Carbon emission prediction method and system based on electric power big data

Abstract

The invention relates to a carbon emission prediction method based on electric power big data, and the method comprises the following steps: S1, collecting electric power production and consumption data, and constructing a data pool; s2, statistical features are extracted and generated through statistical analysis and an automatic encoder technology, spatial features are extracted by using a geographic information system technology, periodicity and tendency of time sequence data are analyzed to obtain time features, a causal inference method is introduced to identify key causal features, and finally a comprehensive feature set is constructed; s3, a regression model is adopted as a basic model, LSTM is introduced to process complex and long-sequence data, the nonlinear fitting ability of the model is enhanced, the generalization ability and accuracy of the model are improved through integrated learning, a carbon emission prediction model is constructed, and training is carried out based on a comprehensive feature set; s4, on the basis of the trained carbon emission prediction model, carbon emission prediction is carried out according to real-time data; and S5, displaying a prediction result through a visual tool, and supporting user-defined view and analysis. The accuracy of carbon emission prediction is improved.

Description

Technical Field

[0001] The present invention relates to the field of carbon emission prediction, and in particular to a carbon emission prediction method and system based on power big data. Background Art

[0002] In modern society, electricity, as a core energy source, is widely used in various fields such as industry, commerce and households. However, the large amount of carbon emissions generated during electricity production and consumption is one of the important factors causing global climate change. Therefore, how to effectively predict electricity-related carbon emissions, and then optimize energy use and reduce environmental impact, has become an important topic of concern to governments, academic research institutions and enterprises. With the rapid development of the Internet of Things, sensor technology and smart grids, the collection, storage and analysis capabilities of power big data have been significantly improved, providing a rich data foundation and technical support for carbon emission prediction.

[0003] Power big data usually contains massive, multi-source, and multi-dimensional data, including time series data, such as real-time power load data and power generation records, as well as static geospatial data, and economic and policy-related information. These data are not only huge in quantity, but also complex and diverse. Traditional linear prediction methods have difficulty capturing the nonlinear relationships and implicit long-term and short-term dependencies. Summary of the invention

[0004] In order to solve the above problems, the purpose of the present invention is to provide a carbon emission prediction method and system based on power big data, which can effectively mine the potential patterns and causal relationships in the data, thereby improving the accuracy of carbon emission prediction and the effectiveness of decision support.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A carbon emission prediction method based on power big data includes the following steps:

[0007] S1: Collect electricity production and consumption data, obtain data from different data sources, integrate and uniformly format data, and build a data pool;

[0008] S2: Extract and generate statistical features through statistical analysis and autoencoder technology, use geographic information system technology to extract spatial features, analyze the periodicity and trend of time series data to obtain temporal features, introduce causal inference methods to identify key causal features, and finally construct a comprehensive feature set;

[0009] S3: The regression model is used as the basic model, and LSTM is introduced to process complex and long sequence data, and the nonlinear fitting ability of the model is enhanced. The generalization ability and accuracy of the model are improved through ensemble learning, and a carbon emission prediction model is constructed, which is trained based on a comprehensive feature set;

[0010] S4: Based on the trained carbon emission prediction model, carbon emission prediction is performed according to real-time data;

[0011] S5: Display prediction results through visualization tools, supporting user-defined views and analysis.

[0012] Furthermore, S1 is specifically:

[0013] Deploy edge computing devices as data aggregation gateways to preprocess, filter, and buffer data;

[0014] Collect electricity production and consumption data in real time through sensors and smart meters, and connect the sensors and smart meters to the data aggregation gateway;

[0015] Develop standardized APIs for power company records, policies and regulations, climate and environmental data, and socio-economic parameter data sources to obtain data from different data sources;

[0016] Use Apache NiFi to implement data extraction, transformation, and loading processes to ensure data quality and consistency;

[0017] Build a data lake based on Hadoop HDFS to store raw and normalized data and support large-scale data analysis and processing.

