Carbon emission trend prediction method based on polynomial regression and neural network

Abstract

The present application provides a kind of carbon emission trend prediction method based on polynomial regression and neural network, comprising: S11, respectively constructs the carbon emission model and flow chart model of whole life cycle;S12, data cleaning and pretreatment are carried out to carbon emission model and flow chart model, and high-quality data are output;S13, high-quality data are updated by dynamic updating mechanism;S14, a carbon emission prediction model is established using polynomial regression analysis method, carbon emission data are input into the carbon emission prediction model, and carbon emission trend prediction result is output.The present application can improve the accuracy and timeliness of high-carbon emission data, shorten the evaluation period, reduce the cost, and provide a strong basis for environmental regulation and protection decision.

Description

Technical Field

[0001] This invention relates to the field of carbon emission prediction technology, and in particular to a method for predicting carbon emission trends based on multinomial regression and neural networks. Background Technology

[0002] Currently, carbon emission quantification is mainly achieved through macro-statistical methods, life-cycle methods, and carbon emission measurement methods.

[0003] Macro-statistical method: This method involves joint efforts between the National Bureau of Statistics and the Ministry of Ecology and Environment, using software and databases to collect and analyze data. This data is then combined with energy consumption data from different industries and sectors to estimate carbon emissions. Basic carbon emission factors (such as CO2 emissions per unit of energy) are multiplied by activity data (such as electricity generation or fuel consumption) to arrive at the total carbon emissions. This method relies on national or regional economic activity statistics and energy consumption data, resulting in long timeframes, low accuracy, and limited real-time availability.

[0004] The life cycle approach encompasses a collection of environmental input and output data throughout the entire life cycle of electricity production and consumption. It provides comprehensive data support for carbon emission assessment of power systems and is one of the fundamental methods for calculating product carbon footprints. Professional life cycle assessment software such as GaBi and SimaPro is used for modeling and data analysis to ultimately obtain the carbon emissions generated throughout the process. However, existing life cycle databases have a large spatiotemporal span and are constantly changing, resulting in insufficient data representativeness and difficulties in assessing and controlling data quality. Traditional life cycle assessments use weighting factors to measure the severity of environmental damage represented by different indicators. Because these weighting factors include subjective factors, the objectivity and scientific rigor of life cycle assessments are not high.

[0005] Carbon emission measurement method: This method uses sensor technologies such as spectrometers and gas chromatographs, along with on-site data acquisition equipment, to monitor and record emission data in real time, directly measuring the CO2 concentration and emission rate of the emission source. This method is suitable for point source emissions such as factory chimneys or vehicle exhaust. Environmental protection departments or third-party environmental monitoring agencies use portable or fixed emission monitoring equipment for on-site testing. These devices can accurately measure and transmit data to a data center wirelessly or via wired means for further analysis and processing. This method requires highly skilled operators, has limited coverage, and incurs high costs.

[0006] In summary, carbon emission assessment of the power system is a crucial step. Traditional carbon metering methods are not precise enough, often calculating at the regional level and unsuitable for distribution networks and microgrid structures. This results in carbon emission data that fails to accurately reflect actual emissions and lacks timeliness. Existing methods fail to provide user-side carbon emission intensity data, cannot reflect carbon emissions during power network transmission, and methods based on historical data fail to capture the calculation characteristics of real-time carbon emission flows, and their reliable data sharing capabilities are insufficient. Therefore, improving the flow and sharing capabilities of carbon emission assessment data is key to promoting a precise comprehensive low-carbon assessment system for the entire lifecycle of the power system. Summary of the Invention

[0007] The purpose of this invention is to provide a carbon emission trend prediction method based on multinomial regression and neural networks, so as to improve the accuracy and timeliness of high carbon emission data, shorten the assessment cycle, reduce costs, and provide a strong basis for environmental supervision and protection decisions.

