results in limited prediction accuracy.

[0038] Optionally, the determining the carbon emission feature matrix of the target object by using the improved temporal convolutional network according to the carbon emission data vectors of all carbon sources in the past preset number of days includes: determining the carbon emission data matrix of the target object in the past preset number of days according to the carbon emission data vectors of all carbon sources in the past preset number of days; and inputting the carbon emission data matrix into the improved temporal convolutional network to output the carbon emission feature matrix of the target object.

[0039] Specifically, first obtain the carbon emission data C of the target object in the past m days1 ~C m , determine n carbon sources that generate carbon emissions of the target object, and analyze the carbon emission data C 1 ~C m to determine the carbon emission data vector x of each carbon source i in the past m days i =[x i,1 ~x i,m ; where i = 1~n, j = 1~m. Then, according to the carbon emission data vectors of all carbon sources in the past preset number of days, determine the carbon emission data matrix X of the target object in the past preset number of days:

[0040]

[0041] Optionally, the improved temporal convolutional network includes a variance calculation unit, a dilation coefficient determination unit, and a temporal convolutional layer; and the operation of inputting the carbon emission data matrix into the improved temporal convolutional network and outputting the carbon emission feature matrix of the target object includes: inputting the carbon emission data matrix into the variance calculation unit to determine the variance vector of the target object; where each element in the variance vector represents the variance calculation result of the carbon emission data vector of the corresponding carbon source in the past preset number of days; inputting the variance vector into the dilation coefficient determination unit to determine the dilation coefficient corresponding to each carbon source, and transmitting the dilation coefficient to the temporal convolutional layer; and using the temporal convolutional layer to perform dilated convolution on the carbon emission data vectors of all carbon sources in the past preset number of days in parallel according to the corresponding dilation coefficient to determine the carbon emission feature matrix of the target object.

[0042] Specifically, the improved temporal convolutional network includes a variance calculation unit, a dilation coefficient determination unit, and a temporal convolutional layer. Among them, the role of the variance function is to calculate the carbon emission fluctuation of each carbon source in the past m days, and the role of the dilation coefficient determination unit is to dynamically adjust the dilation coefficient of the dilated convolution in the convolutional layer according to the variance value. Therefore, the operation of inputting the carbon emission data matrix into the improved temporal convolutional network and outputting the carbon emission feature matrix of the target object includes:

[0043] (1) For the carbon emission data vector x i =[x i,1 ~x i,m of each carbon source i, calculate its variance, and finally obtain a variance vector V = [Var(x i )~Var(x n )];

[0044] (2) Input the variance vector V into the dilation coefficient determination unit (for example, composed of a mapping function) to generate the dilation coefficient d corresponding to each carbon source ii ;

[0045] (3) For each carbon source i, use its corresponding dilation coefficient d i to perform dilated convolution and output a feature vector y i = [y i,1 ~y i,m . By performing dilated convolution on all carbon sources in parallel, the carbon emission feature matrix Y is finally output:

[0046]

[0047] In this way, the fluctuation characteristics of carbon emission data can be better captured, which is suitable for processing carbon source data with different fluctuation patterns.

[0048] In a specific embodiment, the operation of inputting the carbon emission data matrix into the variance calculation unit to determine the variance vector of the target object includes: using the variance calculation unit to extract the carbon emission data vector of each carbon source i in the past preset number of days from the carbon emission data matrix; using the variance calculation unit to calculate the variance value of the carbon emission data vector of each carbon source in the past preset number of days according to the following formula:

[0049]

[0050] where Var(x i ) is the variance value of the carbon emission data vector x i of carbon source i in the past preset number of days, x i = [x i,1 ~x i,m ; m is the past preset number of days, x i,j is the carbon emission data of carbon source i on the j-th day in the past; μ i is the mean of x i ; and

[0051] Based on the variance values of the carbon emission data vectors of all carbon sources in the past preset number of days, determine the variance vector V of the target object; where V = [Var(x i )~Var(x n ), and n is the number of carbon sources.

[0052] In a specific embodiment, the operation of inputting the variance vector into the dilation coefficient determination unit to determine the dilation coefficient corresponding to each carbon source includes: using the dilation coefficient determination unit to calculate the dilation coefficient corresponding to each carbon source according to the following formula:

[0053] d i = a·Var(x i ) + b;

[0054] where d i is the expansion coefficient corresponding to carbon source i; Var(x i ) is the variance value of the carbon emission data vector x of carbon source i in the past preset number of days i ; a and b are constants.

