Dynamic association method and system for power consumption and average carbon column concentration of urban department

Through random forest and XGBoost, the correlation model between power consumption and carbon column concentration in urban departments is constructed, and the data integration and model accuracy problems in carbon emission monitoring are solved, and the high-precision timing correlation analysis is realized, providing scientific pollution reduction and reduction Carbon guidance.

CN119066627AInactive Publication Date: 2024-12-03STATE GRID SICHUAN ELECTRIC POWER CORP ELECTRIC POWER RES INST
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411255984.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-09
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art has problems in the monitoring and analysis of carbon emissions, huge data volume and difficult processing, model accuracy and reliability depend on data quality, and lacks effective instructions and methods to guide urban departments to reduce pollution and carbon reduction work.

Method used

Random forest and XGBoost methods are used to construct a high-precision correlation model of urban sector power consumption and average carbon column concentration. Through multi-source satellite data fusion, time rolling analysis and stacked regression models, importance sorting, feature extraction and model fusion are selected to ensure that Reliability of timing correlation analysis.

Benefits of technology

Achieve high-precision time rolling window technology application, improve the comprehensive prediction capability of the model, reveal the dynamic relationship between power consumption and carbon emissions, and provide scientific analysis tools for pollution reduction and carbon reduction work, improving data accuracy and reliability sex.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066627A_ABST
    Figure CN119066627A_ABST
Patent Text Reader

Abstract

The invention discloses a dynamic correlation method and system for power consumption and average carbon column concentration of urban departments, and relates to the field of carbon emission monitoring and analysis, and the key points of the technical scheme are that the method comprises the following steps: constructing an initial correlation model for power consumption and average carbon column concentration of urban sub-departments by adopting a random forest and an XGBoost method; screening out power consumption data of a plurality of departments through the initial correlation model; analyzing the correlation degree between the screened department power consumption data and the high-resolution XCO2 data; and establishing a change trend chart of different time periods according to the change condition of the correlation degree. According to the method, the random forest and the XGBoost method are adopted to construct the high-precision relevance model, the method plays a crucial role in time scrolling analysis, a high-precision basic prediction result is provided for application of the time scrolling window technology, and the reliability of time sequence relevance analysis is ensured through feature extraction, refined prediction and model fusion.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of carbon emission monitoring and analysis, and more specifically, it relates to a method and system for dynamically correlating urban sector electricity consumption with average carbon column concentration. Background Art

[0002] Global climate change has become one of the most serious environmental problems facing humanity. The massive emission of greenhouse gases such as carbon dioxide (CO2) is the main cause of global warming, which has a profound impact on ecosystems, human health and economic activities.

[0003] At present, relevant research at home and abroad mainly focuses on the following three aspects:

[0004] 1) Use multi-source data fusion technology to provide high-resolution carbon emission monitoring: Use multi-source data fusion technology such as satellite remote sensing, drone monitoring, and ground sensor networks to provide detailed carbon emission spatial and temporal resolution monitoring. These technologies help identify the main sources and hotspots of carbon emissions, but due to the inconsistent temporal and spatial resolutions between different data sources, data integration is difficult and the ability to provide guidance for pollution reduction and carbon reduction is weak. In addition, the amount of high-resolution data is huge, and it is very difficult to process and analyze this data.

[0005] 2) Constructing economic-environmental models to reveal the relationship between carbon emissions and economic activities: Some studies have attempted to reveal the relationship between carbon emissions and economic activities by constructing economic-environmental models and analyzing the impact of different economic activities on carbon emissions. For example, by analyzing data from various aspects such as industrial production, transportation, and building energy consumption, a connection model between carbon emissions and economic activities is established to evaluate the emission reduction effects of different policy measures. However, in actual applications, due to insufficient data, the evaluation of the actual application effect of the model is insufficient, which affects the systematization and refinement of pollution reduction and carbon reduction work.

[0006] 3) Use machine learning and big data technologies to develop intelligent carbon emission analysis and prediction models: Although machine learning and big data technologies can provide powerful analytical capabilities, through the analysis of massive data, potential carbon emission patterns and trends can be discovered, providing a scientific basis for decision makers. However, the accuracy and reliability of the model depend on the data quality and model performance, resulting in the lack of a technology that can provide effective instructions and guidance for pollution reduction and carbon reduction work in various industries and departments at this stage.

[0007] Therefore, how to study and design a method and system for dynamically correlating urban sector electricity consumption with average carbon column concentration that can overcome the above-mentioned defects is a problem that we urgently need to solve. Summary of the invention

[0008] In order to address the deficiencies in the prior art, the purpose of the present invention is to provide a method and system for dynamically correlating urban sector electricity consumption with average carbon column concentration. Random forest and XGBoost methods are used to construct a high-precision correlation model between urban sector electricity consumption and its average carbon column concentration, which plays a vital role in time rolling analysis. It not only provides high-precision basic prediction results for the application of time rolling window technology, but also ensures the reliability of time series correlation analysis through feature extraction, refined prediction and model fusion.