[0008] To achieve the above-mentioned objectives, the technical solution provided by this invention is as follows:

[0009] A carbon emission trend prediction method based on multinomial regression and neural networks, the method comprising:

[0010] S11. Construct a carbon emission model and a flowchart model for the entire life cycle, respectively;

[0011] S12. Perform data cleaning and preprocessing on the carbon emission model and flowchart model to output high-quality data;

[0012] S13. Update high-quality data through a dynamic update mechanism;

[0013] S14. Use the multinomial regression analysis method to establish a carbon emission prediction model, input carbon emission data into the carbon emission prediction model, and output the carbon emission trend prediction results.

[0014] Furthermore, constructing a full life-cycle carbon emission model includes the following operations:

[0015] S21. Integrate the data from the power grid database, power grid construction database, and high-carbon emission product database to achieve mutual correlation and supplementation of data in different databases and obtain integrated data.

[0016] S22. Perform data cleaning on the integrated data;

[0017] S23. Utilize construction process data, operation data, product design and usage data from the integrated data to establish a full life cycle carbon emission model.

[0018] Furthermore, a full lifecycle flowchart model is constructed, which specifically includes the following operations:

[0019] S31. Use the LCA basic database to obtain life cycle inventory data of high carbon emission products, and use literature databases to obtain historical carbon emission data, industry reports and scientific research information of high carbon emission products.

[0020] S32. Clean the acquired data and standardize the data format;

[0021] S33. Based on the cleaned data, clarify the scope of each life cycle stage of high-carbon emission products and construct a flowchart model of the entire life cycle.

[0022] Furthermore, data cleaning and preprocessing are performed on the carbon emission model and flowchart model, specifically including:

[0023] S41. Detect outliers in the carbon emission model and flowchart model data using the DBSCAN clustering algorithm;

[0024] S42. For time series data with missing values ​​in carbon emission models and flowchart models, estimate the missing values ​​using the prediction model, and fill the missing time series data with the estimated values ​​through multiple imputation.

[0025] S43. Perform consistency checks on the carbon emission model data and the flowchart model data.

[0026] Furthermore, the DBSCAN clustering algorithm is used to detect outliers in the carbon emission model and flowchart model data, specifically including the following operations:

[0027] S51. Input model data, radius parameter, and density threshold MinPts;

[0028] S52. Randomly select an unprocessed data object P from the model data, and determine whether the number of objects in its neighborhood meets the density threshold MinPts. If it does, mark P as a kernel object.

[0029] S53. Traverse the carbon emission model and flowchart model data, find all density-reachable objects originating from the kernel object P to form a new cluster, perform density-connection on all data objects within the cluster, and mark them as processed.

[0030] S54. Repeat steps S52 to S53 until all data objects in the model data have been processed.

[0031] Furthermore, consistency checks are performed on the carbon emission model and flowchart model data, specifically including the following operations:

[0032] S61. Define the quality standards for model data and determine the data quality indicators based on the quality standards.

[0033] S62. Based on data quality indicators, establish a data quality model using Apache Griffin;

[0034] S63. Run Apache Griffin to perform data quality checks on the model data, automatically identify data problems that do not meet quality standards, and generate and output data quality reports.

[0035] S64. Analyze the data quality report to identify data areas that need improvement.

[0036] Furthermore, a dynamic update mechanism is used to update high-quality data to enable real-time collection and updating of carbon emission data. This includes the following operations:

[0037] A sliding window mechanism is adopted, with a fixed-size window set, to continuously introduce new high-quality data and discard old data to update carbon emission data in real time;

[0038] Incremental gradient boosting trees are constructed using incremental learning algorithms to adjust the parameters of the carbon emission prediction model when data is updated.

[0039] Real-time acquisition and processing of carbon emission data is achieved through a streaming data processing framework and data stream pipeline.

[0040] Furthermore, the streaming data processing framework is Apache Flink, and the data stream pipeline is Apache Kafka.

[0041] Furthermore, a carbon emission prediction model is established using multinomial regression analysis. Carbon emission data is input into the model, and the carbon emission trend prediction results are output. Specifically, this includes the following operations:

[0042] S71. Construct a carbon emission prediction model based on the multinomial regression analysis method, input carbon emission data into the carbon emission prediction model, and output the basic trend of carbon emission changes over time.