[0055] In this way,

[0056] In a specific embodiment, the operation of using the temporal convolutional layer to perform dilated convolution on the carbon emission data vectors of all carbon sources in the past preset number of days in parallel according to the corresponding expansion coefficients to determine the carbon emission feature matrix of the target object includes: using the temporal convolutional layer to perform dilated convolution on the carbon emission data vectors of all carbon sources in the past preset number of days in parallel according to the corresponding expansion coefficients through the following formula to obtain the carbon emission feature values of each carbon source in the past preset number of days:

[0057]

[0058] where y i,j is the carbon emission feature value of carbon source i on the j-th day in the past; ω k is the k-th weight of the convolutional kernel; k max is the size of the convolutional kernel; d i is the expansion coefficient corresponding to carbon source i; is the carbon emission data vector x of carbon source i in the past preset number of days i at position j - d i ·k; k is the index of the convolutional kernel, k = 0, 1, ···, k;

[0059] Based on the carbon emission feature values of all carbon sources in the past preset number of days, determine the carbon emission feature matrix Y of the target object;

[0060]

[0061] where y n,m is the carbon emission feature value of the n-th carbon source on the m-th day in the past.

[0062] In another specific embodiment, the implementation method of performing dilated convolution on all carbon sources in parallel is: regarding the data of each carbon source as an independent channel, and then using the characteristics of grouped convolution or depthwise separable convolution to allocate independent convolutional kernels and expansion coefficients for each channel (carbon source), specifically:

[0063] (1) Convert the input data X from n×m to 1×n×m, that is, regard it as a one-dimensional signal with multiple channels;

[0064] (2) Use grouped convolution to group the convolutional kernels, with each group corresponding to a carbon source. The number of groups is set to n, that is, each carbon source uses a single convolutional kernel alone;

[0065] (3) Assign an independent dilation coefficient d to each carbon source i , that is, in the convolutional operation, specify the corresponding dilation coefficient for each group.

[0066] (4) Use grouped convolution for parallel dilated convolution, and then concatenate the outputs of all carbon sources to obtain Y · , whose dimension is 1×n×m;

[0067] (5) Convert the dimension of Y · to n×m to obtain Y.

[0068] In this way, all carbon sources can be processed efficiently in parallel while retaining the dimension of the time series.

[0069] Optionally, the operation of using the pre-trained attention mechanism network to process the carbon emission feature matrix to obtain the attention-weighted feature matrix of the target object includes: using the attention mechanism network to obtain the carbon emission feature vector y of carbon source i from the carbon emission feature matrix i ; where y i = [y i,1 ~ y i,m ], y i,m is the carbon emission feature value of the i-th carbon source on the m-th day in the past, and i is any item from 1 to n; based on the carbon emission feature vector y i , calculate the attention weight α of carbon source i i ; use the attention weight α i to perform weighted summation on the carbon emission feature vector y i to obtain the attention-weighted feature vector y of carbon source i attn(i) ; and based on the attention-weighted feature vectors of all carbon sources in the past preset number of days, determine the attention-weighted feature matrix Y of the target object attn :

[0070]

[0071] Among them, y attn(n,m) is the attention-weighted feature value of carbon source n on the m-th day in the past.

[0072] Specifically, for the carbon emission feature vector y of each carbon source i i , calculate its attention weight α i ; the calculation of the attention weight is usually achieved through the following steps:

[0073] (1) Map the feature vector y i to a hidden representation through a fully connected layer (or linear transformation):

[0074] h i = W h y i + b h ;

[0075] where W h is the weight matrix and b h is the bias term;

[0076] (2) Calculate the attention score e i :

[0077] e i = v T tanh(h i );

[0078] where v is a learnable parameter vector;

[0079] (3) Use the softmax function to normalize the attention score to weights α i :

[0080]

[0081] where Y attn is the weighted feature matrix representing the contributions of different carbon sources to the overall carbon emission prediction.

[0082] Then, use the attention weights α i to perform a weighted sum of the carbon emission feature vector y i to obtain the output of the attention mechanism:

[0083]

[0084] where y attn(i) is the attention-weighted feature vector for each carbon source i, y attn(i,m) is the attention-weighted eigenvalue for each carbon source i on the m-th day in the past, and y attn(i) = [y attn(i,1) ~ y attn(i,m) .

