Carbon market price prediction method considering multi-dimensional influence factors

Abstract

The invention discloses a carbon market price prediction method considering multi-dimensional influence factors, and relates to the technical field of carbon market price prediction, comprising the following steps: obtaining carbon market price influence data including energy factor influence data and index influence data; obtaining text data of carbon market internal cause information, processing the text data of the carbon market internal cause information based on a natural language to obtain a plurality of keyword vectors, and obtaining a comprehensive feature vector of the carbon market internal cause information based on cross fusion of the plurality of keyword vectors; obtaining a carbon market price prediction model based on the comprehensive feature vectors of the carbon market price influence data and the carbon market internal cause information on the basis of SVR (Support Vector Regression); and generating short-term and long-term carbon market price curve graphs according to the carbon market price prediction model, and judging the carbon market price prediction trend according to the short-term and long-term carbon market price curve graphs. Through the SVR technology, the problem that the accuracy of a prediction model is reduced due to the fact that data collection and considered factors are often not perfect enough is solved. And some models may have limitations when processing a nonlinear relationship and multivariable interaction, resulting in insufficient predictive ability.

Description

Technical Field

[0001] The present invention relates to the technical field of optimal decision-making for power grid technical transformation projects, and more specifically, to the technical field of a carbon market price prediction method that takes multi-dimensional influencing factors into consideration. Background Art

[0002] The carbon market is a market mechanism established to address global climate change, aiming to reduce greenhouse gas emissions through economic means. Its core concept is "emissions trading", which allows companies to buy and sell emission quotas, thereby providing economic incentives to reduce emissions. As the problem of global warming intensifies, countries have taken measures, and the carbon market has gradually grown as an important tool for emission reduction.

[0003] The ‌SVR algorithm‌ is a regression algorithm based on support vector machine (SVM), which is mainly used to solve regression problems, that is, predicting continuous numerical targets. SVR fits the data by constructing a separating hyperplane. The core idea is to build a model by finding support vectors to achieve modeling of uncertain functional relationships. SVR is used for regression analysis rather than classification. SVR improves the robustness of the model by maximizing the interval of the regression model while allowing some samples to violate a certain error range (called tolerance deviation). ‌

[0004] In recent years, the research on carbon emission price prediction has attracted the attention of many experts, professionals and emission-controlled enterprises. At present, carbon market prediction relies on a large amount of high-quality data, but in many regions, especially developing countries, data collection and factors to be considered are often not perfect, resulting in a decrease in the accuracy of the prediction model. In addition, some models may have limitations in dealing with nonlinear relationships and multivariate interactions, resulting in insufficient prediction capabilities. In view of the above problems, the present invention proposes a solution. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a carbon market price forecasting method taking into account multi-dimensional influencing factors. Through the SVR technology, it solves the problem that data collection and factors considered are often not perfect, resulting in a decrease in the accuracy of the forecasting model, and some models may have limitations when dealing with nonlinear relationships and multivariate interactions, resulting in insufficient forecasting capabilities.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for predicting carbon market prices considering multi-dimensional influencing factors comprises the following steps: obtaining carbon market price influencing data, wherein the carbon market price influencing data comprises energy factor influencing data and indicator influencing data; obtaining text data of carbon market intrinsic information; obtaining a feature vector of the carbon market intrinsic information based on natural language processing for the text data of the carbon market intrinsic information; obtaining a carbon market price prediction model based on SVR for the carbon market price influencing data and the feature vector of the carbon market intrinsic information; and judging the trend of carbon market price prediction based on the carbon market price prediction model.