[0018] Furthermore, S2 is specifically:

[0019] S2.1: Use Python's Pandas and NumPy libraries to perform basic statistical analysis and calculate basic statistics such as mean, variance, maximum, and minimum. Apply principal component analysis to reduce the dimension and extract the main statistical features.

[0020] S2.2: Use the deep learning framework to build an autoencoder model, perform unsupervised feature learning on the data, extract the low-dimensional representation of the data through the encoder part, and obtain nonlinear features;

[0021] S2.3: Use Python's GeoPandas library to integrate geospatial data, combine power data with geographic data, and generate initial spatial features; use Kriging interpolation to analyze the spatial distribution of geographic data; identify geographic hot spots through spatial clustering algorithms to obtain final spatial features;

[0022] S2.4: Use Python's statsmodels library to decompose the time series, separate the trend, seasonality, and residual components, and obtain the temporal characteristics;

[0023] S2.5: Use structural equation modeling to model the causal relationship between variables and obtain causal characteristics;

[0024] S2.6: Based on the features obtained above, construct a comprehensive feature matrix.

[0025] Furthermore, S2.3 is specifically:

[0026] S2.3.1: Import the GeoPandas library and load different datasets, including power consumption, topography, land use, and meteorological conditions data, and use the GeoPandas overlay function to combine the power data with the geographic data to generate an initial spatial feature dataset containing multidimensional attributes;

[0027] S2.3.2: Based on the initial spatial feature data set, use Kriging interpolation to perform spatial interpolation analysis to determine the possible value distribution of power data in unobserved areas:

[0028] S2.3.3: Based on the smoothed spatial data generated by interpolation, perform spatial clustering analysis (DBSCAN) to identify spatially dense hotspots and noise.

[0029] Further, S2.3.2 is as follows:

[0030] Confirm that the projection system of the initial spatial feature dataset is consistent to enable accurate geospatial analysis;

[0031] Analyze the distribution characteristics of power data, check spatial autocorrelation, and preliminarily determine whether the data is suitable for Kriging interpolation;

[0032] Semivariogram is used to describe the spatial autocorrelation between data points:

[0033]

[0034] Among them, γ(h) is the semivariogram function value, h is the distance interval; N(h) is the number of data point pairs with distance h; z(x i ) is the position x i The data observation value, z(x i +h) is the distance h and position x i + another data observation of h;

[0035] A Gaussian model was selected to fit based on the empirical estimate of the semivariogram, and model parameters were determined, including the sill, range, and nugget effect;

[0036] The spatial covariance matrix is ​​determined using the semivariogram through the equation construction of the kriging system:

[0037]

[0038] Among them, λ j is the Kriging weight, C(xi ,x j ) is the position x i and x j The covariance between ; x0 is the predicted position;

[0039] According to the Gaussian model, the covariance matrix between known points is calculated to solve the Kriging weights, and the estimated values ​​of the unknown points are calculated using the obtained Kriging weights:

[0040]

[0041] in, is the interpolation result of the predicted position x0.

[0042] Further, S2.5 is as follows:

[0043] Identify the causal relationships that need to be analyzed, obtain electricity data indicators that have an impact on carbon emissions greater than the threshold, and make assumptions based on theoretical frameworks and domain knowledge;

[0044] List all relevant observed and latent variables and hypothesize the structural relationships between them;

[0045] Construct a causal path diagram based on prior knowledge to indicate the assumed causal relationship, such as how electricity consumption and weather changes affect carbon emissions.

[0046] The variables are connected through directed paths, representing the causal hypothesis relationship, and are divided into a structural part and a measurement part.