[0009] The above technical objectives of the present invention are achieved through the following technical solutions:

[0010] First, a method for dynamically associating urban sector electricity consumption with average carbon column concentration is provided, including the following steps:

[0011] Collect data on electricity consumption by sector and average carbon intensity in urban areas;

[0012] The average carbon concentration data of urban areas is reconstructed using multi-source satellite data fusion technology to obtain high-resolution XCO2 data;

[0013] The random forest and XGBoost methods were used to construct an initial correlation model between electricity consumption and average carbon column concentration in different urban sectors;

[0014] Input the electricity consumption data of each department and the corresponding high-resolution XCO2 data into the initial correlation model to screen out the electricity consumption data of multiple departments ranked top in importance;

[0015] The time rolling analysis method is used to analyze the correlation between the selected departmental electricity consumption data and the corresponding high-resolution XCO2 data;

[0016] According to the changes in the correlation between electricity consumption of each department and the average carbon column concentration, a trend chart of changes in different time periods is established.

[0017] Furthermore, the process of reconstructing the average carbon concentration data of urban areas by using multi-source satellite data fusion technology is specifically as follows:

[0018] Based on the XGBoost machine learning algorithm, the average mole fraction data of CO2 dry air columns from multiple satellite sources are fused and filled to reconstruct the monthly spatial and temporal distribution of XCO2 concentrations in the 1km grid of the corresponding city;

[0019] Among them, the multi-source satellites include the Greenhouse Gas Observation Satellite GOSAT, the Orbiting Carbon Observation Satellite OCO-2 and the Orbiting Carbon Observation Satellite OCO-3.

[0020] Furthermore, the construction process of the initial association model is specifically as follows:

[0021] Build random forest model and XGBoost model respectively;

[0022] The random forest model and the XGBoost model were fused using the stacked regression method to obtain the initial association model;

[0023] Among them, the random forest model and the XGBoost model are both used as base learners. The stacked regression method takes the prediction results of multiple base learners as input and then trains a new regressor to output the final prediction results.

[0024] Furthermore, the process of analyzing the correlation between the screened departmental power consumption data and the corresponding high-resolution XCO2 data using the time rolling analysis method is specifically as follows:

[0025] Divide the departmental electricity consumption data and the corresponding high-resolution XCO2 data into multiple fixed-size time rolling windows in chronological order;

[0026] Calculates the relevance based on the data in a time rolling window.

[0027] Furthermore, the data in the time rolling window is specifically:

[0028] The departmental electricity consumption data and the corresponding high-resolution XCO2 data are expressed as time series data:

[0029]

[0030] Where D represents time series data; x t The feature vector representing time t corresponds to the department's electricity consumption data; y t represents the target variable at time t, corresponding to the high-resolution XCO2 data; T represents the sequence period of the time series data;

[0031] Then the data of the i-th time rolling window is:

[0032]

[0033] Among them, D i represents the data of the i-th time rolling window; w represents the window width of the time rolling window.

[0034] Furthermore, the analysis process of the correlation degree is specifically as follows:

[0035] Calculate the covariance between the departmental electricity consumption data and the corresponding high-resolution XCO2 data within the time rolling window;

[0036] Calculate the standard deviation of the department power consumption data and the high-resolution XCO2 data in the time rolling window respectively, and calculate the first parameter by the product of the two standard deviations;

[0037] The corresponding correlation degree in the time rolling window is calculated by the product of the covariance and the first parameter.

[0038] Furthermore, the process of establishing the change trend graph is specifically as follows:

[0039] Select the correlation of the electricity consumption data of the first m departments, and the correlation within each time rolling window is expressed as:

[0040]

[0041] in, represents the jth department at time t k The amount of electricity consumed; Indicates that at time t k The average carbon column concentration of Indicates that at time t k The correlation between the electricity consumption of the jth sector and the average carbon column concentration under ;

[0042] A change trend graph is established with time as the horizontal axis and the correlation degree as the vertical axis, and each value of m in the change trend graph corresponds to a trend line.

[0043] In a second aspect, a system for dynamically associating urban sector electricity consumption with average carbon column concentration is provided, the system being used to implement a method for dynamically associating urban sector electricity consumption with average carbon column concentration as described in any one of the first aspects, comprising:

[0044] Data collection module, used to collect electricity consumption data of various departments and average carbon concentration data of urban areas;

[0045] The data reconstruction module is used to reconstruct the average carbon concentration data of urban areas using multi-source satellite data fusion technology to obtain high-resolution XCO2 data;

[0046] The model building module is used to build an initial correlation model between electricity consumption and average carbon column concentration in different urban sectors using random forest and XGBoost methods;

[0047] The data screening module is used to input the power consumption data of each department and the corresponding high-resolution XCO2 data into the initial correlation model, and screen out the power consumption data of multiple departments ranked top in importance;

[0048] A correlation analysis module, used to analyze the correlation between the screened departmental power consumption data and the corresponding high-resolution XCO2 data using a time rolling analysis method;

[0049] The trend analysis module is used to establish a trend chart of changes in different time periods based on the changes in the correlation between the electricity consumption of each department and the average carbon column concentration.

[0050] In a third aspect, a computer terminal is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a method for dynamically associating urban sector electricity consumption with average carbon column concentration as described in any one of the first aspects is implemented.