[0085] Finally, based on the attention-weighted feature vectors of all carbon sources over the preset number of days in the past, determine the attention-weighted feature matrix Y attn :

[0086]

[0087] where y attn(n,m)is the attention-weighted eigenvalue of carbon source n on the m-th day in the past.

[0088] In a specific embodiment, the attention-weighted feature matrix Y attn is input into a long short-term memory network to capture the long-term dependencies in the time series data, so as to output the carbon emission data of the target object in the next k days

[0089]

[0090] wherein, the carbon emission data of carbon source i in the next k days is y future(i,m+1) ~y future(i,m+k) .

[0091] In addition, according to this embodiment, a storage medium is also provided. The storage medium includes a stored program, wherein the method of any one of the above is executed by a processor when the program runs.

[0092] This application first determines the carbon emission data vector of each carbon source based on the carbon emission data of the target object in the past preset number of days, so as to fully consider the differences and interactions between different carbon sources, and effectively avoid the problem of key feature loss caused by general processing of all carbon source data. Then, an improved temporal convolutional network is used to determine the carbon emission feature matrix. This network assigns independent convolutional kernels and dilation coefficients to each carbon source, which can more accurately capture the carbon emission characteristics of each carbon source, and the dilation coefficient is determined in real time based on the carbon emission fluctuation of the carbon source, further improving the prediction accuracy. Secondly, a pre-trained attention mechanism network is used to process the carbon emission feature matrix to obtain the attention-weighted feature matrix. The attention mechanism can automatically learn and emphasize the information that has an important impact on the prediction result, further improving the prediction accuracy. Finally, the attention-weighted feature matrix is input into a pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source in the next preset number of days. The long short-term memory network has strong time series processing ability, can capture the long-term dependencies in the carbon emission data, and realize the accurate prediction of future carbon emissions. Thus, this application combines refined data prediction, an improved temporal convolutional network, an attention mechanism network and a long short-term memory network, fully considering the importance differences and fluctuation feature differences of different carbon sources, and significantly improving the accuracy of carbon emission prediction. Furthermore, it solves the technical problem in the prior art that the carbon emission prediction method does not fully consider the importance differences and fluctuation feature differences of different carbon sources, and the simple combination of data results in limited prediction accuracy.

[0093] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0094] Embodiment 2

[0095] Figure 3 The structural schematic diagram of a carbon emission prediction system 300 based on big data analysis according to the present embodiment is shown. The carbon emission prediction system 300 includes: a first determination module 310, configured to determine the carbon emission data vector of each carbon source of the target object in the past preset number of days based on the carbon emission data of the target object in the past preset number of days; wherein, the target object has at least two carbon sources; a second determination module 320, configured to determine the carbon emission feature matrix of the target object by using an improved temporal convolutional network according to the carbon emission data vectors of all carbon sources in the past preset number of days; wherein, the improved temporal convolutional network assigns independent convolutional kernels and dilation coefficients to each carbon source, and the dilation coefficients are determined in real time based on the carbon emission fluctuation conditions of the carbon sources; the dimension of the carbon emission feature matrix is n×m, where n is the number of carbon sources and m is the past preset number of days, and each element in the carbon emission feature matrix represents the carbon emission feature value of the corresponding carbon source on the j-th day in the past; and a third determination module 330, configured to process the carbon emission feature matrix by using a pre-trained attention mechanism network to obtain the attention-weighted feature matrix of the target object, and input the attention-weighted feature matrix into a pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source of the target object in the future preset number of days.

[0096] Optionally, the determining the carbon emission feature matrix of the target object by using an improved temporal convolutional network according to the carbon emission data vectors of all carbon sources in the past preset number of days includes: determining the carbon emission data matrix of the target object in the past preset number of days according to the carbon emission data vectors of all carbon sources in the past preset number of days; and inputting the carbon emission data matrix into the improved temporal convolutional network to output the carbon emission feature matrix of the target object.