[0043] S72. Create a Prophet model, train the Prophet model using historical datasets, evaluate the performance of the Prophet model on unseen data through cross-validation, input carbon emission data into the Prophet model, predict carbon emission trends, and output prediction results.

[0044] S73. Input carbon emission data into a neural network model to predict carbon emission trends, and add features of the carbon emission prediction model and Prophet model outputs before the last layer of the neural network model to output the final carbon emission trend prediction result.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] 1. This invention, by constructing a full life-cycle carbon emission model and flowchart model, can better manage carbon emission data and improve data quality by ensuring data accuracy and integrity.

[0047] 2. This invention updates the model data through a dynamic update mechanism, achieving efficient collection and management of carbon emission data from different sources, thereby improving the effectiveness of the data;

[0048] 3. This invention achieves an efficient carbon emission data processing algorithm through a carbon emission prediction model based on multinomial regression analysis, which can effectively improve the data quality and analysis capabilities of the power system. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a schematic diagram of the overall process of a carbon emission trend prediction method based on multinomial regression and neural networks provided in an embodiment of the present invention.

[0051] Figure 2 This is a schematic diagram of the process for constructing a carbon emission model covering the entire life cycle, as provided in this embodiment of the invention.

[0052] Figure 3 This is a schematic diagram of the full lifecycle flowchart model construction process provided in the embodiments of the present invention.

[0053] Figure 4 This is a schematic diagram of the data cleaning and preprocessing process provided in an embodiment of the present invention.

[0054] Figure 5 This is a schematic diagram of the model data outlier detection process provided in the embodiments of the present invention.

[0055] Figure 6 This is a schematic diagram of the model data consistency detection process provided in the embodiments of the present invention.

[0056] Figure 7 This is a schematic diagram of the carbon emission trend prediction process provided in an embodiment of the present invention. Detailed Implementation

[0057] The principles and features of the present invention are described below with reference to the accompanying drawings. The listed embodiments are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0058] Reference Figure 1 This embodiment provides a carbon emission trend prediction method based on multinomial regression and neural networks. The method includes the following steps:

[0059] S11. Construct a carbon emission model and a flowchart model for the entire life cycle, respectively.

[0060] S12. Perform data cleaning and preprocessing on the carbon emission model and flowchart model to output high-quality data.

[0061] S13. Update high-quality data through a dynamic update mechanism.

[0062] S14. Use the multinomial regression analysis method to establish a carbon emission prediction model, input carbon emission data into the carbon emission prediction model, and output the carbon emission trend prediction results.

[0063] In this embodiment, before constructing the full life cycle carbon emission model and flowchart model, a unified LCA database needs to be established first to address the problems of data opacity, poor quality, poor consistency and standardization, and limited data application in the current power carbon emission database.

[0064] The first step is to establish a power grid and power grid construction database. This involves reviewing relevant literature, analyzing the theoretical foundations of product and enterprise life cycle assessment (LCA) both domestically and internationally, identifying key environmental impact assessment methods, carbon footprint calculation standards, and industry best practices, and collecting necessary data sources such as power system transaction data, electricity consumption data, and production and usage data of high-energy-consuming products. Data acquisition technologies such as API integration, web crawling, or direct database links are employed to automatically collect carbon emission-related data from various sources, including power system transaction data and electricity consumption data. ETL (Extract, Transform, Load) tools are used to regularly collect data from various sources, and the data is preprocessed and cleaned to ensure quality. By standardizing the format and units of data from different sources, a strict data quality control process is implemented to ensure data accuracy and completeness. Based on the collected valid data and LCA theory, a power grid and power grid construction database are established.

[0065] Secondly, a high-carbon emission product database, an LCA (Limited Carbon Attachment) foundation database, and a literature database will be established. This will be achieved by collecting carbon footprint data from various domestic and international products, electricity trading and consumption data from the power system, and relevant data on high-energy-consuming products, as well as fundamental theoretical research materials and data on LCA, to enrich the content and quality of the databases. Data validity and accuracy will be ensured through data screening, organization, and classification. Units and formats from different data sources will be standardized, and a data quality control process will be implemented. Based on the collected valid data and LCA theory, a high-carbon emission product database, an LCA foundation database, and a literature database will be established.