[0051] In a fourth aspect, a computer-readable medium is provided, on which a computer program is stored, and the computer program is executed by a processor to implement the method for dynamically associating urban sector electricity consumption with average carbon column concentration as described in any one of the first aspects.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. The dynamic correlation method between urban sector electricity consumption and average carbon column concentration provided by the present invention adopts random forest and XGBoost methods to construct a high-precision correlation model between urban sector electricity consumption and its average carbon column concentration, which plays a vital role in time rolling analysis. It not only provides high-precision basic prediction results for the application of time rolling window technology, but also ensures the reliability of time series correlation analysis through feature extraction, refined prediction and model fusion;

[0054] 2. The stacked regression model combines the advantages of random forest and XGBoost, significantly improving the comprehensive prediction ability of the model. By integrating the prediction results of the two models, the stacked regression model performs well in capturing complex correlations and reducing prediction errors. It not only performs well in static analysis, but also provides strong support for time series correlation analysis.

[0055] 3. The present invention dynamically calculates the correlation within each time rolling window, thereby revealing the time series changes of electricity consumption of each department on carbon emissions. The fusion effect of stacked regression ensures that the calculation results of the correlation of each department in the time rolling analysis are more stable and accurate, fully reflecting the model's ability to predict carbon emissions in different periods, and can deeply reveal the dynamic relationship between electricity consumption and carbon emissions in different time periods, providing an innovative analysis tool for the scientific guidance of pollution reduction and carbon reduction work;

[0056] 4. The present invention utilizes multi-source satellite data fusion technology to reconstruct high-resolution XCO2 data, thereby improving the accuracy and reliability of the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0058] Figure 1 is a flow chart of Embodiment 1 of the present invention;

[0059] Figure 2 is the monthly average spatial distribution diagram of XCO2 in Example 1 of the present invention;

[0060] Figure 3 is a comparison chart of the actual and model-predicted carbon column concentrations in different cities in Example 1 of the present invention;

[0061] Figure 4 is a graph showing the changing trend of the correlation between the power consumption industries in different cities and the average carbon column concentration in Example 1 of the present invention;

[0062] Figure 5 It is a system block diagram in Example 2 of the present invention. DETAILED DESCRIPTION

[0063] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0064] Example 1: Dynamic correlation method between urban sector electricity consumption and average carbon column concentration, such as Figure 1 As shown, the following steps are included:

[0065] S1: Collect electricity consumption data of various departments and average carbon concentration data of urban areas;

[0066] S2: Use multi-source satellite data fusion technology to reconstruct the average carbon concentration data of urban areas to obtain high-resolution XCO2 data;

[0067] S3: Random forest and XGBoost methods were used to construct an initial correlation model between electricity consumption and average carbon column concentration in different urban sectors;

[0068] S4: Inputting the electricity consumption data of each department and the corresponding high-resolution XCO2 data into the initial correlation model, and screening out multiple departmental electricity consumption data ranked at the top of importance;

[0069] S5: Analyze the correlation between the selected departmental electricity consumption data and the corresponding high-resolution XCO2 data using a time rolling analysis method;

[0070] S6: Establish a trend chart of changes in different time periods based on the changes in the correlation between electricity consumption of each department and the average carbon column concentration.

[0071] In step S1, the departmental electricity consumption data can be classified according to various industries and / or sectors. For example, the present invention collects classified data on total electricity consumption in six cities from January 2017 to December 2021. The data set is extensive and detailed, covering various industries and sectors, and providing rich information, as shown in Table 1. These industries and sectors include traditional primary industries such as agriculture, forestry, animal husbandry and fisheries, as well as resource-intensive industries such as coal mining, oil and gas mining, and metal mining. In addition, it also includes high-energy consumption sectors in the manufacturing industry, such as food processing, textiles, chemical products, metal smelting and rolling, machinery manufacturing and electronic equipment manufacturing, as well as energy supply industries such as electricity and heat production and supply. The data set also comprehensively reflects the electricity consumption in important areas such as construction, transportation, and communication services, as well as the electricity consumption of public services and social service departments such as water management, education, health care, and public administration. This comprehensive data set supports the study of the relationship between electricity consumption and average carbon column concentration in various urban departments.

[0072] Through in-depth analysis of these electricity consumption data, we explore the dynamic temporal relationship between electricity consumption in various urban sectors and their average carbon column concentration. This will help us understand the impact of electricity consumption in different sectors on the city's overall carbon emissions and provide a scientific basis for formulating effective carbon reduction strategies.