[0047] Build the structural model:

[0048] η=Bη+Γξ+ζ;

[0049] Among them, η is the endogenous variable matrix, B is the path coefficient matrix between endogenous variables; ξ is the exogenous latent variable, Γ is the path coefficient matrix of the exogenous variable to the endogenous variable, and ζ is the error term;

[0050] The relationship between latent variables and observed variables is described by the measurement model:

[0051] y=Λ y η+∈

[0052] x=Λ x ξ+δ;

[0053] Where y is the actual measured vector, and x is the external direct measurement vector; Λ y is the latent factor loading matrix, Λ x is the latent factor loading matrix, ∈ and δ are noise vectors;

[0054] Through the size and significance test of the path coefficient, the strength and direction of the causal relationship between the variables can be interpreted;

[0055] Key causal features are extracted from the structural model for subsequent carbon emission prediction models, including correlation coefficients in significant paths and explanatory components of latent variables.

[0056] Furthermore, the carbon emission prediction model is as follows:

[0057] Ridge regression was used as the base model to capture simple linear relationships;

[0058] J(β)=||Y-Xβ|| 2 +λ||β|| 2 ;

[0059] Among them, Y is the target variable vector, X is the comprehensive feature matrix, β is the coefficient vector, and λ is the regularization parameter;

[0060] The long short-term memory network LSTM is introduced to process long time series data and capture the time dependency therein; the long short-term memory network LSTM specifically includes:

[0061] Input layer: x i ;

[0062] Forget gate: f i =σ(W f ·|h i-1 ,x i |+b f );

[0063] Input gate: I i =σ(W I ·|h i-1 ,x i |+b I );

[0064] Candidate values:

[0065] Cell status update:

[0066] Output gate: o t =σ(W o ·|h i-1 ,x i |+b0);

[0067] Hide status update:h t =o i tanh(C i );

[0068] Output layer: y t ;

[0069] Among them, x i is the input feature vector of time step i; h i-1 is the hidden state at time step i; f i For the Forget Gate; i is the input gate; is the candidate value; that is, the candidate memory cell at the current moment; C t is the cell state; it indicates the memory cell state at the current moment; o t is the output gate; h t is the hidden state at the current moment; W f , W I , W C , W o are the weight matrices of the forget gate, input gate, candidate value, and output gate respectively; b f 、b I 、b C , b0 are the bias vectors of the forget gate, input gate, candidate value and output gate respectively;

[0070] Based on the gradient boosting tree XGBoost, the prediction results of different models are integrated to build a carbon emission prediction model:

[0071] Y final＝ XGBoost(Y ridge ,Y LSTM );

[0072] Among them, Y ridge is the output of ridge regression; Y LSTM Output of the LSTM model.

[0073] A carbon emission prediction system based on electric power big data includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the carbon emission prediction method based on electric power big data as described above.

[0074] The present invention has the following beneficial effects:

[0075] 1. The present invention effectively mines potential patterns and causal relationships in data, thereby improving the accuracy of carbon emission forecasts and the effectiveness of decision support;

[0076] 2. The present invention integrates multiple types of features (statistical, nonlinear, spatial, temporal, causal) so that the model can comprehensively analyze the driving factors of carbon emissions from multiple dimensions. Through multivariate feature extraction and comprehensive analysis, it not only improves the prediction ability of carbon emissions, but also enhances the explanatory power of the model;

[0077] 3. The present invention combines ridge regression, LSTM and XGBoost models, which can fully utilize their respective advantages, capture the linear relationship and time dependency in the data, and improve the prediction performance of the overall model through integrated learning, effectively improving the accuracy and reliability of carbon emission prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 is a flow chart of the method of the present invention;

[0079] Figure 2 The figure is a schematic diagram of constructing a comprehensive feature matrix in one embodiment of the present invention. DETAILED DESCRIPTION

[0080] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0081] refer to Figure 1 In this embodiment, a carbon emission prediction method based on power big data is provided, comprising the following steps:

[0082] S1: Collect electricity production and consumption data, obtain data from different data sources, integrate and uniformly format data, and build a data pool;

