Carbon quota transaction-based CCER transaction prediction method
By constructing multiple prediction models, using historical data of carbon quota trading and CCER trading, the problem of difficulty in predicting CCER trading price is solved, and accurate prediction of CCER trading price and trading timing is achieved, and the company's risk of carbon assets is improved. Control ability.
Patent Information
- Application Number
- CN202410982988.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-12-10
AI Technical Summary
It is difficult to predict the CCER trading price. Existing methods rarely specifically predict the CCER trading price, and discontinuous transactions lead to difficulties in prediction.
By obtaining historical data of carbon quota trading and CCER trading, multiple prediction models are constructed, including the next-day trading price prediction model based on the CNN network and the next-day trading prediction model of the BP network, as well as the same-day price prediction model based on the BP network. Improve the accuracy of predictions.
It realizes accurate prediction of CCER transaction prices and trading timing, helps enterprises better manage carbon asset risks, and improves the accuracy of predictions and multi-faceted control capabilities.
Smart Images
Figure CN119107119A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of carbon trading, and in particular to a CCER trading prediction method based on carbon quota trading. Background Art
[0002] As the global climate change problem becomes increasingly serious, countries have taken measures to reduce greenhouse gas emissions, and the carbon trading market has emerged. China Certified Emission Reduction (CCER), as China's unique voluntary greenhouse gas emission reduction mechanism, plays an important role in promoting energy conservation and emission reduction and promoting green and low-carbon development.
[0003] However, the current CCER market transactions are different from traditional stock transactions or carbon quota transactions. CCER transactions are not continuous. There are CCER transactions on some trading days, but not on other trading days. This makes it difficult to predict CCER transactions. At the same time, although the price fluctuations of CCER transactions are affected by many factors, they are always strongly correlated with the transaction price of carbon quotas. Although there are many methods to predict carbon prices, there are few methods that specifically predict the transaction price of CCER.
[0004] Therefore, in order to achieve accurate prediction of CCER transaction prices, grasp the trading opportunities of CCER, and help enterprises better manage their carbon asset risks, we proposed a CCER transaction prediction method based on carbon quota trading, which can make accurate predictions for discontinuous CCER transaction prices. Summary of the invention
[0005] 1. Technical problems to be solved: In response to the above technical problems, the present invention provides a CCER transaction prediction method based on carbon quota trading. By screening specific data and data processing methods, and using BP neural network prediction model and CNN network prediction model, the CCER transaction and transaction price can be predicted on the same day or the next day, thereby improving the prediction accuracy.
[0006] 2. Technical solution: A CCER transaction prediction method based on carbon quota trading, characterized in that it includes the following steps: Step 1: Obtain historical data on carbon quota transactions and CCER transactions and construct a historical data set; the historical data set is composed of carbon quota transactions and CCER transaction-related data on multiple different trading days with or without CCER transactions.
[0007] Step 2: Sort the acquired historical data to generate data sets S1 and S2; data set S1 is used to predict the price of the next day; data set S2 is used to predict whether there will be a transaction on the next day.
[0008] Step 3: Construct the next-day transaction price prediction model A1 based on the CNN network, and use the data set S1 to train the model to obtain the trained next-day transaction price prediction model A1.
[0009] Step 4: Construct a prediction model A2 for whether to trade the next day based on the BP network, and use the data set S2 to train the model to obtain the trained prediction model A2 for whether to trade the next day.
[0010] Step 5: Use the trained transaction price prediction model A1 to predict the historical data S1, and construct the data set S3 with the obtained prediction values.
[0011] Step 6: Construct a daily price prediction model based on the BP network, and use the data set S3 to train to obtain a trained daily price prediction model.
[0012] Furthermore, the carbon quota trading related data in step one include: the daily opening price, highest price, lowest price, closing price, average transaction price, increase, and total carbon emission quota trading volume of the national carbon market; the CCER trading related data include the daily average transaction price of the national CCER market; the daily data in the historical data set include the daily opening price, highest price, lowest price, closing price, average transaction price, increase, total carbon emission quota trading volume, and the daily average transaction price of the national CCER market.
[0013] Furthermore, in step 2, the process of sorting the historical data sets to obtain the data sets S1 and S2 is specifically as follows: S21: Based on whether there is CCER transaction on the trading day, the corresponding date in the daily data set in the historical data set where there is transaction is defined as data set D1, and the national CCER market average transaction price corresponding to each date in the historical data set is obtained to obtain data set Y1; the date without transaction is defined as data set D2.