[0073] Table 1 Names and labels of electricity consumption classification data

[0074] Label Industry Name Label Industry Name 1 agriculture 34 Special equipment manufacturing 2 forestry 35 Automotive 3 Animal Husbandry 36 Electrical machinery and equipment manufacturing 4 Fisheries 37 Computer, communications and other electronic equipment manufacturing 5 Agriculture, forestry, animal husbandry, fishery and auxiliary activities 38 Instrument manufacturing 6 Coal mining and washing industry 39 Other manufacturing industries 7 Oil and gas extraction industry 40 Waste resource comprehensive utilization industry 8 Ferrous metal mining and dressing industry 41 Metal products, machinery and equipment repair 9 Non-ferrous metal mining and dressing industry 42 Electricity, heat production and supply industry 10 Non-metallic mineral mining and dressing industry 43 Gas production and supply industry 11 Other mining industries 44 Water production and supply industry 12 Agricultural and sideline food processing industry 45 Residential Construction Industry 13 Food Manufacturing 46 Civil Engineering and Construction 14 Beverage manufacturing and refined tea manufacturing 47 Construction and installation industry 15 Tobacco products industry 48 Building decoration, renovation and other construction industries 16 Textile Industry 49 Railway transportation industry 17 Textile, clothing and apparel industry 50 Road transport industry 18 Leather, fur, feathers and their products and footwear 51 Water transport industry 19 Wood processing and wood, bamboo, rattan, palm and grass products industry 52 Air transport industry 20 Furniture Manufacturing 53 Pipeline transportation industry 21 Paper and paper products industry 54 Multimodal transport and transport forwarding 22 Printing and recorded media reproduction industry 55 Stevedoring and warehousing 23 Cultural, educational, arts and crafts, sports and entertainment products manufacturing 56 Postal Industry 24 Oil processing, coking and nuclear fuel processing industries 57 Telecommunications, broadcasting and satellite transmission services 25 Chemical raw materials and chemical products manufacturing industry 58 Internet and related services 26 Pharmaceutical Manufacturing 59 Software and information technology services 27 Chemical fiber manufacturing industry 60 Scientific research and technical services 28 Rubber and plastic products industry 61 Water conservancy, environment and public facilities management industry 29 Non-metallic mineral products industry 62 Residential services, repairs and other services 30 Ferrous metal smelting and rolling processing industry 63 Education, culture, sports and entertainment 31 Non-ferrous metal smelting and rolling processing industry 64 Health and Social Work 32 Metal products industry 65 Public administration, social organizations and international organizations 33 General equipment manufacturing industry

[0075] In step S2, the present invention uses the XGBoost machine learning algorithm to fuse and fill the CO2 dry air column average mole fraction data from multiple source satellites to reconstruct the monthly spatial and temporal distribution of the 1km grid XCO2 concentration of the corresponding city, such as Figure 2 Among them, the multi-source satellites include the Greenhouse Gas Observation Satellite GOSAT, the Orbiting Carbon Observation Satellite OCO-2 and the Orbiting Carbon Observation Satellite OCO-3.

[0076] In step S3, the construction process of the initial association model is specifically as follows: construct a random forest model and an XGBoost model respectively; use a stacked regression method to fuse the random forest model and the XGBoost model to obtain an initial association model; wherein the random forest model and the XGBoost model are both used as base learners, and the stacked regression method uses the prediction results of multiple base learners as input, and then trains a new regressor to output the final prediction result.

[0077] Random Forest is an ensemble learning method based on decision trees. It improves the accuracy and stability of the model by building multiple decision trees and averaging their prediction results (regression problems) or voting (classification problems). Random Forest uses "bagging" and random feature selection to enable each tree to make high-variance and low-bias predictions, thereby reducing the risk of overfitting.

[0078] A random forest consists of multiple decision trees, each of which is trained on a subsample of the training data. The model's prediction result is the average of all tree prediction results (regression) or the majority vote result (classification), as follows:

[0079]

[0080] in, is the final prediction result of the random forest, N is the number of decision trees, is the prediction result of the i-th tree.

[0081] XGBoost (Extreme Gradient Boosting) is an efficient implementation based on Gradient Boosted Decision Trees (GBDT), which is popular for its advantages in processing large-scale data and providing high accuracy. XGBoost gradually optimizes the loss function so that each newly added tree can compensate for the prediction error of the previous tree, thereby improving the performance of the overall model.

[0082] The XGBoost model iteratively optimizes the model's prediction results by adding new trees. The goal of each new tree is to minimize the residual (prediction error) of all previous trees, as follows:

[0083]

[0084] in, is the prediction result of round t, f t (x) is the prediction result of the tth tree. The loss function is optimized by the following formula:

[0085]

[0086] Among them, L (t) is the objective function of the tth round, is the loss function, Ω(f k ) is the regularization term, and n is the number of samples.

[0087] In order to further improve the prediction performance of the model, the present invention adopts the stacking regressor method to fuse the random forest and XGBoost models. Stacking regression takes the prediction results of multiple base learners as input and then trains a new regressor to output the final prediction result. The specific expression is:

[0088]

[0089] Among them, stacked_model is a correlation model, estimators is a list of base learners, including the best random forest model best_rf and the best XGBoost model best_xgb, and final_estimator is the final regressor. The present invention selects Ridge Regression (RidgeCV).

[0090] In step S5, the process of using the time rolling analysis method to analyze the correlation between the screened department power consumption data and the corresponding high-resolution XCO2 data is specifically as follows: the department power consumption data and the corresponding high-resolution XCO2 data are divided into multiple fixed-size time rolling windows in chronological order; and the correlation is calculated based on the data in the time rolling window.