[0083] S2: Extract and generate statistical features through statistical analysis and autoencoder technology, use geographic information system technology to extract spatial features, analyze the periodicity and trend of time series data to obtain temporal features, introduce causal inference methods to identify key causal features, and finally construct a comprehensive feature set;

[0084] S3: The regression model is used as the basic model, and LSTM is introduced to process complex and long sequence data, and the nonlinear fitting ability of the model is enhanced. The generalization ability and accuracy of the model are improved through ensemble learning, and a carbon emission prediction model is constructed, which is trained based on a comprehensive feature set;

[0085] S4: Based on the trained carbon emission prediction model, carbon emission prediction is performed according to real-time data;

[0086] S5: Display prediction results through visualization tools (such as Tableau and PowerBI), supporting user-defined views and analysis.

[0087] In this embodiment, S1 is specifically:

[0088] Deploy edge computing devices as data aggregation gateways to preprocess, filter, and buffer data;

[0089] Collect electricity production and consumption data in real time through sensors and smart meters, and connect the sensors and smart meters to the data aggregation gateway;

[0090] Develop standardized APIs for power company records, policies and regulations, climate and environmental data, and socio-economic parameter data sources to obtain data from different data sources;

[0091] Use Apache NiFi to implement data extraction, transformation, and loading processes to ensure data quality and consistency;

[0092] Build a data lake based on Hadoop HDFS to store raw and normalized data and support large-scale data analysis and processing.

[0093] In this embodiment, S2 is specifically:

[0094] S2.1: Use Python's Pandas and NumPy libraries to perform basic statistical analysis and calculate basic statistics such as mean, variance, maximum, and minimum. Apply principal component analysis to reduce the dimension and extract the main statistical features.

[0095] S2.2: Use a deep learning framework (such as TensorFlow or PyTorch) to build an autoencoder model to perform unsupervised feature learning on the data, extract a low-dimensional representation of the data through the encoder part, and obtain nonlinear features;

[0096] S2.3: Use Python's GeoPandas library to integrate geospatial data, combine power data with geographic data (such as topography, land use, and meteorological conditions) to generate initial spatial features; use Kriging interpolation to analyze the spatial distribution of geographic data; identify geographic hotspots through spatial clustering algorithms (such as DBSCAN) to obtain final spatial features;

[0097] S2.4: Use Python's statsmodels library to decompose the time series, separate the trend, seasonality, and residual components, and obtain the temporal characteristics;

[0098] S2.5: Use structural equation modeling to model the causal relationship between variables and obtain causal characteristics;

[0099] S2.6: Based on the features obtained above, construct a comprehensive feature matrix.

[0100] In this embodiment, S2.3 is specifically:

[0101] S2.3.1: Import the GeoPandas library and load different datasets, including power consumption, topography, land use, and meteorological conditions data, and use the GeoPandas overlay function to combine the power data with the geographic data to generate an initial spatial feature dataset containing multidimensional attributes;

[0102] S2.3.2: Based on the initial spatial feature data set, use Kriging interpolation to perform spatial interpolation analysis to determine the possible value distribution of power data in unobserved areas:

[0103] S2.3.3: Based on the smoothed spatial data generated by interpolation, perform spatial clustering analysis (DBSCAN) to identify spatially dense hotspots and noise.

[0104] In this embodiment, S2.3.2 is as follows:

[0105] Confirm that the projection system (CRS) of the initial spatial feature dataset is consistent to enable accurate geospatial analysis;

[0106] Analyze the distribution characteristics of power data, check spatial autocorrelation, and preliminarily determine whether the data is suitable for Kriging interpolation;

[0107] Semivariogram is used to describe the spatial autocorrelation between data points:

[0108]

[0109] Among them, γ(h) is the semivariogram function value, h is the distance interval; N(h) is the number of data point pairs with distance h; z(x i ) is the position x i The data observation value, z(x i +h) is the distance h and position x i + another data observation of h;

