Carbon market transaction price influence factor analysis method and system based on adaptive neural fuzzy inference system
Through the adaptive neural fuzzy inference system model, the influencing factors of the carbon market transaction price was analyzed, and the problem of the inability to accurately analyze the driving factors of the carbon market transaction price fluctuations in the existing technology was solved, and more accurate analysis of influencing factors and price fluctuations were achieved. .
Patent Information
- Application Number
- CN202311144420.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-06
- Publication Date
- 2024-11-12
AI Technical Summary
The existing technology lacks systematic methods to analyze the influencing factors of the carbon market trading price, and cannot accurately analyze the driving factors of fluctuations in the carbon market trading price sequence at different stages.
Adopting the adaptive neural fuzzy inference system model, by obtaining the carbon market transaction price and series of influencing factors, establishing an influencing factor index system, conducting descriptive statistics and fully integrated empirical modal decomposition of adaptive noise, and establishing adaptive neural fuzzy Inference system model, performance testing and influencing factor analysis.
It improves the accuracy and pertinence of influencing factor analysis, effectively reduces noise, and can be directly used for the analysis of influencing factors of carbon market transaction prices at different stages, providing a more accurate analysis of the driving factors of price fluctuations.
Smart Images

Figure CN118941310A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of carbon trading price influencing factor analysis, and in particular to a carbon market transaction price influencing factor analysis method and system based on an adaptive neural fuzzy inference system. Background Art
[0002] As an emission reduction tool operated by a market mechanism, the carbon trading market has the function of price discovery and the characteristic of uncertainty in carbon emission reduction costs. Compared with emission reduction tools such as carbon taxation, the carbon market has higher political acceptance, so the initiative of using the carbon market to promote energy conservation and emission reduction has been used by many countries and regions around the world.
[0003] Fluctuations in carbon market trading prices will be transmitted to other traditional physical markets through the supply chain and cause uncertainty in related industry price indices.
[0004] The adjustment of policy orientation and the actions of all parties involved in the carbon trading market are of great significance to the operation of the carbon trading market. However, in the existing technology, there is no systematic analysis method or analysis model for the factors affecting the fluctuations of carbon market trading prices, and it is impossible to accurately analyze the driving factors of the fluctuations of carbon market trading price series in different periods. Summary of the invention
[0005] Purpose of the invention: The purpose of the present invention is to provide an accurate and directly applicable method and system for analyzing factors affecting carbon market transaction prices based on an adaptive neuro-fuzzy inference system model.
[0006] Technical solution: The method for analyzing factors affecting carbon market transaction prices based on an adaptive neural fuzzy inference system model described in the present invention specifically includes the following steps:
[0007] S1. Obtain a series of instance data on the closing price of carbon market transactions and factors affecting the carbon market transaction price;
[0008] S2. Establish an index system of factors influencing carbon market transaction prices based on the example data obtained in S1, and obtain a data column of the index of factors influencing carbon market transaction prices;
[0009] S3. To explore the volatility of carbon market transaction prices, descriptive statistics are conducted on the carbon market transaction price series;
[0010] S4. Based on the volatility characteristics of the carbon market transaction price series in S3, the fully integrated empirical mode decomposition method with adaptive noise is used to obtain the decomposition series of carbon market transaction prices;
[0011] S5. Based on the decomposition sequence of carbon market transaction prices in S4 and the data column of carbon market transaction price influencing factor indicators in S2, an adaptive neural fuzzy inference system model is established;
[0012] S6, performing a performance test on the adaptive neuro-fuzzy inference system model established in S5 based on mean square error;
[0013] S7, analyzing the influencing factors of the transaction price decomposition sequences of each carbon market obtained in S4 according to the adaptive neural fuzzy inference system model established in S5;
[0014] S8. Determine the influencing factors of the decomposition sequence of different carbon market transaction prices in S7 based on the trend analysis method.
[0015] Specifically, the carbon market transaction price influencing factor index system in step S2 includes energy prices, macroeconomic shocks, international carbon futures prices, financial variables, and weather conditions.