[0007] In a preferred embodiment, the energy factor impact data is specifically acquired by the following method: acquiring the first data of energy factor impact data in real time through market research, wherein the first data includes coal energy prices, natural gas energy prices, and oil energy prices; analyzing historical energy demand data based on an econometric model to find the relationship curve between energy demand and energy prices, and obtaining the energy demand sensitivity coefficient by performing a regression analysis on the relationship curve between energy demand and energy prices; counting the carbon quotas of all enterprises in the region and the total carbon emissions of all enterprises; and obtaining the energy factor impact data based on the carbon quotas and total carbon emissions of the enterprises combined with the first data of energy factor impact data and the energy demand sensitivity coefficient based on a preset energy factor impact formula.

[0008] In a preferred embodiment, the indicator impact data is specifically obtained by the following method: obtaining basic data of the indicator impact data, the basic data including GDP growth rate and market transaction volume; obtaining price fluctuation patterns of the carbon market based on data analysis by obtaining historical carbon trading prices; real-time monitoring of the increase and decrease of industrial activities to obtain the industrial production impact index; obtaining the price of the carbon market reporting period and the price of the base period based on the price fluctuation pattern of the carbon market by recording the price changes of the carbon market in different regions, and obtaining the carbon market output impact coefficient based on the weighted average method according to the weight of the carbon market output in each region; obtaining the weight of the carbon market production impact coefficient based on the industrial production impact index of industrial activities, and obtaining the carbon market production impact coefficient through the weighted average method; obtaining the indicator impact data based on the preset indicator impact data calculation formula according to the carbon market output impact coefficient and the carbon market production impact coefficient combined with the basic data of the indicator impact data.

[0009] In a preferred embodiment, the text data of the carbon market information obtains the feature vector of the carbon market intrinsic information based on natural language processing, and the specific steps are as follows: preprocessing the text data of the carbon market information to remove words, punctuation marks and special characters that do not contribute much to the meaning of the text, and performing word segmentation on the preprocessed text data of the carbon market information to obtain multiple keyword sequences; capturing the semantic relationship and context information between the keyword sequences based on a word embedding model and performing model sequence encoding to convert the multiple keyword sequences into multiple keyword vectors; and fusing the multiple keyword vectors to obtain a comprehensive feature vector of the carbon market information.

[0010] In a preferred embodiment, the characteristic vectors of carbon market price impact data and carbon market intrinsic information are used to obtain a carbon market price prediction model based on SVR, and the specific steps are as follows: data cleaning is performed on the characteristic vectors of historical carbon market price impact data and historical carbon market intrinsic information, and the cleaned data is converted into characteristic data of the same dimension based on data normalization processing; based on SVR, RBF kernel function is selected as the function for constructing the initial carbon market price prediction model, and the characteristic vectors of the processed historical carbon market price impact data and historical carbon market intrinsic information are used based on the RBF kernel function to generate an initial carbon market price prediction model; the initial carbon market price prediction model is optimized based on cross-validation and grid search methods to generate a complete carbon market price prediction model.

[0011] In a preferred embodiment, the cross-validation comprises the following specific steps: dividing the processed feature vectors of historical carbon market price impact data and historical carbon market internal information into a training set, a validation set and a test set; further dividing the training set into K subsets; for each iteration, selecting one of the K subsets as a validation set and the remaining K-1 subsets as training sets to train the model; using the validation set to evaluate the performance of the model; repeating the above process K times, each time selecting a different subset as a validation set; and calculating the average performance index of the K evaluations as the final performance index of the model.

[0012] In a preferred embodiment, the grid search comprises the following specific steps: defining the range and combination of feature vector data of historical carbon market price impact data and historical carbon market intrinsic information; for each parameter combination, using cross-validation to evaluate the performance of the model; recording the performance indicators of each parameter combination, and selecting the parameter combination with the best performance indicators as the final model parameters; and retraining the model using the best parameter combination based on the results of the grid search to ensure that the model has good generalization ability and accuracy.