[0110] A Gaussian model was selected to fit based on the empirical estimate of the semivariogram, and model parameters were determined, including the sill, range, and nugget effect;

[0111] The spatial covariance matrix is ​​determined using the semivariogram through the equation construction of the kriging system:

[0112]

[0113] Among them, λ j is the Kriging weight, C(x i ,x j ) is the position x i and x j The covariance between ; x0 is the predicted position;

[0114] According to the Gaussian model, the covariance matrix between known points is calculated to solve the Kriging weights, and the estimated values ​​of the unknown points are calculated using the obtained Kriging weights:

[0115]

[0116] in, is the interpolation result of the predicted position x0.

[0117] In this embodiment, S2.5 is as follows:

[0118] Identify the causal relationships that need to be analyzed, obtain electricity data indicators that have an impact on carbon emissions greater than the threshold, and make assumptions based on theoretical frameworks and domain knowledge;

[0119] List all relevant observed and latent variables and hypothesize the structural relationships between them;

[0120] Construct a causal path diagram based on prior knowledge to indicate the assumed causal relationship, such as how electricity consumption and weather changes affect carbon emissions.

[0121] The variables are connected through directed paths, representing the causal hypothesis relationship, and are divided into a structural part and a measurement part.

[0122] Build the structural model:

[0123] η=Bη+Γξ+ζ;

[0124] Among them, η is the endogenous variable matrix, B is the path coefficient matrix between endogenous variables; ξ is the exogenous latent variable, Γ is the path coefficient matrix of the exogenous variable to the endogenous variable, and ζ is the error term;

[0125] The relationship between latent variables and observed variables is described by the measurement model:

[0126] y=Λ y η+∈

[0127] x=Λ x ξ+δ;

[0128] Where y is the actual measured vector, and x is the external direct measurement vector; Λ y is the latent factor loading matrix, Λ x is the latent factor loading matrix, ∈ and δ are noise vectors;

[0129] Through the size and significance test of the path coefficient, the strength and direction of the causal relationship between the variables can be interpreted;

[0130] Key causal features are extracted from the structural model for subsequent carbon emission prediction models, including correlation coefficients in significant paths and explanatory components of latent variables.

[0131] In this embodiment, the carbon emission prediction model is as follows:

[0132] Ridge regression was used as the base model to capture simple linear relationships;

[0133] J(β)=||Y-Xβ||2 +λ||β|| 2 ;

[0134] Among them, Y is the target variable vector, X is the comprehensive feature matrix, β is the coefficient vector, and λ is the regularization parameter;

[0135] The long short-term memory network LSTM is introduced to process long time series data and capture the time dependency therein; the long short-term memory network LSTM specifically includes:

[0136] Input layer: xi;

[0137] Forget gate: f i =σ(W f ·|h i-1 ,x i |+b f );

[0138] Input gate: I i =σ(W I ·|h i-1 ,x i |+b I );

[0139] Candidate values:

[0140] Cell status update:

[0141] Output gate: o t =σ(W o ·|h i-1 ,x i |+b0);

[0142] Hide status update:h t =o i tanh(C i );

[0143] Output layer: y t ;

[0144] Among them, x i is the input feature vector of time step i; h i-1 is the hidden state at time step i; f i For the Forget Gate; i is the input gate; is the candidate value; that is, the candidate memory cell at the current moment; C t is the cell state; it indicates the memory cell state at the current moment; o t is the output gate; h t is the hidden state at the current moment; W f , W I , WC , W o are the weight matrices of the forget gate, input gate, candidate value, and output gate respectively; b f 、b I 、b C , b0 are the bias vectors of the forget gate, input gate, candidate value and output gate respectively;

[0145] Based on the gradient boosting tree XGBoost, the prediction results of different models are integrated to build a carbon emission prediction model:

[0146] Y final＝ XGBoost(Y ridge ,Y LSTM );

[0147] Among them, Y ridge is the output of ridge regression; Y LSTM Output of the LSTM model.