[0066] Based on this, a full life-cycle carbon emission model is constructed, referring to... Figure 2 Specifically, it includes the following steps:

[0067] S21. Integrate the data from the power grid database, power grid construction database, and high-carbon emission product database to achieve mutual correlation and supplementation of data from different databases and obtain integrated data.

[0068] S22. Perform data cleaning on the integrated data.

[0069] In this step, the data cleaning includes, but is not limited to, removing duplicate records, correcting errors, and handling missing values, in order to improve data quality and accuracy.

[0070] S23. Utilize construction process data, operation data, product design and usage data from the integrated data to establish a full life cycle carbon emission model.

[0071] In this step, the construction process data in the integrated data comes from the power grid construction database, the operation data comes from the power grid power database, and the product design and usage data comes from the high carbon emission product database.

[0072] Reference Figure 3 Construct a full lifecycle flowchart model, which includes the following steps:

[0073] S31. Use the LCA basic database to obtain life cycle inventory data of high carbon emission products, and use literature databases to obtain historical carbon emission data, industry reports and scientific research information of high carbon emission products.

[0074] S32. Clean the acquired data and standardize the data format.

[0075] S33. Based on the cleaned data, clarify the scope of each life cycle stage of high-carbon emission products and construct a flowchart model of the entire life cycle.

[0076] As one possible implementation method, refer to Figure 4 Data cleaning and preprocessing were performed on the carbon emission model and flowchart model, specifically including:

[0077] S41. Detect outliers in the carbon emission model and flowchart model data using the DBSCAN clustering algorithm.

[0078] DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a clustering algorithm used to identify high-density regions in a dataset and group them into clusters, thus identifying outliers. Unlike traditional clustering algorithms such as K-means, DBSCAN does not require pre-specifying the number of clusters and can automatically identify different clusters in the data while effectively handling noisy points. By using the machine learning-based DBSCAN clustering algorithm to identify outliers in the data, the accuracy of outlier detection can be improved.

[0079] S42. For time series data with missing values ​​in carbon emission models and flowchart models, the missing values ​​are estimated by the prediction model, and the estimated values ​​are filled in the missing time series data by multiple imputation.

[0080] In this implementation, a model-based multiple interpolation method is used to estimate missing values ​​for time-series power grid operation data, such as forward imputation, backward imputation, or prediction models based on historical trends, and then the missing values ​​are imputed based on the estimation results.

[0081] S43. Perform consistency checks on the carbon emission model data and the flowchart model data.

[0082] This implementation method improves the completeness and accuracy of carbon emission model data and process model data through the above steps.

[0083] Reference Figure 5 The DBSCAN clustering algorithm is used to detect outliers in the carbon emission model and flowchart model data. Specifically, the following operations are performed:

[0084] S51. Input model data, radius parameter, and density threshold MinPts.

[0085] S52. Randomly select an unprocessed data object P from the model data, and determine whether the number of objects in its neighborhood meets the density threshold MinPts. If it does, mark P as a kernel object.

[0086] S53. Traverse the carbon emission model and flowchart model data, find all density-reachable objects originating from the kernel object P to form a new cluster, perform density-connection on all data objects within the cluster, and mark them as processed.

[0087] S54. Repeat steps S52 to S53 until all data objects in the model data have been processed, and data objects not included in any cluster are outliers.

[0088] Reference Figure 6 The consistency check between the carbon emission model and the flowchart model data includes the following operations:

[0089] S61. Define the quality standards for model data and determine the data quality indicators based on the quality standards.

[0090] This step defines quality standards for the model data, including completeness, consistency, accuracy, timeliness, and relevance. These standards ensure that the model data comprehensively and accurately reflects the environmental impact of the power system. Then, based on these quality standards, specific data quality indicators are set, such as data coverage, data time validity, data source reliability, and data missing rate.

[0091] S62. Based on data quality indicators, establish a data quality model using Apache Griffin.

[0092] In this step, the established model data quality standards and metrics are converted into an executable data quality model within the Apache Griffin framework.