[0013] In a preferred embodiment, the carbon market price forecasting model is used to judge the trend of carbon market price forecast, and the specific steps are as follows: obtaining short-term and long-term carbon market price impact data and text data of carbon market information in different regions with rich carbon markets; obtaining the feature vector of the text data of short-term carbon market information in different regions based on natural language processing and inputting it into the carbon market price forecasting model in combination with the short-term carbon market price impact data in different regions to generate different short-term carbon market price forecast values, and draw a short-term-carbon market price curve chart; obtaining the feature vector of the text data of long-term carbon market information in different regions based on natural language processing and inputting it into the carbon market price forecasting model in combination with the short-term carbon market price impact data in different regions to generate different long-term carbon market price forecast values ​​and draw a long-term-carbon market price curve chart; based on the moving average method, the intersection and trend changes of the moving averages of the short-term-carbon market price curve chart and the long-term-carbon market price curve chart are used to judge the trend of carbon market price forecast.

[0014] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. Through the SVR technology, the problem of imperfect data collection and factors considered, which leads to reduced accuracy of prediction models, is solved. In addition, some models may have limitations in dealing with nonlinear relationships and multivariate interactions, resulting in insufficient prediction capabilities.

[0015] 2. The accuracy of the model generalization ability evaluation was improved through cross-validation and grid search methods, and the optimal parameter combination was found through systematic traversal, further improving the performance and stability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of the structure of a carbon market price prediction method considering multi-dimensional influencing factors provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Embodiment 1, Figure 1 A structural diagram of a carbon market price prediction method considering multi-dimensional influencing factors provided in an embodiment of the present application includes the following steps: S1. Obtaining carbon market price impact data, wherein the carbon market price impact data includes energy factor impact data and indicator impact data.

[0019] The carbon market price impact data refers to various information and indicators that can affect the trading price of carbon emission rights. By understanding the impact of carbon market prices, enterprises and investors can make more accurate decisions based on market data, formulate more effective emission reduction strategies and investment plans, thereby reducing costs and increasing profits. Moreover, understanding the various factors that affect prices helps predict market trends, enabling participants to seize investment opportunities and adjust strategies in a timely manner to adapt to market changes. The carbon market price impact data mentioned include energy factor impact data and indicator impact data.

[0020] Energy factor impact data refers to the price information of relevant energy commodities (such as coal, natural gas, oil, etc.) that affect the price of carbon market. These data can help analyze the cost and trend of carbon emissions. Energy prices are usually closely related to carbon emissions. For example, rising coal prices may cause power generation companies to reduce coal use and switch to cleaner energy, thereby affecting the supply and demand relationship in the carbon market. When the price of a certain energy source rises, it may drive companies and consumers to look for alternative energy sources, which may affect the demand in the carbon market. For example, rising natural gas prices may cause more companies to turn to renewable energy, thereby reducing carbon emissions. Changes in energy prices may affect the market's expectations of future carbon prices, and investors and companies may adjust their behavior in the carbon market based on their predictions of energy price trends. Therefore, energy factor impact data is an important variable in the carbon market model.

[0021] Analyzing the impact of energy factors on carbon market price forecasts has the following benefits: Analyzing the impact data of energy factors can improve the accuracy of forecasts, help decision makers make more effective decisions, and promote strategic adjustments of enterprises. The following are some of the main benefits: First, there is a close connection between energy prices and carbon emissions. The production and consumption of energy directly affects the level of carbon emissions. By analyzing changes in energy prices, analysts can identify which energy sources are dominating the market and how they affect overall carbon emissions. For example, when coal prices rise, many power generation companies may turn to natural gas or renewable energy, which will directly lead to a decrease in carbon emissions. Such analysis helps predict fluctuations in carbon market prices, as carbon prices are usually driven by increased demand for renewable energy.

[0022] Second, changes in energy prices can affect the expectations of market participants. The behavior of companies and investors in the carbon market is often based on predictions of future energy price trends. For example, if natural gas prices continue to rise, companies may expect carbon prices to rise as well, because higher energy costs may lead to reduced carbon emissions. By monitoring and analyzing these trends, market participants are able to adjust their investment strategies more accurately, thereby improving the effectiveness of forecasts.