[0148] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0149] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0150] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0151] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0152] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A carbon emission prediction method based on power big data, characterized in that: The following steps are involved: S1: Collect electricity production and consumption data, obtain data from different data sources, integrate and uniformly format data, and build a data pool; S2: Extract and generate statistical features through statistical analysis and autoencoder technology, use geographic information system technology to extract spatial features, analyze the periodicity and trend of time series data to obtain temporal features, introduce causal inference methods to identify key causal features, and finally construct a comprehensive feature set; S3: The regression model is used as the basic model, and LSTM is introduced to process complex and long sequence data, and the nonlinear fitting ability of the model is enhanced. The generalization ability and accuracy of the model are improved through ensemble learning, and a carbon emission prediction model is constructed, which is trained based on a comprehensive feature set; S4: Based on the trained carbon emission prediction model, carbon emission prediction is performed according to real-time data; S5: Display prediction results through visualization tools, supporting user-defined views and analysis.

2. The carbon emission prediction method based on power big data according to claim 1 is characterized in that: The S1 is specifically: Deploy edge computing devices as data aggregation gateways to preprocess, filter, and buffer data; Collect electricity production and consumption data in real time through sensors and smart meters, and connect the sensors and smart meters to the data aggregation gateway; Develop standardized APIs for power company records, policies and regulations, climate and environmental data, and socio-economic parameter data sources to obtain data from different data sources; Use Apache NiFi to implement data extraction, transformation, and loading processes to ensure data quality and consistency; Build a data lake based on Hadoop HDFS to store raw and normalized data and support large-scale data analysis and processing.

3. The carbon emission prediction method based on power big data according to claim 1 is characterized in that: The S2 is specifically: S2.1: Use Python's Pandas and NumPy libraries to perform basic statistical analysis and calculate basic statistics such as mean, variance, maximum, and minimum. Apply principal component analysis to reduce the dimension and extract the main statistical features. S2.2: Use the deep learning framework to build an autoencoder model, perform unsupervised feature learning on the data, extract the low-dimensional representation of the data through the encoder part, and obtain nonlinear features; S2.3: Use Python's GeoPandas library to integrate geospatial data, combine power data with geographic data, and generate initial spatial features; And use Kriging interpolation to analyze the spatial distribution of geographic data; identify geographic hot spots through spatial clustering algorithms to obtain the final spatial characteristics; S2.4: Use Python's statsmodels library to decompose the time series, separate the trend, seasonality, and residual components, and obtain the temporal characteristics; S2.5: Use structural equation modeling to model the causal relationship between variables and obtain causal characteristics; S2.6: Based on the features obtained above, construct a comprehensive feature matrix.

4. The carbon emission prediction method based on power big data according to claim 3 is characterized in that: The S2.3 is specifically: S2.3.1: Import the GeoPandas library and load different datasets, including power consumption, topography, land use, and meteorological conditions data, and use the GeoPandas overlay function to combine the power data with the geographic data to generate an initial spatial feature dataset containing multidimensional attributes; S2.3.2: Based on the initial spatial feature data set, use Kriging interpolation to perform spatial interpolation analysis to determine the possible value distribution of power data in unobserved areas: S2.3.3: Based on the smoothed spatial data generated by interpolation, perform spatial clustering analysis (DBSCAN) to identify spatially dense hotspots and noise.