[0091] The departmental electricity consumption data and the corresponding high-resolution XCO2 data are expressed as time series data:

[0092]

[0093] Where D represents time series data; x t The feature vector representing time t corresponds to the department's electricity consumption data; y t represents the target variable at time t, corresponding to the high-resolution XCO2 data; T represents the sequence period of the time series data.

[0094] Then the data of the i-th time rolling window is:

[0095]

[0096] Among them, D i represents the data of the i-th time rolling window; w represents the window width of the time rolling window.

[0097] The analysis process of the correlation is specifically as follows: calculating the covariance between the department power consumption data and the corresponding high-resolution XCO2 data within the time rolling window; calculating the standard deviation of the department power consumption data and the high-resolution XCO2 data within the time rolling window respectively, and calculating the first parameter by the product of the two standard deviations; calculating the corresponding correlation within the time rolling window by the product of the covariance and the first parameter.

[0098] Specifically, the correlation calculation formula is:

[0099]

[0100] Among them, corr(x,y) is the correlation within the time rolling window, cov(x,y) is the covariance, σ x and σ y are the standard deviations of the feature and target variables, respectively.

[0101] For each time rolling window, extract the corresponding feature matrix x window and the target variable y window , calculate the correlation between electricity consumption of various departments in the city and the average carbon column concentration.

[0102] In step S6, the present invention can obtain the change in the correlation between the power consumption of each department in the city and the average carbon column concentration every month through the time rolling window method. Select several power consumption departments with the highest correlation every month, and analyze the change trends in different time periods to better understand and reveal the potential impact of power consumption of each department on the average carbon column concentration.

[0103] Select the correlation of the electricity consumption data of the first m departments, and the correlation within each time rolling window is expressed as:

[0104]

[0105] in, represents the jth department at time t k The amount of electricity consumed; Indicates that at time t k The average carbon column concentration of Indicates that at time t k The correlation between the electricity consumption of the jth sector and the average carbon column concentration under .

[0106] A change trend graph is established with time as the horizontal axis and the correlation degree as the vertical axis, and each value of m in the change trend graph corresponds to a trend line.

[0107] Experimental verification and analysis

[0108] 1. Model training and performance evaluation

[0109] The training process of the correlation model between urban sub-sector electricity consumption and its average carbon column concentration consists of two steps: first, training the base learners (random forest and XGBoost), and then using the outputs of these base learners as new features to train the final regressor (ridge regression). The specific steps are as follows:

[0110] 1) Base learner training: Cross-validate the training data to obtain the prediction results of each base learner.

[0111] 2) Final regressor training: Use the prediction results of the base learner as input features to train the final regressor.

[0112]

[0113] in, is the final prediction result of the initial association model, and They are the prediction results of random forest and XGBoost base learners respectively.

[0114] The model performance was measured by the coefficient of determination (R 2 ) and root mean square error (RMSE)

[32] :

[0115]

[0116]

[0117] Among them, y i is the true value, is the predicted value, y is the mean of the true values, and n is the number of samples.

[0118] 2. Model parameter configuration

[0119] 1) The main hyperparameters of random forest include:

[0120] n_estimators: represents the number of decision trees built in the random forest. More trees generally improve the performance of the model, but also increase the computational overhead.

[0121] max_depth: controls the maximum depth of the tree. Setting it to None means that the tree will expand until all leaf nodes are pure or the number of samples contained in the leaf nodes is less than min_samples_split.

[0122] min_samples_split: controls the minimum number of samples a node needs before it is split. Larger values ​​can prevent the model from overfitting.

[0123] min_samples_leaf: controls the minimum number of samples required for a leaf node. Larger values ​​prevent the model from learning noise in the training data.

[0124] Randomized Search Cross Validation (RandomizedSearchCV)

[33] is used for hyperparameter adjustment. The specific settings of the hyperparameter search space are shown in Table 2.

[0125] Table 2 Main hyperparameters of random forest

[0126] Hyperparameters Search Space n_estimators [100,200,500] max_depth [None,10,20,30] min_samples_split [2,5,10] min_samples_leaf [1,2,4]

[0127] 2) The main hyperparameters of XGBoost include:

[0128] n_estimators: represents the number of boosted trees. More trees can improve the performance of the model, but it will also increase the training time.

[0129] max_depth: controls the maximum depth of the tree. A larger depth may cause the model to overfit, while a smaller depth may cause underfitting.

[0130] learning_rate: controls the learning rate of each boost. A smaller learning rate can make the model more robust, but requires more boosted trees.

[0131] subsample: indicates the sample ratio used to train each tree. The value is between (0,1] to prevent overfitting.

[0132] colsample_bytree: indicates the feature ratio used to build each tree. The value is between (0,1] to prevent overfitting.

[0133] Randomized Search Cross Validation (RandomizedSearchCV) is used for hyperparameter adjustment. The specific settings of the hyperparameter search space are shown in Table 3.

[0134] Table 3 Main hyperparameters of XGBoost

[0135]

[0136]

[0137] By tuning these hyperparameters, we can find the best model configuration and improve the predictive performance of the model.