[0097] Optionally, the improved temporal convolutional network includes a variance calculation unit, a dilation coefficient determination unit, and a temporal convolutional layer; and the operation of inputting the carbon emission data matrix into the improved temporal convolutional network and outputting the carbon emission feature matrix of the target object includes: inputting the carbon emission data matrix into the variance calculation unit to determine the variance vector of the target object; where each element in the variance vector represents the variance calculation result of the carbon emission data vector of the corresponding carbon source in the past preset number of days; inputting the variance vector into the dilation coefficient determination unit to determine the dilation coefficient corresponding to each carbon source, and transmitting the dilation coefficient to the temporal convolutional layer; and using the temporal convolutional layer to perform dilation convolution on the carbon emission data vectors of all carbon sources in the past preset number of days in parallel according to the corresponding dilation coefficient to determine the carbon emission feature matrix of the target object.

[0098] Optionally, the operation of inputting the carbon emission data matrix into the variance calculation unit to determine the variance vector of the target object includes: using the variance calculation unit to extract the carbon emission data vector of each carbon source in the past preset number of days from the carbon emission data matrix; using the variance calculation unit to calculate the variance value of the carbon emission data vector of each carbon source in the past preset number of days according to the following formula:

[0099]

[0100] where, Var(x i ) is the variance value of the carbon emission data vector x i of carbon source i in the past preset number of days, x i = [x i,1 ~ x i,m ; m is the past preset number of days, x i,j is the carbon emission data of carbon source i on the j-th day in the past; μ i is the mean of x i ; and

[0101] based on the variance values of the carbon emission data vectors of all carbon sources in the past preset number of days, determine the variance vector V of the target object; where, V = [Var(x i ) ~ Var(x n ), and n is the number of carbon sources.

[0102] Optionally, the operation of inputting the variance vector into the dilation coefficient determination unit to determine the dilation coefficient corresponding to each carbon source includes: using the dilation coefficient determination unit to calculate the dilation coefficient corresponding to each carbon source according to the following formula:

[0103] d i = a·Var(x i ) + b;

[0104] Among them, d i is the expansion coefficient corresponding to carbon source i; Var(x i ) is the variance value of the carbon emission data vector x i of carbon source i in the past preset number of days; a and b are constants.

[0105] Optionally, the operation of using the temporal convolutional layer to perform dilated convolution on the carbon emission data vectors of all carbon sources in the past preset number of days in parallel according to the corresponding expansion coefficients to determine the carbon emission feature matrix of the target object includes: using the temporal convolutional layer to perform dilated convolution on the carbon emission data vectors of all carbon sources in the past preset number of days in parallel according to the corresponding expansion coefficients through the following formula to obtain the carbon emission feature values of each carbon source in the past preset number of days:

[0106]

[0107] Among them, y i,j is the carbon emission feature value of carbon source i on the j-th day in the past; ω k is the k-th weight of the convolutional kernel; k max is the size of the convolutional kernel; d i is the expansion coefficient corresponding to carbon source i; is the carbon emission data vector x i of carbon source i at position j - d i ·k; k is the index of the convolutional kernel, k = 0, 1, ···, k;

[0108] Based on the carbon emission feature values of all carbon sources in the past preset number of days, determine the carbon emission feature matrix Y of the target object;

[0109]

[0110] Among them, y n,m is the carbon emission feature value of the n-th carbon source on the m-th day in the past.

[0111] Optionally, the operation of using the pre-trained attention mechanism network to process the carbon emission feature matrix to obtain the attention-weighted feature matrix of the target object includes: using the attention mechanism network to obtain the carbon emission feature vector y i of carbon source i from the carbon emission feature matrix; among them, y i = [y i,1 ~y i,m , y i,m is the carbon emission feature value of the i-th carbon source on the m-th day in the past, and i is any one of 1 to n; based on the carbon emission feature vector y i, calculate the attention weight α of the carbon source i i ; use the attention weight α i to perform weighted summation on the carbon emission feature vector y i to obtain the attention weighted feature vector y of the carbon source i attn(i) ; and determine the attention weighted feature matrix Y of the target object based on the attention weighted feature vectors of all carbon sources in the past preset number of days attn :

[0112]

[0113] where y attn(n,m) is the attention weighted feature value of carbon source n on the mth day in the past.