[0023] Finally, as global attention to climate change intensifies, companies increasingly need to consider changes in energy prices when formulating long-term strategies. By systematically analyzing energy factors, companies can better formulate emission reduction targets and investment plans to adapt to changes in the carbon market. This forward-looking analysis can not only reduce corporate risks, but also improve their market competitiveness.

[0024] In summary, analyzing the impact of energy factors on carbon market prices has significant benefits. It not only improves the accuracy of forecasts, but also provides an important basis for corporate strategy and promotes sustainable development goals. As the carbon market continues to develop, in-depth analysis of this data will become increasingly important.

[0025] The energy factor influences the data, and the specific method for obtaining the data is as follows: Obtaining first data of energy factor impact data in real time through market research, wherein the first data includes coal energy price, natural gas energy price and oil energy price; Analyze historical energy demand data based on an econometric model to find a relationship curve between energy demand and energy price, and obtain an energy demand sensitivity coefficient by performing a regression analysis on the relationship curve between energy demand and energy price; Count the carbon quotas of all companies and the total carbon emissions of all companies; The energy factor impact data is obtained by calculating based on a preset energy factor impact formula according to the carbon quota and total carbon emissions of the enterprise in combination with the first data of energy factor impact data and the energy demand sensitivity coefficient.

[0026] The energy factors affect the data, and the specific calculation method is as follows:

[0027] In the formula, L is the energy factor impact data, , , For coal energy prices, natural gas energy prices, oil energy prices, is the energy demand sensitivity coefficient, is the total carbon emissions of the enterprise, and Z is the carbon quota of the enterprise.

[0028] It should be noted that fossil fuels (such as coal, oil and natural gas) are one of the main sources of carbon dioxide emissions. Therefore, changes in fossil fuel prices will directly affect the production costs and carbon emission costs of enterprises. When the price of fossil fuels rises, enterprises may reduce their use, thereby reducing carbon emissions, which may have a certain impact on carbon market prices. Conversely, when the price of fossil fuels falls, enterprises may increase their use, resulting in increased carbon emissions, which in turn puts pressure on carbon market prices.

[0029] Indicator impact data refers to statistical data and economic variables that reflect the overall economic situation and development trends. These indicators can help analyze the relationship between economic activities and carbon emissions, thereby affecting carbon market prices.

[0030] Analyzing the impact of indicators on carbon market prices has the following benefits: Indicator impact data such as GDP growth rate, unemployment rate and industrial output directly reflect the overall state of a country's economic activities. Economic growth is usually accompanied by an increase in energy demand, which drives up carbon emissions. This phenomenon makes carbon prices likely to rise. By analyzing these indicators, market participants can predict the price trend of the carbon market and optimize their investment strategies.

[0031] Inflation rates and interest rates also have an impact on the carbon market. High inflation usually means rising commodity prices and increased corporate costs, which may prompt companies to reduce carbon emissions to reduce costs. In addition, changes in interest rates will affect corporate investment decisions, especially investments in renewable energy and carbon reduction technologies. Therefore, understanding the changes in these macroeconomic indicators can help investors better grasp the dynamics of the carbon market.

[0032] In addition, the international economic environment also has an important impact on the carbon market. Changes in the global economy, fluctuations in international trade, and environmental protection in other countries will have an impact on the carbon market. For example, economic recovery may lead to an increase in demand for carbon quotas, which in turn drives up global carbon prices. Therefore, combined with the analysis of global economic indicators, a more comprehensive perspective can be provided for price forecasts in the carbon market.

[0033] Finally, by building a model between the indicator impact data and carbon market prices, the researchers were able to quantify the relationship between the two. This quantitative analysis can not only help market participants understand historical price trends, but also provide a basis for future price forecasts. This model can reveal the specific impact of various macroeconomic variables on carbon market prices through regression analysis, time series analysis and other methods, thereby enhancing the scientific nature of decision-making.