[0014] S22: Take each data in the data set Y1 as a label and correspond it to the data set D1, and then obtain the opening price, highest price, lowest price, closing price, average transaction price, and increase data of the national carbon market in the carbon quota trading history data of the previous five days corresponding to each label, construct them into a matrix and normalize them, and use the normalized matrix data as the training data X1; correspond the data set X1 with the elements in the data set Y1 and generate the data set S1.
[0015] S23: Based on whether there is CCER transaction on the trading day, the trading day corresponding to the data set D1 is marked as 1, and the trading day corresponding to the data set D2 is marked as 0, and the marked data is made into a label data set Y2 indicating whether there is transaction.
[0016] S24: Normalize the average transaction price, closing price, increase, and total transaction volume of the national carbon market in the carbon quota trading history data of the day before the trading day corresponding to each data in the data set Y2, and use the normalized data as the training data set X2; the data set X2 and the label data set Y2 of the corresponding trading day jointly construct the data set S2.
[0017] Furthermore, the training process in step three is: the data set S1 is passed into the next day transaction price prediction model A1 for training until the prediction value of the training data set converges or reaches a preset number of iterations.
[0018] Furthermore, the training process in step 4 is: the data set S2 is passed into the prediction model A2 for whether there will be a transaction the next day for training until the prediction value of the training data set converges or reaches a preset number of iterations.
[0019] Furthermore, the training process in step 5 is: S51: Use the trained transaction price prediction model A1 to predict the data set X1 to obtain the CCER transaction price prediction value for each trading day in the data set D1.
[0020] S52: Normalize the predicted value obtained in step 51 and the opening price, highest price, lowest price, closing price, average transaction price, and increase data of the national carbon market in the carbon quota trading history data of the trading day corresponding to the data set D1.
[0021] S53: The normalized data set obtained in step S52 is used as the training data set X3, and X3 and Y1 are used together to construct the data set S3 based on the trading day.
[0022] Furthermore, the training process in step six is: the data set S3 is passed into the said daily price prediction model A3 for training until the prediction value of the training data set converges or reaches a preset number of iterations.
[0023] 3. Beneficial effects: (1) This method discloses a CCER transaction prediction method based on carbon quota trading. By constructing three models with different purposes for training, it can not only use carbon price trading data to predict whether CCER will be traded, but also predict the transaction price, thereby improving the company's control over CCER transaction risks.
[0024] (2) This method discloses a CCER transaction prediction method based on carbon quota trading. The method constructs a system that is responsible for predicting whether CCER trading will occur on the second day, a system that is responsible for predicting the price of CCER trading on the second day, and a system that is responsible for making real-time predictions of the CCER trading price on the current day. The accuracy of the prediction is ensured from multiple aspects. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Figure 1 A flow chart of a CCER transaction prediction method based on carbon quota trading according to the present invention; Figure 2 Schematic diagram of the process of obtaining data sets S1 and S2 in the present invention; Figure 3 Schematic diagram of the construction of the training matrix in this method; Figure 4 Schematic diagram of the process of obtaining the data set S3 in this method. DETAILED DESCRIPTION
[0026] The present invention will be described in detail below with reference to the accompanying drawings.
[0027] As attached Figure 1 As shown, a CCER transaction prediction method based on carbon quota trading is characterized by comprising the following steps: Step 1: Obtain historical data on carbon quota transactions and CCER transactions and construct a historical data set; the historical data set is composed of carbon quota transactions and CCER transaction-related data on multiple different trading days with or without CCER transactions.
[0028] Step 2: Sort the acquired historical data to generate data sets S1 and S2; data set S1 is used to predict the price of the next day; data set S2 is used to predict whether there will be a transaction on the next day.
[0029] Step 3: Construct the next-day transaction price prediction model A1 based on the CNN network, and use the data set S1 to train the model to obtain the trained next-day transaction price prediction model A1.
[0030] Step 4: Construct a prediction model A2 for whether to trade the next day based on the BP network, and use the data set S2 to train the model to obtain the trained prediction model A2 for whether to trade the next day.
[0031] Step 5: Use the trained transaction price prediction model A1 to predict the historical data S1, and construct the data set S3 with the obtained prediction values.
[0032] Step 6: Construct a daily price prediction model based on the BP network, and use the data set S3 to train to obtain a trained daily price prediction model.