[0114] Therefore, according to this embodiment, first, based on the carbon emission data of the target object in the past preset number of days, the carbon emission data vector of each carbon source is determined to fully consider the differences and interactions between different carbon sources, effectively avoiding the problem of key feature loss caused by general processing of all carbon source data. Then, the improved temporal convolutional network is used to determine the carbon emission feature matrix. This network assigns independent convolutional kernels and dilation coefficients to each carbon source, can capture the carbon emission characteristics of each carbon source more accurately, and the dilation coefficient is determined in real time based on the carbon emission fluctuation of the carbon source, further improving the prediction accuracy. Second, the pre-trained attention mechanism network is used to process the carbon emission feature matrix to obtain the attention weighted feature matrix. The attention mechanism can automatically learn and emphasize the information that has an important impact on the prediction result, further improving the prediction accuracy. Finally, the attention weighted feature matrix is input into the pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source in the future preset number of days. The long short-term memory network has a strong ability to process time series, can capture the long-term dependence relationship in the carbon emission data, and achieve accurate prediction of future carbon emissions. Thus, this application combines refined data prediction, the improved temporal convolutional network, the attention mechanism network and the long short-term memory network, fully considering the importance differences and fluctuation feature differences of different carbon sources, significantly improving the accuracy of carbon emission prediction. Furthermore, it solves the technical problem in the prior art that the carbon emission prediction method does not fully consider the importance differences and fluctuation feature differences of different carbon sources, and the prediction accuracy is limited by simply merging data.

[0115] Embodiment 3

[0116] Figure 4The carbon emission prediction system 400 based on big data analysis according to this embodiment is shown, including: a processor 410; and a memory 420, connected to the processor 410, for providing instructions for the processor 410 to process the following steps: based on the carbon emission data of the target object in the past preset number of days, determining the carbon emission data vectors of each carbon source of the target object in the past preset number of days; wherein, the target object has at least two carbon sources; according to the carbon emission data vectors of all carbon sources in the past preset number of days, using the improved temporal convolutional network, determining the carbon emission feature matrix of the target object; wherein, the improved temporal convolutional network assigns independent convolutional kernels and dilation coefficients to each carbon source, and the dilation coefficients are determined in real time based on the carbon emission fluctuation conditions of the carbon sources; the dimension of the carbon emission feature matrix is n×m, where n is the number of carbon sources and m is the past preset number of days, and each element in the carbon emission feature matrix represents the carbon emission feature value of the corresponding carbon source on the j-th day in the past; and using the pre-trained attention mechanism network to process the carbon emission feature matrix to obtain the attention-weighted feature matrix of the target object, and inputting the attention-weighted feature matrix into the pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source of the target object in the future preset number of days.

[0117] Optionally, the step of determining the carbon emission feature matrix of the target object by using the improved temporal convolutional network according to the carbon emission data vectors of all carbon sources in the past preset number of days includes: determining the carbon emission data matrix of the target object in the past preset number of days according to the carbon emission data vectors of all carbon sources in the past preset number of days; and inputting the carbon emission data matrix into the improved temporal convolutional network to output the carbon emission feature matrix of the target object.

[0118] Optionally, the improved temporal convolutional network includes a variance calculation unit, a dilation coefficient determination unit, and a temporal convolutional layer; and the operation of inputting the carbon emission data matrix into the improved temporal convolutional network to output the carbon emission feature matrix of the target object includes: inputting the carbon emission data matrix into the variance calculation unit to determine the variance vector of the target object; wherein each element in the variance vector represents the variance calculation result of the carbon emission data vector of the corresponding carbon source in the past preset number of days; inputting the variance vector into the dilation coefficient determination unit to determine the dilation coefficient corresponding to each carbon source, and transmitting the dilation coefficient to the temporal convolutional layer; and using the temporal convolutional layer to perform dilated convolution on the carbon emission data vectors of all carbon sources in the past preset number of days in parallel according to the corresponding dilation coefficients to determine the carbon emission feature matrix of the target object.

[0119] Optionally, the operation of inputting the carbon emission data matrix into the variance calculation unit to determine the variance vector of the target object includes: using the variance calculation unit to extract the carbon emission data vector of each carbon source in the past preset number of days from the carbon emission data matrix; using the variance calculation unit to calculate the variance value of the carbon emission data vector of each carbon source in the past preset number of days according to the following formula:

[0120]

[0121] where Var(x i ) is the variance value of the carbon emission data vector x i of carbon source i in the past preset number of days, x i = [x i,1 ~ x i,m ; m is the past preset number of days, x i,j is the carbon emission data of carbon source i on the jth day in the past; μ i is the mean value of x i ; and

[0122] based on the variance values of the carbon emission data vectors of all carbon sources in the past preset number of days, determine the variance vector V of the target object; where V = [Var(x i ) ~ Var(x n ), and n is the number of carbon sources.