[0034] The indicators affect the data, and the specific acquisition method is as follows: Obtaining basic data on the impact of the indicator, including GDP growth rate and market transaction volume; Obtain the price fluctuation pattern of the carbon market through data analysis by obtaining historical carbon trading prices; Real-time monitoring of changes in industrial activities to obtain the industrial production impact index; By recording the price changes of carbon markets in different regions, we can obtain the prices of the carbon market reporting period and the base period based on the price fluctuation rules of the carbon market, and obtain the carbon market output impact coefficient based on the weighted average method according to the output weight of the carbon market in each region; The weight of the carbon market production impact coefficient is obtained based on the industrial production impact index of industrial activities, and the carbon market production impact coefficient is obtained through the weighted average method; The indicator impact data is obtained by calculating the carbon market output impact coefficient and the carbon market production impact coefficient in combination with basic data of the indicator impact data based on a preset indicator impact data calculation formula.

[0035] The indicators affect the data, and the specific calculation method is as follows:

[0036] In the formula, is the indicator impact data, CPI is the carbon market output impact coefficient, Y is the GDP growth rate, G is the market transaction volume, and PPI is the carbon market production impact coefficient.

[0037] It should be noted that the GDP growth rate reflects the level of overall economic activity and has a direct impact on carbon emission demand. The industrial production index measures the increase or decrease in industrial production activities and is closely related to carbon emissions. The Consumer Price Index (CPI) and the Producer Price Index (PPI): reflect changes in price levels and indirectly affect energy consumption and carbon emissions. Market trading volume reflects the activity of the market and the enthusiasm of participants. Analyzing these data plays a significant role in obtaining indicator impact data.

[0038] S2, obtain the text data of the carbon market's intrinsic information, obtain multiple keyword vectors based on natural language processing for the text data of the carbon market's intrinsic information, and obtain the comprehensive feature vector of the carbon market's intrinsic information based on cross-fusion of the multiple keyword vectors. Among them, the carbon market's intrinsic information covers market expectations, the determination of key emission areas, the supervision of carbon emission data quality, and the allocation of carbon emission quotas and greenhouse gas emission reports. The feature vector of the carbon market's intrinsic information is obtained based on natural language processing for the text data of the carbon market's intrinsic information. It should be understood that the carbon market's intrinsic information includes market expectations, the determination of key emission areas, the supervision of carbon emission data quality, and the allocation of carbon emission quotas and greenhouse gas emission reports. These rule information usually exists in text form and cannot be directly used as input for the carbon market price model. Therefore, it is necessary to process the text data of these carbon market's intrinsic information through natural language technology and convert these data into feature vectors to facilitate the input of the model. The specific steps are as follows: Preprocessing the text data of the carbon market internal information to remove words, punctuation marks and special characters that do not contribute much to the meaning of the text, and performing word segmentation processing on the preprocessed text data of the carbon market internal information to obtain multiple keyword sequences; Capturing semantic relations and context information between multiple keyword sequences based on a word embedding model and performing model sequence encoding to convert the multiple keyword sequences into multiple keyword vectors; The multiple keyword vectors are fused to obtain a comprehensive feature vector of the carbon market internal information.

[0039] The cross-fusion of the multiple keyword vectors to obtain a comprehensive feature vector of the carbon market internal information is specifically calculated as follows:

[0040] In the formula, is the comprehensive characteristic vector of the carbon market internal information, is the nth keyword vector, n=1,2,3,...,n, n is an integer, and N is the number of keyword vectors.

[0041] It should be noted that fusing the multiple keyword vectors can be understood as combining multiple keyword vectors to form a feature vector of the entire keyword sequence. A specific fusion method is to average the vectors of all words to obtain a feature vector of a fixed length, which is used as a comprehensive feature vector of the carbon market internal information.