[0033] Furthermore, the carbon quota trading related data in step one include: the daily opening price, highest price, lowest price, closing price, average transaction price, increase, and total carbon emission quota trading volume of the national carbon market; the CCER trading related data include the daily average transaction price of the national CCER market; the daily data in the historical data set include the daily opening price, highest price, lowest price, closing price, average transaction price, increase, total carbon emission quota trading volume, and the daily average transaction price of the national CCER market.
[0034] Furthermore, as attached Figure 2 As shown, the process of sorting the historical data set in step 2 to obtain data sets S1 and S2 is specifically as follows: S21: Based on whether there is CCER transaction on the trading day, the corresponding date in the daily data set in the historical data set where there is transaction is defined as data set D1, and the national CCER market average transaction price corresponding to each date in the historical data set is obtained to obtain data set Y1; the date without transaction is defined as data set D2.
[0035] Taking the trading days in June 2024 as an example, the trading days in June include [6.3, 6.4, 6.5, 6.6, 6.7, 6.11, 6.12, 6.13, 6.14, 6.17, 6.18, 6.19, 6.20, 6.21, 6.24, 6.25, 6.26, 6.27, 6.28], among which the dates with CCER transactions are [6.4, 6.5, 6.11, 6.12, 6.19, 6.20]. The data of these days are classified as data set D1, and the rest of the data are classified as data set D2.
[0036] S22: Take each data in the data set Y1 as a label and correspond it to the data set D1, and then obtain the opening price, highest price, lowest price, closing price, average transaction price, and increase data of the national carbon market in the carbon quota trading history data of the previous five days corresponding to each label, construct them into a matrix and normalize them, and use the normalized matrix data as the training data X1; correspond the data set X1 with the elements in the data set Y1 and generate the data set S1.
[0037] like Figure 3 As shown, taking June 19 as an example, the carbon quota trading days before it include [6.12, 6.13, 6.14, 6.17, 6.18]. The opening price, highest price, lowest price, closing price, average transaction price, and increase data of the national carbon market on these five days are constructed into a matrix.
[0038] S23: Based on whether there is CCER transaction on the trading day, the trading day corresponding to the data set D1 is marked as 1, and the trading day corresponding to the data set D2 is marked as 0, and the marked data is made into a label data set Y2 indicating whether there is transaction.
[0039] S24: Normalize the average transaction price, closing price, increase, and total transaction volume of the national carbon market in the carbon quota trading history data of the day before the trading day corresponding to each data in the data set Y2, and use the normalized data as the training data set X2; the data set X2 and the label data set Y2 of the corresponding trading day jointly construct the data set S2.
[0040] Furthermore, the training process in step three is: the data set S1 is passed into the next day transaction price prediction model A1 for training until the prediction value of the training data set converges or reaches a preset number of iterations.
[0041] Furthermore, the training process in step 4 is: the data set S2 is passed into the prediction model A2 for whether there will be a transaction the next day for training until the prediction value of the training data set converges or reaches a preset number of iterations.
[0042] Furthermore, as attached Figure 4 As shown, the training process in step five is: S51: Use the trained transaction price prediction model A1 to predict the data set X1 to obtain the CCER transaction price prediction value for each trading day in the data set D1.
[0043] S52: Normalize the predicted value obtained in step 51 and the opening price, highest price, lowest price, closing price, average transaction price, and increase data of the national carbon market in the carbon quota trading history data of the trading day corresponding to the data set D1.
[0044] S53: The normalized data set obtained in step S52 is used as the training data set X3, and X3 and Y1 are used together to construct the data set S3 based on the trading day.
[0045] Furthermore, the training process in step six is: the data set S3 is passed into the said daily price prediction model A3 for training until the prediction value of the training data set converges or reaches a preset number of iterations.
[0046] Using the above method, three prediction models were established respectively. Model A1 is responsible for predicting the CCER transaction price on the second day, model A2 is responsible for predicting whether CCER transaction will occur on the second day, and model A3 is responsible for real-time prediction of the CCER transaction price on the day; the data transmitted to A3 include the predicted value of the CCER transaction price on the day by A1 and the opening price, highest price, lowest price, real-time price, average transaction price, and increase data of the national carbon market in the carbon quota trading history data on D1.
[0047] Although the present invention has been disclosed as above in terms of preferred embodiments, they are not intended to limit the present invention. Anyone skilled in the art can make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection defined by the claims of this application.