[0123] Optionally, the operation of inputting the variance vector into the expansion coefficient determination unit to determine the expansion coefficient corresponding to each carbon source includes: using the expansion coefficient determination unit to calculate the expansion coefficient corresponding to each carbon source according to the following formula:

[0124] d i = a·Var(x i ) + b;

[0125] where d i is the expansion coefficient corresponding to carbon source i; Var(x i ) is the variance value of the carbon emission data vector x i of carbon source i in the past preset number of days; a and b are constants.

[0126] Optionally, the operation of using the temporal convolutional layer to perform dilated convolution on the carbon emission data vectors of all carbon sources in the past preset number of days in parallel according to the corresponding expansion coefficients to determine the carbon emission feature matrix of the target object includes: using the temporal convolutional layer to perform dilated convolution on the carbon emission data vectors of all carbon sources in the past preset number of days in parallel according to the corresponding expansion coefficients through the following formula to obtain the carbon emission feature values of each carbon source in the past preset number of days:

[0127]

[0128] Among them, y i,j is the carbon emission characteristic value of carbon source i on the j-th day in the past; ω k is the k-th weight of the convolutional kernel; k max is the size of the convolutional kernel; d i is the dilation coefficient corresponding to carbon source i; is the carbon emission data vector x of carbon source i in the past preset number of days i at position j - d i ·k; k is the index of the convolutional kernel, k = 0, 1, ···, k;

[0129] Based on the carbon emission characteristic values of all carbon sources in the past preset number of days, determine the carbon emission characteristic matrix Y of the target object;

[0130]

[0131] Among them, y n,m is the carbon emission characteristic value of the n-th carbon source on the m-th day in the past.

[0132] Optionally, the operation of using the pre-trained attention mechanism network to process the carbon emission characteristic matrix to obtain the attention-weighted characteristic matrix of the target object includes: using the attention mechanism network to obtain the carbon emission characteristic vector y of carbon source i from the carbon emission characteristic matrix i ; Among them, y i = [y i,1 ~y i,m , y i,m is the carbon emission characteristic value of the i-th carbon source on the m-th day in the past, and i is any one of 1 to n; based on the carbon emission characteristic vector y i , calculate the attention weight α i of carbon source i; use the attention weight α i to perform weighted summation on the carbon emission characteristic vector y i to obtain the attention-weighted characteristic vector y attn(i) of carbon source i; and based on the attention-weighted characteristic vectors of all carbon sources in the past preset number of days, determine the attention-weighted characteristic matrix Y attn of the target object:

[0133]

[0134] Among them, y attn(n,m) is the attention-weighted characteristic value of carbon source n on the m-th day in the past.

[0135] Therefore, according to this embodiment, first, based on the carbon emission data of the target object in the past preset number of days, the carbon emission data vector of each carbon source is determined, so as to fully consider the differences and interactions between different carbon sources, effectively avoiding the problem of loss of key features caused by generalizing all carbon source data. Then, the improved temporal convolutional network is used to determine the carbon emission feature matrix. This network assigns independent convolutional kernels and dilation coefficients to each carbon source, which can capture the carbon emission characteristics of each carbon source more accurately, and the dilation coefficient is determined in real time based on the carbon emission fluctuation of the carbon source, further improving the prediction accuracy. Secondly, the pre-trained attention mechanism network is used to process the carbon emission feature matrix to obtain the attention-weighted feature matrix. The attention mechanism can automatically learn and emphasize the information that has an important impact on the prediction result, further improving the prediction accuracy. Finally, the attention-weighted feature matrix is input into the pre-trained long short-term memory network to determine the carbon emission prediction data of each carbon source in the future preset number of days. The long short-term memory network has a powerful time series processing ability, which can capture the long-term dependence relationship in the carbon emission data and achieve accurate prediction of future carbon emissions. Thus, through the combination of refined data prediction, the improved temporal convolutional network, the attention mechanism network and the long short-term memory network, this application fully considers the importance differences and fluctuation feature differences of different carbon sources, significantly improving the accuracy of carbon emission prediction. Furthermore, it solves the technical problem in the prior art that the carbon emission prediction method does not fully consider the importance differences and fluctuation feature differences of different carbon sources, and the simple combination of data results in limited prediction accuracy.