[0016] Specifically, the energy prices include energy commodity price index, natural gas futures closing price, thermal coal index, and Brent crude oil price.
[0017] Specifically, the macroeconomic shocks include the CSI 300 Index, the Shanghai Composite Index, the High Grade Green Bond Index, the China Green Bond Index, the China Bond Composite Index, and the Euro Stoxx 50 Price Index.
[0018] Specifically, the international carbon futures price includes the European carbon emission quota price; the financial variable includes the exchange rate of the US dollar to the RMB; and the weather conditions include the air quality index and the weather futures heating day index.
[0019] Specifically, the calculation method for descriptive statistics of the carbon market transaction price series in step S3 is:
[0020] The ADF test is used for the carbon market transaction price series, and the calculation formula is:
[0021]
[0022] Where: P t is the carbon market transaction price at time t; α is a constant term; β is a trend term; θ represents P t-1 The coefficient of represents the difference term; ε t is a random error term.
[0023] The core is to conduct a hypothesis test on θ, that is, to test whether θ is equal to 1. If θ = 1, the null hypothesis cannot be rejected, that is, the time series data has a unit root, indicating that the data is unstable and there is random drift; if θ < 1, the null hypothesis can be rejected, that is, the time series data does not have a unit root, indicating that the data tends to be stable.
[0024] Specifically, the specific steps of using the fully integrated empirical mode decomposition method with adaptive noise to obtain each decomposition sequence of the carbon market transaction price in step S4 include:
[0025] S401, add white noise to get a new time series:
[0026] f i (t)=f(t)+ε0ω i (t), i=1,2,...,n.
[0027] Where: f i (t) is the new sequence, f(t) is the original data sequence of carbon market transaction prices; ω i (t) is Gaussian white noise that conforms to the standard normal distribution; ε0 is the initial scalar coefficient;
[0028] S402, decompose f using empirical mode decomposition i (t) sequence n times, and calculate the average value to get the first eigenmode function sequence of CEEMDAN algorithm:
[0029]
[0030]
[0031] Where: is the first intrinsic mode function sequence, EMD1(·) represents the first IMF sequence obtained by EMD decomposition; r1(t) is the first residual;
[0032] S403: Add the adaptive noise term to obtain a new time series:
[0033]
[0034]
[0035] Where: is the second eigenmode function sequence; r2(t) is the second residual; ε1 is the first scalar coefficient;
[0036] S404: Add the new adaptive noise term to the new residual with white noise:
[0037]
[0038] k = 2, 3, ..., k.
[0039]
[0040] Where: is the kth eigenmode function sequence; r k-1(t) is the k-1th residual; r k (t) is the kth residual;
[0041] S405: Finally, when the residual does not exceed two extreme points, that is, it becomes a monotonic signal and cannot be further decomposed, the CEEMDAN algorithm terminates and decomposes into k IMF sequences:
[0042]
[0043] Where: R(t) is the residual term obtained by the final decomposition.
[0044] Specifically, the specific steps of establishing the adaptive neural fuzzy inference system model in step S5 include:
[0045] S501: In the first layer, the input variables are fuzzified and output according to the membership of the fuzzy set. Each node k in this layer is an adaptive node with a node function:
[0046] P 1,k =μ Ak (ε), k = 1, 2, ..., n
[0047] Where: P 1,k is the membership degree of the fuzzy set; μ Ak is a membership function; ε is an input variable, namely, the data column of the carbon market transaction price influencing factor index in step S2 and the decomposition sequence of the carbon market transaction price in step S4; k is the number of specific membership functions and forms selected according to actual conditions;
[0048] S502: The second layer implements all input signal operations. Each node in this layer is a fixed node, and its output is the algebraic product of all input signals:
[0049] P 2,k =ω k =μ Ak (ε)·μ Bk (ε), k = 1, 2, ..., n
[0050] Where: k and P 2,k For the applicability of the rules;
[0051] S503: The third layer regulates the incentive strength of each rule. The nodes in this layer are also fixed nodes:
[0052]
[0053] Where: P 3,k and is the normalized fitness;
[0054] S504: The fourth layer is defuzzification, that is, calculating the output of each rule. Each node k in this layer is an adaptive node with a node function:
[0055]
[0056] Where P 4,k is the output layer; k Representative rules;
[0057] S505: The fifth layer is a single node that calculates the total output of the sum of all input signals:
[0058]
[0059] Where: P 5,k is the total output obtained by summing the output layers in S504.