[0138] 3. Performance Verification of Different Urban Models

[0139] The prediction performance of the correlation model (initial correlation model) varies among cities, as shown in Table 4, but the overall performance is good. In some cities, such as City 1, City 2, City 3, and City 4, the R 2 The values ​​are relatively high, namely 0.8528, 0.8068, 0.8327 and 0.7817, which indicates that the model has a strong ability to explain the relationship between electricity consumption and carbon concentration in these cities. In addition, the RMSE values ​​are low, namely 1.4939, 1.7246, 1.5082 and 1.9029, which indicates that the prediction error is small.

[0140] However, for cities 5 and 6, the R 2 The values ​​are relatively low, 0.7352 and 0.7389, respectively, and the prediction errors are large, 1.9446 and 2.0167, respectively. This may be due to the more complex relationship between electricity consumption and carbon concentration in these cities, or some noise in the data. For these cities, it may be necessary to further optimize the model or introduce additional features.

[0141] Overall, through the performance verification of the correlation model, it is proved that the constructed model can fully predict the relationship between electricity consumption and average carbon column concentration in most cities, verifying the effectiveness and reliability of the model. This lays the foundation for further revealing the dynamic time correlation between electricity consumption in major industries and urban carbon concentration.

[0142] Table 4 Performance test verification of the initial correlation model in different cities

[0143] City <![CDATA[R 2 ]]> RMSE 1 0.8528 1.4939 2 0.8068 1.7246 3 0.8327 1.5082 4 0.7817 1.9029 5 0.7352 1.9446 6 0.7389 2.0167

[0144] The present invention compares the actual carbon column concentration of six cities (Cities 1-6) with the carbon column concentration predicted by the model. Figure 3 Overall, the model's predicted values ​​(red line) are relatively close to the actual values ​​(blue line) in most time periods, which shows that the model can effectively capture the relationship between electricity consumption and average carbon column concentration in various urban industries.

[0145] Specifically: For City 1, the model performed well in capturing the trend of carbon column concentration, and although there were slight deviations in some short-term fluctuations, the overall trend was consistent with the actual values. For City 2, the model accurately predicted the peaks and troughs in 2019 and 2020, but performed slightly worse during periods of greater fluctuations. For City 3, despite slight differences in some short-term fluctuations, the model captured the long-term trend well. For City 4, the model performed well on the data for 2020 and 2021, with a high degree of consistency with the actual values, showing strong predictive ability. For City 5, the model performed well overall, although there were slight deviations during peak periods. For City 6, the model effectively captured the overall trend of carbon concentration, although there were slight deficiencies in the prediction of some trough periods.

[0146] Through the analysis of these six cities, it can be seen that the model can predict the relationship between electricity consumption and average carbon column concentration well in most time periods. This verifies the effectiveness and reliability of the correlation model. Although there are some shortcomings in the prediction of certain periods and cities, the model generally performs well in reflecting the relationship between electricity consumption and carbon column concentration.

[0147] In order to deeply analyze the temporal changes in the correlation between electricity consumption of various industries and the average carbon column concentration of the city, the present invention adopts a time rolling analysis method, which dynamically calculates the potential impact of electricity consumption of different industries on the change of average carbon column concentration in different periods.

[0148] Time trends of the top three industries consuming the most electricity in each city, such as Figure 4 The following is a detailed analysis:

[0149] 1) Time correlation trend of electricity consumption industries and average carbon column concentration (top 3) in city 1

[0150] In City 1, the coal mining and washing industry (label 6) showed a significantly high correlation with carbon concentration in 2018 and 2020, with a correlation exceeding 0.3. However, this correlation dropped significantly after 2021, which may reflect the effectiveness of pollution reduction and carbon emission reduction measures in this industry. At the same time, the civil engineering construction industry (label 46) suddenly increased its correlation after 2021, indicating that this industry may have made a significant contribution to the increase in carbon emissions during this period, which deserves further attention.

[0151] 2) Time correlation trend between electricity consumption industries and average carbon column concentration (top 3) in city 2

[0152] In City 2, the real estate industry (label 45) shows a high correlation of more than 0.25 in 2018 and 2021. However, the correlation drops significantly after 2022, indicating that the industry may have successfully reduced its impact on carbon intensity through effective emission reduction measures. On the other hand, the Internet and related service industries (label 58) have a sudden increase in correlation between 2019 and 2020, which may indicate an increase in carbon emissions in these years, possibly due to the expansion of data centers and other infrastructure. Further investigation and regulation are needed.

[0153] 3) Time correlation trend between electricity consumption industries and average carbon column concentration in City 3 (top 3)

[0154] In City 3, the chemical raw materials and chemical products manufacturing industry (label 25) showed a high correlation in 2018 and 2020, but this correlation decreased after 2021, indicating that emission reduction measures in this industry may have begun to take effect. In contrast, the textile industry (label 16) showed a sudden increase in correlation in 2021, which may indicate that the impact of carbon emissions during this period increased and that stronger regulation was needed.