[0136] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

[0137] In the above embodiments of the present invention, the descriptions of the respective embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0138] In several embodiments provided by this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the units or modules can be in an electrical or other form.

[0139] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0140] In addition, in each embodiment of the present invention, each functional unit may be integrated in a processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0141] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The foregoing storage medium includes: USB flash drive, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disc and other various media that can store program codes.

[0142] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A carbon emission prediction method based on big data analysis, characterized in that: include: Based on the carbon emission data of the target object in the past preset days, determining the carbon emission data vector of each carbon source of the target object in the past preset days; wherein the target object has at least two carbon sources; According to the carbon emission data vectors of all carbon sources in the past preset days, the carbon emission feature matrix of the target object is determined using an improved temporal convolutional network; wherein the improved temporal convolutional network assigns an independent convolution kernel and expansion coefficient to each carbon source, and the expansion coefficient is determined in real time based on the carbon emission fluctuation of the carbon source; the dimension of the carbon emission feature matrix is ​​n×m, wherein n is the number of carbon sources, m is the past preset number of days, and each element in the carbon emission feature matrix represents the carbon emission feature value of the corresponding carbon source on the jth day in the past; and The carbon emission feature matrix is ​​processed using a pre-trained attention mechanism network to obtain an attention weighted feature matrix of the target object, and the attention weighted feature matrix is ​​input into a pre-trained long short-term memory network to determine the carbon emission forecast data of each carbon source of the target object in the future preset number of days.

2. The method according to claim 1, characterized in that The method of determining the carbon emission feature matrix of the target object by using the improved temporal convolutional network based on the carbon emission data vectors of all carbon sources in the past preset days includes: Determining a carbon emission data matrix of the target object in the past preset days based on the carbon emission data vectors of all carbon sources in the past preset days; and The carbon emission data matrix is ​​input into the improved temporal convolutional network, and the carbon emission feature matrix of the target object is output.

3. The method according to claim 2, characterized in that The improved temporal convolutional network includes a variance calculation unit, a dilation coefficient determination unit and a temporal convolutional layer; Furthermore, the operation of inputting the carbon emission data matrix into the improved temporal convolutional network and outputting the carbon emission feature matrix of the target object includes: Inputting the carbon emission data matrix into the variance calculation unit to determine the variance vector of the target object; wherein each element in the variance vector represents the variance calculation result of the carbon emission data vector of the corresponding carbon source in the past preset number of days; Inputting the variance vector into the expansion coefficient determination unit, determining the expansion coefficient corresponding to each carbon source, and transmitting the expansion coefficient to the temporal convolution layer; and The temporal convolution layer is used to perform dilation convolution on the carbon emission data vectors of all carbon sources in the past preset days in parallel according to the corresponding dilation coefficients to determine the carbon emission feature matrix of the target object.

4. The method according to claim 3, characterized in that The operation of inputting the carbon emission data matrix into the variance calculation unit to determine the variance vector of the target object includes: Utilizing the variance calculation unit, extracting the carbon emission data vector of each carbon source in the past preset days from the carbon emission data matrix; The variance calculation unit is used to calculate the variance value of the carbon emission data vector of each carbon source in the past preset days according to the following formula: Among them, Var(x i ) is the carbon emission data vector x of carbon source i in the past preset days i The variance value of x i =[x i,1 ~x i,m ]; m is the preset number of days in the past, x i,j is the carbon emission data of carbon source i on the past j day; μ i is x i The mean of Based on the variance values ​​of the carbon emission data vectors of all carbon sources in the past preset number of days, the variance vector V of the target object is determined; wherein V=[Var(x i )~Var(x n )], n is the number of carbon sources.

5. The method according to claim 3, characterized in that: The operation of inputting the variance vector into the expansion coefficient determination unit to determine the expansion coefficient corresponding to each carbon source includes: The expansion coefficient determination unit is used to calculate the expansion coefficient corresponding to each carbon source according to the following formula: d i =a·Exists(x i )+b; Among them, d i is the expansion coefficient corresponding to carbon source i; Var(x i ) is the carbon emission data vector x of carbon source i in the past preset days i The variance value of ; a and b are constants.