[0060] Specifically, in step S6, the performance test of the adaptive neuro-fuzzy inference system model established in step S5 adopts the mean square error (MSE):
[0061]
[0062] Where: MSE is the mean square error, n is the number of measurements, y t is the kth measurement value, is the kth value estimated by the model.
[0063] Specifically, the decomposition sequence of each carbon market transaction price in step S7 refers to inputting each decomposition sequence of the carbon market transaction price obtained in step S4 into the adaptive neural fuzzy inference system model constructed in step S5.
[0064] A carbon market transaction price influencing factor analysis system is constructed by applying a carbon market transaction price influencing factor analysis method based on an adaptive neuro-fuzzy inference system model.
[0065] Beneficial effects: Compared with the prior art, the present invention has the following significant effects:
[0066] 1. The present invention obtains five types of indicators by sorting out, screens and processes the example data under each type of indicator, and constructs an indicator system of factors affecting carbon market transaction prices, thereby improving the accuracy and pertinence of the analysis of influencing factors.
[0067] 2. The present invention uses descriptive statistical methods to process the original data sequence of carbon market transaction prices, characterizes the volatility characteristics of the original sequence of carbon market transaction prices, and effectively solves the problem of large errors in the price sequence analysis effect of the common linear method in most studies.
[0068] 3. After analyzing the volatility characteristics of the original sequence of carbon market transaction prices, the present invention uses a fully integrated empirical mode decomposition method with adaptive noise to decompose the original sequence of carbon prices, thereby effectively reducing noise.
[0069] 4. The adaptive neuro-fuzzy inference system model finally established by the present invention can be directly used for analyzing the factors affecting the carbon market transaction price in different time periods. BRIEF DESCRIPTION OF THE DRAWINGS
[0070] Figure 1 It is the overall flow chart of the analytical method of the present invention.
[0071] Figure 2 It is a schematic diagram of a specific implementation mode of the present invention.
[0072] Figure 3 4 is a flow chart of step S4 of the analysis method of the present invention.
[0073] Figure 4 4 is a flow chart of step S5 of the analysis method of the present invention. DETAILED DESCRIPTION
[0074] The present invention is further explained below in conjunction with the accompanying drawings and specific implementation methods.
[0075] See also Figure 1 As shown, the present invention discloses a method for analyzing factors affecting carbon market transaction prices based on an adaptive neural fuzzy inference system model, which specifically includes the following steps:
[0076] S1. Obtain a series of instance data on the closing price of carbon market transactions and factors affecting the carbon market transaction price;
[0077] S2. Establish an index system of factors influencing carbon market transaction prices based on the example data obtained in S1, and obtain a data column of the index of factors influencing carbon market transaction prices;
[0078] S3. To explore the volatility of carbon market transaction prices, descriptive statistics are conducted on the carbon market transaction price series;
[0079] S4. Based on the volatility characteristics of the carbon market transaction price series in S3, the fully integrated empirical mode decomposition method with adaptive noise is used to obtain the decomposition series of carbon market transaction prices;
[0080] S5. Based on the decomposition sequence of carbon market transaction prices in S4 and the data column of carbon market transaction price influencing factor indicators in S2, an adaptive neural fuzzy inference system model is established;
[0081] S6, performing a performance test on the adaptive neuro-fuzzy inference system model established in S5 based on mean square error;
[0082] S7, analyzing the influencing factors of the transaction price decomposition sequences of each carbon market obtained in S4 according to the adaptive neural fuzzy inference system model established in S5;
[0083] S8. Determine the influencing factors of the decomposition sequence of different carbon market transaction prices in S7 based on the trend analysis method, and put forward countermeasures and suggestions to all participating parties.