[0093] S63. Run Apache Griffin to perform data quality checks on the model data, automatically identify data problems that do not meet quality standards, and generate and output data quality reports.

[0094] S64. Analyze the data quality report to identify areas where data improvement is needed. This may include validating data from the original data source, updating or supplementing missing data, or correcting data entry errors.

[0095] In this implementation, Apache Griffin can be configured to run data quality checks automatically on a regular basis, thereby establishing a regular review and monitoring mechanism to ensure that all data in the power system LCA database continues to meet requirements. This helps the power system LCA database maintain data quality, supports more reliable and effective environmental impact assessments, and provides data support for the sustainable development of the power industry.

[0096] As another possible implementation, high-quality data is updated through a dynamic update mechanism, specifically including the following operations:

[0097] A. A sliding window mechanism is adopted, with a fixed-size window. Carbon emission data is updated in real time by continuously introducing new high-quality data and discarding old data.

[0098] B. Use incremental learning algorithms to construct incremental gradient boosting trees and adjust the parameters of the carbon emission prediction model when the data is updated.

[0099] C. Real-time acquisition and processing of carbon emission data is achieved through a streaming data processing framework and data stream pipeline.

[0100] This implementation uses a sliding window mechanism to update carbon emission data, ensuring the real-time and dynamic nature of data analysis. By constructing an incremental gradient boosting tree, model parameters can be quickly adjusted when new data arrives without retraining the entire model, thus improving real-time response capabilities and enabling real-time updates to the carbon emission prediction model.

[0101] For example, the streaming data processing framework is Apache Flink, and the data stream pipeline is Apache Kafka. Apache Kafka is a distributed stream processing platform primarily used to build real-time data stream pipelines and stream processing applications; Apache Flink is a distributed stream processing framework focused on real-time data processing and analysis. Kafka can serve as both a data source and a data receiver for Flink, providing inlet and outlet ports for real-time data streams, while Flink can read data from Kafka and process it in real time. By using the streaming data processing frameworks Apache Kafka and Apache Flink together, an efficient data stream pipeline is built, ensuring low latency from data generation to analysis result output. This enables the entire system to efficiently process large-scale real-time data streams, providing flexible data processing and analysis functions, and achieving real-time acquisition and processing of carbon emission data.

[0102] As another possible implementation method, refer to Figure 7 A carbon emission prediction model is established using multinomial regression analysis. Carbon emission data is input into the model, and the carbon emission trend prediction results are output. The specific operations include the following:

[0103] S71. Construct a carbon emission prediction model based on the multinomial regression analysis method, input carbon emission data into the carbon emission prediction model, and output the basic trend of carbon emission changes over time.

[0104] S72. Create a Prophet model, train the Prophet model using historical datasets, evaluate the performance of the Prophet model on unseen data through cross-validation, input carbon emission data into the Prophet model, predict carbon emission trends, and output prediction results.

[0105] S73. Input carbon emission data into a neural network model to predict carbon emission trends, and add features of the carbon emission prediction model and Prophet model outputs before the last layer of the neural network model to output the final carbon emission trend prediction result.

[0106] This implementation uses multinomial regression analysis to build a carbon emission prediction model. By fitting a trend line to the carbon emission data, it captures the nonlinear relationships in the data to predict future carbon emission trends. Simultaneously, the Prophet time-series forecasting algorithm is applied to simulate and predict carbon emissions, capturing seasonal effects, trend changes, and periodic fluctuations in the time-series data to improve prediction accuracy. Finally, features from the carbon emission prediction model and the Prophet model are added before the last layer of the neural network model to output the final carbon emission trend prediction result. This implementation fully utilizes traditional statistical methods (multinomial regression), algorithms specifically designed for time series (Prophet), and powerful machine learning tools (DNN), thus providing a comprehensive and flexible solution for carbon emission trend prediction.

[0107] For example, the neural network model can be a BP neural network, a fully connected neural network, or other types of neural networks. This embodiment does not specifically limit it.