[0155] 4) Temporal correlation trend between electricity consumption industries and average carbon column concentration (top 3) in City 4

[0156] In City 4, the coal mining and washing industry (Tag 6) showed a high correlation in 2018 and 2021, but this correlation declined after 2022, possibly reflecting progress in carbon emission reduction. However, the electrical machinery and equipment manufacturing industry (Tag 36) suddenly increased its correlation after 2020, indicating that this industry may become a significant contributor to carbon emissions. This trend emphasizes the importance of implementing stricter control measures and promoting energy efficiency in this industry.

[0157] 5) Time correlation trend between electricity consumption industries and average carbon column concentration in the top 5 cities (top 3)

[0158] In City 5, the fishery industry (label 4) showed extremely high correlation in early 2018 and 2021, and then declined, indicating that pollution reduction and carbon reduction efforts may have been effective. At the same time, the instrument manufacturing industry (label 38) showed a significant increase in correlation in 2021, indicating that the industry may contribute more to carbon emissions during this period and needs to be closely monitored.

[0159] 6) Time correlation trend between electricity consumption industries and average carbon column concentration (top 3) in cities 6

[0160] In city 6, the correlation of coal mining and washing (label 6) was high in 2018 and 2020, and decreased after 2021, which may indicate the effectiveness of carbon reduction measures. However, the correlation of agriculture, forestry, animal husbandry and fishery (labels 1, 2, 3, 4) suddenly increased in 2021, indicating that these industries had a greater impact on carbon emissions during this period. This trend needs further investigation to determine the direction of improvement in pollution control.

[0161] By conducting a detailed analysis of the time trends of the top three industries with the most electricity consumption in each city, the present invention not only identifies industries that are likely to successfully reduce pollution and carbon emissions, but also highlights industries where the correlation suddenly increases in certain years. This information is essential for accurately evaluating and guiding pollution reduction and carbon reduction efforts in various industries, and helps to develop more effective urban carbon emission control strategies.

[0162] Example 2: A system for dynamically associating urban sector electricity consumption with average carbon column concentration, the system is used to implement the method for dynamically associating urban sector electricity consumption with average carbon column concentration as described in Example 1, such as Figure 5 As shown, it includes a data acquisition module, a data reconstruction module, a model building module, a data screening module, a correlation analysis module and a trend analysis module.

[0163] Among them, the data acquisition module is used to collect the electricity consumption data of each department and the average carbon concentration data of the urban area; the data reconstruction module is used to reconstruct the average carbon concentration data of the urban area by using multi-source satellite data fusion technology to obtain high-resolution XCO2 data; the model construction module is used to use random forest and XGBoost methods to construct an initial correlation model between the electricity consumption of each department and the average carbon column concentration in the city; the data screening module is used to input the electricity consumption data of each department and the corresponding high-resolution XCO2 data into the initial correlation model to screen out multiple departmental electricity consumption data with top importance rankings; the correlation analysis module is used to analyze the correlation between the screened departmental electricity consumption data and the corresponding high-resolution XCO2 data by time rolling analysis method; the trend analysis module is used to establish a trend chart of changes in different time periods according to the changes in the correlation between the electricity consumption of each department and the average carbon column concentration.

[0164] The present invention also records a computer terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for dynamically associating urban sector electricity consumption with average carbon column concentration as described in Example 1 is implemented.

[0165] The present invention also records a computer-readable medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the method for dynamically associating urban sector electricity consumption with average carbon column concentration as described in Example 1.

[0166] Working principle: The present invention adopts random forest and XGBoost methods to construct a high-precision correlation model between urban sector electricity consumption and its average carbon column concentration, which plays a vital role in time rolling analysis. It not only provides high-precision basic prediction results for the application of time rolling window technology, but also ensures the reliability of time series correlation analysis through feature extraction, refined prediction and model fusion; in addition, the present invention combines the advantages of random forest and XGBoost with a stacked regression model, which significantly improves the comprehensive prediction ability of the model. By fusing the prediction results of the two models, the stacked regression model performs well in capturing complex correlations and reducing prediction errors. It not only performs well in static analysis, but also provides strong support for time series correlation analysis; in addition, the present invention dynamically calculates the correlation within each time rolling window, thereby revealing the time series changes of electricity consumption of each department to carbon emissions. The fusion effect of stacked regression ensures that the calculation results of the correlation of each department in the time rolling analysis are more stable and accurate, which fully reflects the model's prediction ability for carbon emissions in different periods, and can deeply reveal the dynamic relationship between electricity consumption and carbon emissions in different time periods, providing innovative analysis tools for scientific guidance of pollution reduction and carbon reduction.

[0167] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0168] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes 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 generate 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.

[0169] 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.

[0170] 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. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0171] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A dynamic correlation method between urban sector electricity consumption and average carbon column concentration, characterized by: The following steps are involved: Collect data on electricity consumption by sector and average carbon intensity in urban areas; The average carbon concentration data of urban areas is reconstructed using multi-source satellite data fusion technology to obtain high-resolution XCO2 data; The random forest and XGBoost methods were used to construct an initial correlation model between electricity consumption and average carbon column concentration in different urban sectors; Input the electricity consumption data of each department and the corresponding high-resolution XCO2 data into the initial correlation model to screen out the electricity consumption data of multiple departments ranked top in importance; The correlation between the selected departmental electricity consumption data and the corresponding high-resolution XCO2 data was analyzed using a time-rolling analysis method; According to the changes in the correlation between electricity consumption of each department and the average carbon column concentration, a trend chart of changes in different time periods is established.