[0084] See also Figure 2 As shown, the carbon market transaction price influencing factor indicators in step S2 include energy prices, macroeconomic shocks, international carbon futures prices, financial variables, weather conditions, etc. The energy prices include energy commodity price index, natural gas futures closing price, thermal coal index, Brent crude oil price; the macroeconomic shocks include the CSI 300 Index, the Shanghai Composite Index, the High-Grade Green Bond Index, the China Green Bond Index, the China Bond Composite Index, and the European Stoxx 50 Price Index; the international carbon futures prices include European carbon emission quota prices; the financial variables include the US dollar to RMB exchange rate; the weather conditions include the air quality index and the weather futures heating day index.
[0085] The specific data collected for the above-mentioned carbon market transaction price influencing factor indicators are screened and cleaned for subsequent analysis and calculation.
[0086] Specifically, the calculation method for descriptive statistics of the carbon market transaction price series in step S3 is:
[0087] The ADF test is used for the carbon market transaction price series, and the calculation formula is:
[0088]
[0089] Where: P t is the carbon market transaction price at time t; α is a constant term; β is a trend term; θ represents P t-1 The coefficient of represents the difference term; ε t is a random error term.
[0090] The core is to conduct a hypothesis test on θ, that is, to test whether θ is equal to 1. If θ = 1, the null hypothesis cannot be rejected, that is, the time series data has a unit root, indicating that the data is unstable and there is random drift; if θ < 1, the null hypothesis can be rejected, that is, the time series data does not have a unit root, indicating that the data tends to be stable.
[0091] See also Figure 3 As shown, the specific steps of using the fully integrated empirical mode decomposition method with adaptive noise to obtain each decomposition sequence of the carbon market transaction price in step S4 include:
[0092] S401, add white noise to get a new time series:
[0093] f i (t)=f(t)+ε0ω i (t), i = 1, 2, ..., n.
[0094] Where: f i (t) is the new sequence, f(t) is the original data sequence of carbon market transaction prices; ω i (t) is Gaussian white noise that conforms to the standard normal distribution; ε0 is the initial scalar coefficient;
[0095] S402, decompose f using empirical mode decomposition i (t) sequence n times, and calculate the average value to get the first eigenmode function sequence of CEEMDAN algorithm:
[0096]
[0097]
[0098] Where: is the first intrinsic mode function sequence, EMD1(·) represents the first IMF sequence obtained by EMD decomposition; r1(t) is the first residual;
[0099] S403: Add the adaptive noise term to obtain a new time series:
[0100]
[0101]
[0102] Where: is the second eigenmode function sequence; r2(t) is the second residual; ε1 is the first scalar coefficient;
[0103] S404: Add the new adaptive noise term to the new residual with white noise:
[0104]
[0105] k = 2, 3, ..., k.
[0106]
[0107] Where: is the kth eigenmode function sequence; r k-1 (t) is the k-1th residual; r k (t) is the kth residual;
[0108] S405: Finally, when the residual does not exceed two extreme points, that is, it becomes a monotonic signal and cannot be further decomposed, the CEEMDAN algorithm terminates and decomposes into k IMF sequences:
[0109]
[0110] Where: R(t) is the residual term obtained by the final decomposition.
[0111] See also Figure 4 As shown, the specific steps of establishing the adaptive neural fuzzy inference system model in step S5 include:
[0112] S501: In the first layer, the input variables are fuzzified and output according to the membership of the fuzzy set. Each node k in this layer is an adaptive node with a node function:
[0113] P 1,k =μ Ak (ε), k = 1, 2, ..., n
[0114] Where: P 1,k is the membership degree of the fuzzy set; μ Ak is a membership function; ε is an input variable, namely, the data column of the carbon market transaction price influencing factor index in step S2 and the decomposition sequence of the carbon market transaction price in step S4; k is the number of specific membership functions and forms selected according to actual conditions;
[0115] S502: The second layer implements all input signal operations. Each node in this layer is a fixed node, and its output is the algebraic product of all input signals:
[0116] P 2,k =ω k =μ Ak (ε)·μ Bk (ε), k = 1, 2, ..., n
[0117] Where: k and P 2,k For the applicability of the rules;
[0118] In addition, the output of each node represents the regular excitation intensity; the node function of this layer can also take the form of small, bounded or strong product.