6. The method according to claim 3, characterized in that: The operation of using the temporal convolution layer to perform dilated convolution on the carbon emission data vectors of all carbon sources in the past preset days in parallel according to the corresponding dilation coefficient to determine the carbon emission feature matrix of the target object includes: By using the temporal convolution layer, according to the corresponding dilation coefficient, the carbon emission data vectors of all carbon sources in the past preset days are dilated and convolved in parallel by the following formula to obtain the carbon emission characteristic value of each carbon source in the past preset days: Among them, y i,j is the carbon emission characteristic value of carbon source i on the jth day in the past; ω k is the kth weight of the convolution kernel; k max is the convolution kernel size; d i is the expansion coefficient corresponding to carbon source i; is the carbon emission data vector x of carbon source i in the past preset days i In position jd i The value at k; k is the index of the convolution kernel, k=0,1,···,k; Determine the carbon emission characteristic matrix Y of the target object based on the carbon emission characteristic values ​​of all carbon sources in the past preset days; Among them, y n,m is the carbon emission characteristic value of the nth carbon source in the past mth day.

7. The method according to claim 1, characterized in that The operation of processing the carbon emission feature matrix using the pre-trained attention mechanism network to obtain the attention weighted feature matrix of the target object includes: Using the attention mechanism network, the carbon emission feature vector y of carbon source i is obtained from the carbon emission feature matrix i ; Among them, y i =[y i,1 ~y i,m ],y i,m is the carbon emission characteristic value of the i-th carbon source in the past m-th day, where i is any one of 1 to n; Based on the carbon emission characteristic vector y i , calculate the attention weight α of the carbon source i i ; Use attention weight α i For the carbon emission characteristic vector y i Perform weighted summation to obtain the attention weighted feature vector y of the carbon source i attn(i) ;as well as Based on the attention weighted feature vectors of all carbon sources in the past preset number of days, determine the attention weighted feature matrix Y of the target object attn : Among them, y attn(n,m) is the attention-weighted feature value of carbon source n in the past m days.

8. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is run, the processor executes the method according to any one of claims 1 to 7.

9. A carbon emission prediction system based on big data analysis, comprising: A first determination module is used to determine the carbon emission data vector of each carbon source of the target object in the past preset days based on the carbon emission data of the target object in the past preset days; wherein the target object has at least two carbon sources; A second determination module is used to determine the carbon emission characteristic matrix of the target object using an improved temporal convolutional network based on the carbon emission data vectors of all carbon sources in the past preset days; wherein the improved temporal convolutional network assigns an independent convolution kernel and expansion coefficient to each carbon source, and the expansion coefficient is determined in real time based on the carbon emission fluctuation of the carbon source; the dimension of the carbon emission characteristic matrix is ​​n×m, wherein n is the number of carbon sources, m is the past preset number of days, and each element in the carbon emission characteristic matrix represents the carbon emission characteristic value of the corresponding carbon source on the jth day in the past; and The third determination module is used to process the carbon emission feature matrix using a pre-trained attention mechanism network to obtain an attention weighted feature matrix of the target object, and input the attention weighted feature matrix into a pre-trained long short-term memory network to determine the carbon emission forecast data of each carbon source of the target object in a preset number of days in the future.

10. A carbon emission prediction system based on big data analysis, characterized in that: include: processor; as well as A memory, connected to the processor, configured to provide the processor with instructions for processing the following processing steps: Based on the carbon emission data of the target object in the past preset days, determining the carbon emission data vector of each carbon source of the target object in the past preset days; wherein the target object has at least two carbon sources; According to the carbon emission data vectors of all carbon sources in the past preset days, the carbon emission feature matrix of the target object is determined using an improved temporal convolutional network; wherein the improved temporal convolutional network assigns an independent convolution kernel and expansion coefficient to each carbon source, and the expansion coefficient is determined in real time based on the carbon emission fluctuation of the carbon source; the dimension of the carbon emission feature matrix is ​​n×m, wherein n is the number of carbon sources, m is the past preset number of days, and each element in the carbon emission feature matrix represents the carbon emission feature value of the corresponding carbon source on the jth day in the past; and The carbon emission feature matrix is ​​processed using a pre-trained attention mechanism network to obtain an attention weighted feature matrix of the target object, and the attention weighted feature matrix is ​​input into a pre-trained long short-term memory network to determine the carbon emission forecast data of each carbon source of the target object in the future preset number of days.

Images