5. The carbon emission prediction method based on power big data according to claim 3 is characterized in that: The S2.3.2 is as follows: Confirm that the projection system of the initial spatial feature dataset is consistent to enable accurate geospatial analysis; Analyze the distribution characteristics of power data, check spatial autocorrelation, and preliminarily determine whether the data is suitable for Kriging interpolation; Semivariogram is used to describe the spatial autocorrelation between data points: Among them, γ(h) is the semivariogram function value, h is the distance interval; N(h) is the number of data point pairs with distance h; z(x i ) is the position x i The data observation value, z(x i +h) is the distance h and position x i + another data observation of h; A Gaussian model was selected to fit based on the empirical estimate of the semivariogram, and model parameters were determined, including the sill, range, and nugget effect; The spatial covariance matrix is ​​determined using the semivariogram through the equation construction of the kriging system: Among them, λ j is the Kriging weight, C(x i ,x j ) is the position x i and x j The covariance between ; x0 is the predicted position; According to the Gaussian model, the covariance matrix between known points is calculated to solve the Kriging weights, and the estimated values ​​of the unknown points are calculated using the obtained Kriging weights: in, is the interpolation result of the predicted position x0.

6. The carbon emission prediction method based on power big data according to claim 3 is characterized in that: The S2.5 is as follows: Identify the causal relationships that need to be analyzed, obtain electricity data indicators that have an impact on carbon emissions greater than the threshold, and make assumptions based on theoretical frameworks and domain knowledge; List all relevant observed and latent variables and hypothesize the structural relationships between them; Construct causal path diagrams based on prior knowledge to indicate assumed causal relationships. Such as electricity consumption and how weather changes affect carbon emissions. The variables are connected through directed paths, representing the causal hypothesis relationship, and are divided into a structural part and a measurement part. Build the structural model: η=Bη+Γξ+ζ; Among them, η is the endogenous variable matrix, B is the path coefficient matrix between endogenous variables; ξ is the exogenous latent variable, Γ is the path coefficient matrix of the exogenous variable to the endogenous variable, and ζ is the error term; The relationship between latent variables and observed variables is described by the measurement model: y=Λ y n+∈ x=Λ x ξ+δ; Where y is the actual measured vector, and x is the external direct measurement vector; Λ y is the latent factor loading matrix, Λ x is the latent factor loading matrix, ∈ and δ are noise vectors; Through the size and significance test of the path coefficient, the strength and direction of the causal relationship between the variables can be interpreted; Key causal features are extracted from the structural model for subsequent carbon emission prediction models, including correlation coefficients in significant paths and explanatory components of latent variables.

7. The carbon emission prediction method based on power big data according to claim 1 is characterized in that: The carbon emission prediction model is as follows: Ridge regression was used as the base model to capture simple linear relationships; J(β)=||Y-Xβ|| 2 +λ||β|| 2 ; Among them, Y is the target variable vector, X is the comprehensive feature matrix, β is the coefficient vector, and λ is the regularization parameter; The long short-term memory network LSTM is introduced to process long time series data and capture the time dependency therein; the long short-term memory network LSTM specifically includes: Input layer: x i ; Forget gate: f i =σ(W f ·|h i-1 ,x i |+b f ); Input gate: I i =σ(W I ·|h i-1 ,x i |+b I ); Candidate values: Cell status update: Output gate: o t =σ(W o ·|h i-1 ,x i |+b0); Hide status update:h t =o i tanh(C i ); Output layer: y t ; Among them, x i is the input feature vector of time step i; h i-1 is the hidden state at time step i; f i For the Forget Gate; i is the input gate; is the candidate value; that is, the candidate memory cell at the current moment; C t is the cell state; it indicates the memory cell state at the current moment; o t is the output gate; h t is the hidden state at the current moment; W f , W I , W C , W o are the weight matrices of the forget gate, input gate, candidate value, and output gate respectively; b f 、b I 、b C , b0 are the bias vectors of the forget gate, input gate, candidate value and output gate respectively; Based on the gradient boosting tree XGBoost, the prediction results of different models are integrated to build a carbon emission prediction model: AND final＝ XGBoost(Y ridge ,AND LSTM ); Among them, Y ridge is the output of ridge regression; Y LSTM Output of the LSTM model.

8. A carbon emission prediction system based on power big data, characterized in that: It includes a processor, a memory and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the carbon emission prediction method based on power big data as described in any one of claims 1 to 7.

Images