[0119] S503: The third layer regulates the incentive strength of each rule. The nodes in this layer are also fixed nodes:
[0120]
[0121] Where: P 3,k and is the normalized fitness;
[0122] S504: The fourth layer is defuzzification, that is, calculating the output of each rule. Each node k in this layer is an adaptive node with a node function:
[0123]
[0124] Where P 4,k is the output layer; k Representative rules;
[0125] S505: The fifth layer is a single node that calculates the total output, which is the sum of all input signals:
[0126]
[0127] Where: P 5,k is the total output obtained by summing the output layers in S504.
[0128] In step S6, the performance test of the adaptive neuro-fuzzy inference system model established in step S5 is performed using mean square error (MSE):
[0129]
[0130] Where: MSE is the mean square error, n is the number of measurements, y t is the kth measurement value, is the kth value estimated by the model.
[0131] Decomposing the transaction price sequences of each carbon market in step S7 refers to inputting the k IMF sequences obtained in step S4 into the adaptive neuro-fuzzy inference system model constructed in step S5 respectively.
[0132] Based on the above-mentioned method for analyzing factors affecting carbon market transaction prices based on the adaptive neuro-fuzzy inference system model, a system for analyzing factors affecting carbon market transaction prices is constructed.
Claims
1. A method for analyzing factors affecting carbon market transaction prices based on an adaptive neural fuzzy inference system model, characterized in that: The following steps are involved: S1. Obtain a series of instance data on the closing price of carbon market transactions and factors affecting the carbon market transaction price; S2. Establish an index system of factors influencing carbon market transaction prices based on the example data obtained in S1, and obtain a data column of the index of factors influencing carbon market transaction prices; S3. To explore the volatility of carbon market transaction prices, descriptive statistics are conducted on the carbon market transaction price series; S4. Based on the volatility characteristics of the carbon market transaction price series in S3, the fully integrated empirical mode decomposition method with adaptive noise is used to obtain the decomposition series of carbon market transaction prices; S5. Based on the decomposition sequence of carbon market transaction prices in S4 and the data column of carbon market transaction price influencing factor indicators in S2, an adaptive neural fuzzy inference system model is established; S6, performing a performance test on the adaptive neuro-fuzzy inference system model established in S5 based on mean square error; S7, analyzing the influencing factors of the transaction price decomposition sequences of each carbon market obtained in S4 according to the adaptive neural fuzzy inference system model established in S5; S8. Determine the influencing factors of the decomposition sequence of different carbon market transaction prices in S7 based on the trend analysis method.
2. The method for analyzing factors affecting carbon market transaction prices according to claim 1 is characterized by: The carbon market transaction price influencing factor index system described in step S2 includes energy prices, macroeconomic shocks, international carbon futures prices, financial variables, and weather conditions.
3. The method for analyzing factors affecting carbon market transaction prices according to claim 2 is characterized by: The energy prices include energy commodity price index, natural gas futures closing price, thermal coal index, and Brent crude oil price; The macroeconomic shocks include the CSI 300 Index, the Shanghai Composite Index, the High Grade Green Bond Index, the China Green Bond Index, the China Bond Composite Index, and the Euro Stoxx 50 Price Index.
4. The method for analyzing factors affecting carbon market transaction prices according to claim 2 is characterized by: The international carbon futures prices include the European carbon emission quota prices; The financial variables include the exchange rate between the U.S. dollar and the RMB; The weather conditions include an air quality index and a weather futures heating day index.