Claims
1. A CCER transaction prediction method based on carbon quota trading, characterized by: The following steps are involved: Step 1: Obtain historical data on carbon quota transactions and CCER transactions and construct a historical data set; the historical data set consists of carbon quota transactions and CCER transaction-related data on multiple different trading days with or without CCER transactions; Step 2: Sort the acquired historical data to generate data sets S1 and S2; data set S1 is used to predict the price of the next day; Dataset S2 is a dataset used to predict whether there will be a transaction on the next day; Step 3: Construct the next day transaction price prediction model A1 based on the CNN network, and use the data set S1 to train the model to obtain the trained next day transaction price prediction model A1; Step 4: Construct a prediction model A2 for whether to trade the next day based on the BP network, and use the data set S2 to train the model to obtain the trained prediction model A2 for whether to trade the next day; Step 5: Use the trained transaction price prediction model A1 to predict the historical data S1, and construct the data set S3 with the obtained prediction values; Step 6: Construct a daily price prediction model based on the BP network, and use the data set S3 to train to obtain a trained daily price prediction model.
2. A CCER transaction prediction method based on carbon quota trading according to claim 1, characterized in that: The carbon quota trading related data in step one include: the daily opening price, highest price, lowest price, closing price, average transaction price, increase, and total carbon emission quota trading volume of the national carbon market; the CCER trading related data include the daily average transaction price of the national CCER market; the daily data in the historical data set include the daily opening price, highest price, lowest price, closing price, average transaction price, increase, total carbon emission quota trading volume, and the daily average transaction price of the national CCER market.
3. A CCER transaction prediction method based on carbon quota trading according to claim 2, characterized in that: The specific process of sorting the historical data sets in step 2 to obtain data sets S1 and S2 is as follows: S21: According to whether there is CCER transaction on the trading day, the corresponding date in the daily data set in the historical data set with transaction is defined as data set D1, and the national CCER market transaction average price corresponding to each date in the historical data set is obtained to obtain data set Y1; the date without transaction is defined as data set D2; S22: Take each data in the data set Y1 as a label, correspond it to the data set D1, and then obtain the opening price, highest price, lowest price, closing price, average transaction price, and increase data of the national carbon market in the carbon quota trading history data of the previous five days corresponding to each data label to construct a matrix and normalize it, and use the normalized matrix data as training data X1; correspond the data set X1 with the elements in the data set Y1 to generate the data set S1; S23: Based on whether CCER transactions exist on the trading day, the trading day corresponding to the data set D1 is marked as 1, and the trading day corresponding to the data set D2 is marked as 0, and the marked data is made into a label data set Y2 indicating whether transactions exist; S24: normalize the average transaction price, closing price, increase, and total transaction volume of carbon emission quotas in the national carbon market in the carbon quota transaction history data of the previous trading day corresponding to each data in the data set Y2, and use the normalized data as the training data set X2; The dataset X2 and the label dataset Y2 of the corresponding trading day together construct the dataset S2.
4. A CCER transaction prediction method based on carbon quota trading according to claim 3, characterized in that: The training process in step three is: the data set S1 is passed into the next day transaction price prediction model A1 for training until the prediction value of the training data set converges or reaches a preset number of iterations.
5. A CCER transaction prediction method based on carbon quota trading according to claim 4, characterized in that: The training process in step 4 is: the data set S2 is passed into the prediction model A2 for whether to trade the next day for training until the prediction value of the training data set converges or reaches a preset number of iterations.
6. A CCER transaction prediction method based on carbon quota trading according to claim 5, characterized in that: The training process in step 5 is: S51: Use the trained transaction price prediction model A1 to predict the data set X1, and obtain the CCER transaction price prediction value of each trading day in the data set D1; S52: normalize the predicted value obtained in step 51 and the opening price, highest price, lowest price, closing price, average transaction price, and increase data of the national carbon market in the carbon quota trading history data of the trading day corresponding to the data set D1; S53: The normalized data set obtained in step S52 is used as the training data set X3, and X3 and Y1 are used together to construct the data set S3 based on the trading day.
7. A CCER transaction prediction method based on carbon quota trading according to claim 6, characterized in that: The training process in step six is: the data set S3 is passed into the said daily price prediction model A3 for training until the prediction value of the training data set converges or reaches a preset number of iterations.