[0108] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A carbon emission trend prediction method based on multinomial regression and neural networks, characterized in that, The method includes: S11. Construct a carbon emission model and a flowchart model for the entire life cycle, respectively; S12. Perform data cleaning and preprocessing on the carbon emission model and flowchart model to output high-quality data; S13. Update high-quality data through a dynamic update mechanism; S14. Use the multinomial regression analysis method to establish a carbon emission prediction model, input carbon emission data into the carbon emission prediction model, and output the carbon emission trend prediction results. Constructing a full life-cycle carbon emission model includes the following steps: S21. Integrate the data from the power grid database, power grid construction database, and high-carbon emission product database to achieve mutual correlation and supplementation of data in different databases and obtain integrated data. S22. Perform data cleaning on the integrated data; S23. Utilize construction process data, operation data, product design and usage data from the integrated data to establish a full life cycle carbon emission model; Constructing a full lifecycle flowchart model includes the following operations: S31. Use the LCA basic database to obtain life cycle inventory data of high carbon emission products, and use literature databases to obtain historical carbon emission data, industry reports and scientific research information of high carbon emission products. S32. Clean the acquired data and standardize the data format; S33. Based on the cleaned data, clarify the scope of each life cycle stage of high-carbon emission products and construct a flowchart model of the entire life cycle. A carbon emission prediction model is established using multinomial regression analysis. Carbon emission data is input into the model, and the output is a carbon emission trend prediction result. The specific operations include the following: S71. Construct a carbon emission prediction model based on the multinomial regression analysis method, input carbon emission data into the carbon emission prediction model, and output the basic trend of carbon emission changes over time. S72. Create a Prophet model, train the Prophet model using historical datasets, evaluate the performance of the Prophet model on unseen data through cross-validation, input carbon emission data into the Prophet model, predict carbon emission trends and output prediction results. S73. Input carbon emission data into a neural network model to predict carbon emission trends, and add features of the carbon emission prediction model and Prophet model outputs before the last layer of the neural network model to output the final carbon emission trend prediction result.

2. The method according to claim 1, characterized in that, Data cleaning and preprocessing were performed on the carbon emission model and flowchart model, specifically including: S41. Detect outliers in the carbon emission model and flowchart model data using the DBSCAN clustering algorithm; S42. For time series data with missing values ​​in carbon emission models and flowchart models, estimate the missing values ​​using the prediction model, and fill the missing time series data with the estimated values ​​through multiple imputation. S43. Perform consistency checks on the carbon emission model data and the flowchart model data.

3. The method according to claim 2, characterized in that, The DBSCAN clustering algorithm is used to detect outliers in the carbon emission model and flowchart model data. The specific operations include the following: S51, Input model data, radius parameters and density threshold MinPts; S52. Randomly select an unprocessed data object P from the model data, and determine whether the number of objects in its neighborhood meets the density threshold MinPts. If it does, mark P as a kernel object. S53. Traverse the carbon emission model and flowchart model data, find all density-reachable objects originating from the kernel object P to form a new cluster, perform density-connection on all data objects within the cluster, and mark them as processed. S54. Repeat steps S52 to S53 until all data objects in the model data have been processed.

4. The method according to claim 2, characterized in that, The consistency check between the carbon emission model and the flowchart model data includes the following steps: S61. Define the quality standards for model data and determine the data quality indicators based on the quality standards. S62. Based on data quality indicators, establish a data quality model using Apache Griffin; S63. Run Apache Griffin to perform data quality checks on the model data, automatically identify data problems that do not meet quality standards, and generate and output data quality reports. S64. Analyze the data quality report to identify data areas that need improvement.

5. The method according to claim 1, characterized in that, High-quality data is updated through a dynamic update mechanism to enable real-time collection and updating of carbon emission data. This includes the following operations: A sliding window mechanism is adopted, with a fixed-size window set, to continuously introduce new high-quality data and discard old data to update carbon emission data in real time; Incremental gradient boosting trees are constructed using incremental learning algorithms to adjust the parameters of the carbon emission prediction model when data is updated. Real-time acquisition and processing of carbon emission data is achieved through a streaming data processing framework and data stream pipeline.

6. The method according to claim 5, characterized in that, The streaming data processing framework is Apache Flink, and the data stream pipeline is Apache Kafka.

Images