[0042] Among them, word embedding technology is a technology widely used in natural language processing (NLP). It effectively captures the semantic relationship between words. The specific content is to convert words from traditional discrete symbol representations (such as one-hot encoding) into continuous high-dimensional vector representations. This distributed representation method enables words to occupy a position in the vector space, and the distance and direction between vectors can reflect the semantic relationship between words. Word embedding technology learns the contextual information of words by training models on large-scale corpora. The model captures the usage patterns of words in different contexts, thereby learning the semantic connection between words. Specifically, if two words frequently co-occur in multiple contexts, their positions in the vector space will also be relatively close.

[0043] S3, based on SVR, the characteristic vectors of carbon market price impact data and carbon market internal information are used to obtain the carbon market price prediction model.

[0044] Among them, SVR can handle nonlinear relationships and is suitable for the complex dynamic characteristics of carbon market prices. Through the kernel function, SVR can effectively find the best regression hyperplane in high-dimensional space to capture the potential pattern of market prices. It is also highly resistant to outliers and can provide stable predictions even when there is noise in the data. This is especially important in the carbon market because market fluctuations and emergencies may lead to abnormal prices. The SVR model can be adjusted according to different data sets and is applicable to various types of market data, thus providing flexible solutions for different forecasting needs.

[0045] The carbon market price prediction model is obtained by using the characteristic vectors of the carbon market price impact data and the carbon market internal information based on SVR. The specific steps are as follows: Performing data cleaning on the characteristic vectors of the historical carbon market price impact data and the historical carbon market internal information. The data cleaning can be understood as removing duplicate values ​​from the characteristic vectors of the historical carbon market price impact data and the historical carbon market internal information to ensure that there are no duplicate records, and converting the characteristic data of different dimensions into characteristic data of the same dimension based on data normalization processing after the cleaning; Based on SVR, the RBF kernel function is selected as the function for constructing the initial carbon market price prediction model. The feature vectors of the processed historical carbon market price impact data and historical carbon market internal information are based on the RBF kernel function to generate the initial carbon market price prediction model. The initial carbon market price prediction model is optimized based on cross-validation and grid search methods to generate a complete carbon market price prediction model.

[0046] The specific calculation formula of the RBF kernel function is as follows:

[0047] In the formula, K is the kernel function, L is the energy factor impact data, For indicator impact data, is the comprehensive feature vector of carbon market information.

[0048] When processing the characteristic vectors of historical carbon market price impact data and historical carbon market internal information, using cross-validation and grid search methods for optimization is an effective method to improve the generalization ability of the model and find the best parameter combination. The specific steps of cross-validation are as follows: Divide the processed historical carbon market price impact data and the feature vectors of historical carbon market internal information into training set, validation set and test set; The training set is further divided into K subsets; For each iteration, one of the K subsets is selected as the validation set, and the remaining K-1 subsets are used as the training set to train the model; Use the validation set to evaluate the performance of the model; Iterate K times, selecting a different subset as the validation set each time; The average performance index of K evaluations is calculated as the final performance index of the model.

[0049] Grid search is a method that systematically traverses multiple parameter combinations and evaluates the performance of each combination through cross-validation to find the best parameter combination. The specific steps of the grid search are as follows: Define the scope and combination of the characteristic vector data of historical carbon market price impact data and historical carbon market internal information; For each parameter combination, cross-validation was used to evaluate the performance of the model; Record the performance indicators of each parameter combination, and select the parameter combination with the best performance indicators as the final model parameters; Based on the results of the grid search, the model is retrained using the best parameter combination to ensure that the model has good generalization ability and accuracy.