2. The method for dynamically associating urban sector electricity consumption with average carbon column concentration according to claim 1 is characterized in that: The process of reconstructing the average carbon concentration data of urban areas by using multi-source satellite data fusion technology is specifically as follows: Based on the XGBoost machine learning algorithm, the average mole fraction data of CO2 dry air columns from multiple satellite sources are fused and filled to reconstruct the monthly spatial and temporal distribution of XCO2 concentrations in the 1km grid of the corresponding city; Among them, the multi-source satellites include the Greenhouse Gas Observation Satellite GOSAT, the Orbiting Carbon Observation Satellite OCO-2 and the Orbiting Carbon Observation Satellite OCO-3.

3. The method for dynamically associating urban sector electricity consumption with average carbon column concentration according to claim 1 is characterized in that: The construction process of the initial association model is specifically as follows: Build random forest model and XGBoost model respectively; The random forest model and the XGBoost model were fused using the stacked regression method to obtain the initial association model; Among them, the random forest model and the XGBoost model are both used as base learners. The stacked regression method takes the prediction results of multiple base learners as input and then trains a new regressor to output the final prediction results.

4. The method for dynamically associating urban sector electricity consumption with average carbon column concentration according to claim 1 is characterized in that: The process of analyzing the correlation between the selected departmental power consumption data and the corresponding high-resolution XCO2 data using the time rolling analysis method is specifically as follows: Divide the departmental electricity consumption data and the corresponding high-resolution XCO2 data into multiple fixed-size time rolling windows in chronological order; Calculates the relevance based on the data in a time rolling window.

5. The method for dynamically associating urban sector electricity consumption with average carbon column concentration according to claim 4 is characterized in that: The data in the time rolling window are specifically: The departmental electricity consumption data and the corresponding high-resolution XCO2 data are expressed as time series data: Where D represents time series data; x t The feature vector representing time t corresponds to the department's electricity consumption data; y t represents the target variable at time t, corresponding to the high-resolution XCO2 data; T represents the sequence period of the time series data; Then the data of the i-th time rolling window is: Among them, D i represents the data of the i-th time rolling window; w represents the window width of the time rolling window.

6. The method for dynamically associating urban sector electricity consumption with average carbon column concentration according to claim 4 is characterized in that: The analysis process of the correlation degree is specifically as follows: Calculate the covariance between the departmental electricity consumption data and the corresponding high-resolution XCO2 data within the time rolling window; Calculate the standard deviation of the department power consumption data and the high-resolution XCO2 data in the time rolling window respectively, and calculate the first parameter by the product of the two standard deviations; The corresponding correlation degree in the time rolling window is calculated by the product of the covariance and the first parameter.

7. The method for dynamically associating urban sector electricity consumption with average carbon column concentration according to claim 1 is characterized in that: The process of establishing the change trend graph is specifically as follows: Select the correlation of the electricity consumption data of the first m departments, and the correlation within each time rolling window is expressed as: in, represents the jth department at time t k The amount of electricity consumed; Indicates that at time t k The average carbon column concentration of Indicates that at time t k The correlation between the electricity consumption of the jth sector and the average carbon column concentration under ; A change trend graph is established with time as the horizontal axis and the correlation degree as the vertical axis, and each value of m in the change trend graph corresponds to a trend line.

8. A dynamic correlation system between urban sector electricity consumption and average carbon column concentration, characterized by: The system is used to implement the method for dynamically associating urban sector electricity consumption with average carbon column concentration as described in any one of claims 1 to 7, comprising: Data collection module, used to collect electricity consumption data of various departments and average carbon concentration data of urban areas; The data reconstruction module is used to reconstruct the average carbon concentration data of urban areas using multi-source satellite data fusion technology to obtain high-resolution XCO2 data; The model building module is used to build an initial correlation model between electricity consumption and average carbon column concentration in different urban sectors using random forest and XGBoost methods; The data screening module is used to input the power consumption data of each department and the corresponding high-resolution XCO2 data into the initial correlation model, and screen out the power consumption data of multiple departments ranked top in importance; A correlation analysis module, used to analyze the correlation between the screened departmental power consumption data and the corresponding high-resolution XCO2 data using a time rolling analysis method; The trend analysis module is used to establish a trend chart of changes in different time periods based on the changes in the correlation between the electricity consumption of each department and the average carbon column concentration.

9. A computer terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for dynamically associating urban sector electricity consumption with average carbon column concentration as described in any one of claims 1 to 7 is implemented.

10. A computer readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the method for dynamically associating urban sector electricity consumption with average carbon column concentration as described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Energy industry carbon emission prediction method based on electric power data

    CN114298381A

  • Regional carbon emission prediction method based on electric power data

    CN115907070A

  • Power carbon emission prediction method and device based on integrated learning module

    CN116108963A

  • Steel industry carbon emission prediction method and system based on Bayesian optimization

    CN117709524A

  • Forest carbon sink potential prediction method and system

    CN117909686A