5. The method for analyzing factors affecting carbon market transaction prices according to claim 1, characterized in that: The calculation method for descriptive statistics of the carbon market transaction price series in step S3 is: The ADF test is used for the carbon market transaction price series, and the calculation formula is: Where: P t is the carbon market transaction price at time t; α is a constant term; β is a trend term; θ represents P t-1 The coefficient of represents the difference term; ε t is the random error term; This method is to perform a hypothesis test on θ, that is, to test whether θ is equal to 1; if θ = 1, the null hypothesis cannot be rejected, that is, the time series data has a unit root, indicating that the data is unstable and there is random drift; if θ < 1, the null hypothesis can be rejected, that is, the time series data does not have a unit root, indicating that the data tends to be stable.
6. The method for analyzing factors affecting carbon market transaction prices according to claim 1, characterized in that: The specific steps of using the fully integrated empirical mode decomposition method with adaptive noise to obtain each decomposition sequence of the carbon market transaction price in step S4 include: S401, add white noise to get a new time series: f i (t)=f(t)+ε0ω i (t),i=1,2,...,n. Where: f i (t) is the new sequence, f(t) is the original data sequence of carbon market transaction prices; ω i (t) is Gaussian white noise that conforms to the standard normal distribution; ε0 is the initial scalar coefficient; S402, decompose f using empirical mode decomposition i (t) sequence n times, and calculate the average value to get the first eigenmode function sequence of CEEMDAN algorithm: Where: is the first intrinsic mode function sequence, EMD1(·) represents the first IMF sequence obtained by EMD decomposition; r1(t) is the first residual; S403: Add the adaptive noise term to obtain a new time series: Where: is the second eigenmode function sequence; r2(t) is the second residual; ε1 is the first scalar coefficient; S404: Add the new adaptive noise term to the new residual with white noise: Where: is the kth eigenmode function sequence; r k-1 (t) is the k-1th residual; r k (t) is the kth residual; S405: Finally, when the residual does not exceed two extreme points, that is, it becomes a monotonic signal and cannot be further decomposed, the CEEMDAN algorithm terminates and decomposes into k IMF sequences. These k IMF sequences have the following relationship: Where: R(t) is the residual term obtained by the final decomposition.
7. The method for analyzing factors affecting carbon market transaction prices according to claim 1, characterized in that: The specific steps of establishing the adaptive neuro-fuzzy inference system model in step S5 include: S501: In the first layer, the input variables are fuzzified and output according to the membership of the fuzzy set. Each node k in this layer is an adaptive node with a node function: P 1,k =μ Ak (e),k=1,2,…,n Where: P 1,k is the membership degree of the fuzzy set; μ Ak is a membership function; ε is an input variable, namely, the data column of the carbon market transaction price influencing factor index in step S2 and the decomposition sequence of the carbon market transaction price in step S4; k is the number of specific membership functions and forms selected according to actual conditions; S502: The second layer implements all input signal operations. Each node in this layer is a fixed node, and its output is the algebraic product of all input signals: P 2,k =ω k =μ Ak (e)·m Bk (e),k=1,2,…,n Where: k and P 2,k For the applicability of the rules; S503: The third layer regulates the incentive strength of each rule. The nodes in this layer are also fixed nodes: Where: P 3,k and is the normalized fitness; S504: The fourth layer is defuzzification, that is, calculating the output of each rule. Each node k in this layer is an adaptive node with a node function: Where P 4,k is the output layer; k Representative rules; S505: The fifth layer is a single node that calculates the total output of the sum of all input signals: Where: P 5,k is the total output obtained by summing the output layers in S504.
8. The method for analyzing factors affecting carbon market transaction prices according to claim 1, characterized in that: In step S6, the performance test of the adaptive neuro-fuzzy inference system model established in step S5 is performed using mean square error: Where: MSE is the mean square error, n is the number of measurements, y t is the kth measurement, is the kth value estimated by the model.
9. The method for analyzing factors affecting carbon market transaction prices according to claim 1, characterized in that: The step S7 analyzes the influencing factors of each carbon market transaction price decomposition sequence respectively, which means that each decomposition sequence of the carbon market transaction price obtained in step S4 is input into the adaptive neural fuzzy inference system model constructed in step S5 respectively.
10. A carbon market transaction price influencing factors analysis system constructed by applying the carbon market transaction price influencing factors analysis method described in claims 1-9.