[0050] S5, based on the carbon market price prediction model, determining the carbon market price prediction trend, the specific steps are as follows: Obtain text data on the carbon market's short-term and long-term carbon market price impacts in different regions and carbon market internal factors; The text data of the short-term carbon market internal information in different regions is used to obtain its feature vector based on natural language processing and input into the carbon market price prediction model in combination with the short-term carbon market price impact data in different regions to generate different short-term carbon market price prediction values ​​and draw a short-term-carbon market price curve chart; The text data of the internal factors of the long-term carbon market in different regions are processed based on natural language to obtain their feature vectors and combined with the short-term carbon market price impact data of different regions to input into the carbon market price prediction model, generate different long-term carbon market price prediction values ​​and draw a long-term carbon market price curve. Based on the moving average method, the forecast trend of carbon market prices is determined by utilizing the intersection and trend changes of the moving averages of the short-term carbon market price curve and the long-term carbon market price curve.

[0051] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0052] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0053] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0054] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0055] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0056] Finally: 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 in the protection scope of the present invention.

Claims

1. A carbon market price prediction method considering multi-dimensional influencing factors, characterized in that: The steps include: Obtaining carbon market price impact data, wherein the carbon market price impact data includes energy factor impact data and indicator impact data; Acquire text data of carbon market intrinsic information, obtain multiple keyword vectors based on natural language processing for the text data of carbon market intrinsic information, and obtain a comprehensive feature vector of carbon market intrinsic information based on cross-fusion of the multiple keyword vectors; Obtaining a carbon market price prediction model based on SVR using the comprehensive feature vector of the carbon market price impact data and the carbon market internal information; A short-term and long-term carbon market price curve graph is generated according to the carbon market price prediction model, and the carbon market price prediction trend is determined according to the short-term and long-term carbon market price curve graph.

2. A carbon market price prediction method considering multi-dimensional influencing factors according to claim 1, characterized in that: The energy factor influences the data, and the specific method for obtaining the data is as follows: Acquire first data of energy factor impact data, wherein the first data includes coal energy price, natural gas energy price and oil energy price; Analyze historical energy demand data based on an econometric model to obtain a relationship curve between energy demand and energy price, and obtain an energy demand sensitivity coefficient by performing a regression analysis on the relationship curve between energy demand and energy price; Count the carbon quotas of all companies and the total carbon emissions of all companies; The energy factor impact data is obtained by calculating based on a preset energy factor impact formula according to the carbon quota and total carbon emissions of the enterprise in combination with the first data of energy factor impact data and the energy demand sensitivity coefficient.

3. A carbon market price prediction method considering multi-dimensional influencing factors according to claim 1, characterized in that: The indicators affect the data, and the specific acquisition method is as follows: Obtaining basic data on the impact of the indicator, including GDP growth rate and market transaction volume; Obtain the price fluctuation pattern of the carbon market through data analysis by obtaining historical carbon trading prices; Real-time monitoring of changes in industrial activities to obtain the industrial production impact index; By recording the price changes of carbon markets in different regions, we can obtain the prices of the carbon market reporting period and the base period based on the price fluctuation rules of the carbon market, and obtain the carbon market output impact coefficient based on the weighted average method according to the output weight of the carbon market in each region; The weight of the carbon market production impact coefficient is obtained based on the industrial production impact index of industrial activities, and the carbon market production impact coefficient is obtained through the weighted average method; The indicator impact data is obtained by calculating the carbon market output impact coefficient and the carbon market production impact coefficient in combination with basic data of the indicator impact data based on a preset indicator impact data calculation formula.

4. A carbon market price prediction method considering multi-dimensional influencing factors according to claim 1, characterized in that: The steps of obtaining text data of carbon market internal information, obtaining multiple keyword vectors based on natural language processing for the text data of carbon market internal information, and obtaining a comprehensive feature vector of carbon market internal information based on cross-fusion of the multiple keyword vectors are as follows: Preprocessing the text data of the carbon market internal information to remove words, punctuation marks and special characters that do not contribute much to the meaning of the text, and performing word segmentation on the preprocessed text data of the carbon market internal information to obtain multiple keyword sequences; Capturing semantic relations and context information between multiple keyword sequences based on a word embedding model and performing model sequence encoding to convert the multiple keyword sequences into multiple keyword vectors; The multiple keyword vectors are cross-fused to obtain a comprehensive feature vector of the carbon market internal information.

5. The method for predicting carbon market prices considering multi-dimensional influencing factors according to claim 1, characterized in that: The steps of obtaining a carbon market price prediction model based on SVR using the comprehensive feature vector of the carbon market price impact data and the carbon market internal information are as follows: The characteristic vectors of historical carbon market price impact data and historical carbon market internal information are cleaned, and the cleaned data are converted into characteristic data of the same dimension based on data normalization; Based on SVR, the RBF kernel function is selected as the function for constructing the initial carbon market price prediction model. The feature vectors of the processed historical carbon market price impact data and historical carbon market internal information are based on the RBF kernel function to generate the initial carbon market price prediction model. The initial carbon market price prediction model is optimized based on cross-validation and grid search methods to generate a complete carbon market price prediction model.

6. A carbon market price prediction method considering multi-dimensional influencing factors according to claim 5, characterized in that: The specific steps of the cross validation are as follows: Divide the processed historical carbon market price impact data and the feature vectors of historical carbon market internal information into training set, validation set and test set; The training set is further divided into K subsets; For each iteration, one of the K subsets is selected as the validation set, and the remaining K-1 subsets are used as the training set to train the model; Use the validation set to evaluate the performance of the model; Iterate K times, selecting a different subset as the validation set each time; The average performance index of K evaluations is calculated as the final performance index of the model.

7. A method for predicting carbon market prices considering multi-dimensional influencing factors according to claim 5, characterized in that: The specific steps of the grid search are as follows: Define the scope and combination of the characteristic vector data of historical carbon market price impact data and historical carbon market internal information; For each parameter combination, cross-validation was used to evaluate the performance of the model; Record the performance indicators of each parameter combination, and select the parameter combination with the best performance indicators as the final model parameters; Based on the results of the grid search, the model is retrained using the best parameter combination to ensure that the model has good generalization ability and accuracy.

8. The method for predicting carbon market prices considering multi-dimensional influencing factors according to claim 1, characterized in that: The steps of generating short-term and long-term carbon market price curves according to the carbon market price forecasting model and judging the carbon market price forecasting trend according to the short-term and long-term carbon market price curves are as follows: Obtain text data on the expected short-term and long-term carbon market price impacts and carbon market internal factors in different regions; The text data of the short-term carbon market internal information in different regions is used to obtain its feature vector based on natural language processing and input into the carbon market price prediction model in combination with the short-term carbon market price impact data in different regions to generate different short-term carbon market price prediction values ​​and draw a short-term-carbon market price curve chart; The text data of the internal factors of the long-term carbon market in different regions are processed based on natural language to obtain their feature vectors and combined with the short-term carbon market price impact data of different regions to input into the carbon market price prediction model, generate different long-term carbon market price prediction values ​​and draw a long-term carbon market price curve. Based on the moving average method, the forecast trend of carbon market prices is determined by utilizing the intersection and trend changes of the moving averages of the short-term carbon market price curve and the long-term carbon market price curve.

9. A carbon market price prediction method considering multi-dimensional influencing factors according to claim 3, characterized in that: The indicators affect the data, and the specific calculation method is as follows: In the formula, is the indicator impact data, CPI is the carbon market output impact coefficient, Y is the GDP growth rate, G is the market transaction volume, and PPI is the carbon market production impact coefficient.

10. A carbon market price prediction method considering multi-dimensional influencing factors according to claim 5, characterized in that: The specific calculation formula of the RBF kernel function is as follows: In the formula, K is the kernel function, L is the energy factor impact data, For indicator impact data, is the comprehensive characteristic vector of the carbon market’s internal information.

Images