Power consumer carbon emission prediction system and method based on carbon-electricity association
Through the power user carbon emission prediction system based on carbon electricity correlation, the carbon emission prediction model is trained using long and short-term memory models, the problem of insufficient carbon emission prediction in the existing technology is solved, and the accurate prediction of power user carbon emissions and the intelligentization of carbon emission management is achieved.
Patent Information
- Application Number
- CN202411858438.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-17
- Publication Date
- 2025-04-11
AI Technical Summary
The existing carbon emission forecasting methods lack correlation analysis of the behavioral patterns of power users and the dynamic changes of carbon emissions, resulting in inaccurate carbon emission forecasts, making it difficult to adjust the power consumption patterns in real time to cope with carbon emissions over limits and fluctuations.
A carbon emission prediction system for power users based on carbon electricity correlation was designed. The carbon emission correlation coefficient of power users was calculated through the strategy module, the prediction strategy was selected, and the carbon emission prediction model was trained using long and short-term memory models to achieve accurate prediction of power users' carbon emissions.
By combining carbon-electricity correlation characteristics, the accuracy of carbon emission forecasts is improved, helping power users understand the impact of electricity consumption behavior on carbon emissions, and adopt environmental protection and energy-saving measures to achieve the accuracy and intelligence of carbon emission management.
Smart Images
Figure CN119809029A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power system data calculation, and in particular to a system and method for predicting carbon emissions of power users based on carbon-electricity correlation. Background Art
[0002] The consumption of electricity is one of the main sources of carbon emissions for power users. However, due to the imbalance of grid load, the complexity of user electricity consumption behavior and the limitation of grid dispatching capacity, carbon emission intensity often fluctuates and exceeds the limit. Especially in high-energy-consuming enterprises, carbon emission levels are closely related to electricity consumption patterns, but most existing carbon emission prediction methods only focus on the total amount prediction at the macro level, lacking the correlation analysis between power user behavior patterns and dynamic changes in carbon emissions. With power user electricity consumption data as the core and combined with deep learning technology, the accurate prediction of carbon emissions of power users can be effectively achieved by dynamically capturing the deep correlation between electricity consumption and carbon emission levels. Power dispatching rules still have many shortcomings in dealing with carbon emission exceeding the limit and fluctuation problems. When the carbon emission intensity of power users approaches the warning critical value, there is a lack of effective strategies to adjust the power consumption pattern in real time. Summary of the invention
[0003] The purpose of the present invention is to provide a carbon emission prediction system and method for power users based on carbon-electricity correlation. The present invention can optimize the electricity consumption behavior of power users and provide a reference for reducing carbon emissions.
[0004] To achieve this purpose, the present invention designs a carbon emission prediction system for power users based on carbon-electricity correlation, which includes:
[0005] The strategy module is used to use the collected electricity consumption data and carbon emission data of the power users in the selected first historical time period as the power user carbon-electricity correlation prediction data set, calculate the power user carbon-electricity correlation coefficient through the power user carbon-electricity correlation prediction data set, and select the power user carbon emission prediction strategy through the power user carbon-electricity correlation coefficient;
[0006] The carbon emission prediction model module is used to divide the power user carbon-electricity correlation prediction data set into a training set according to the selected power user carbon emission prediction strategy, and train the long-short-term memory model through the long-short-term memory model prediction algorithm to obtain a carbon emission prediction model.
[0007] The carbon emission prediction module is used to input the power consumption data and carbon emission data of the power users in the selected second historical time period into the carbon emission prediction model to perform carbon emission prediction.
[0008] Preferably, the specific method of collecting the power consumption data and carbon emission data of power users in the selected first historical time period includes:
[0009] The data sources of the power consumption data and carbon emission data of power users in the selected first historical time period include carbon emission verification reports of power users, carbon emission monitoring service platforms, data collected by smart meters, monthly electricity bills, and statistical data of power companies;
[0010] The electricity consumption data and carbon emission data of the selected first historical time period to be collected include: the electricity consumption data of the electricity user M years before the forecast month, the electricity user's historical monthly electricity consumption data, a total of n points, X = [x 1 , x 2 , ..., x n ]; The electricity user forecast month M years ago, the electricity user's historical monthly carbon emissions data has a total of n points, C = [c 1 , c 2 , ..., c n ]; X represents the power consumption data of the power users in the selected first historical time period, x n represents the electricity consumption data of the power user in the nth month in the selected first historical time period, C represents the carbon emission data of the power user in the selected first historical time period, c n Represents the carbon emission data of electricity users in the nth month in the selected first historical time period.
[0011] Preferably, the specific method for calculating the carbon-electricity correlation coefficient of power users is:
[0012] Calculate the average value of n points of historical monthly electricity consumption data of electricity users where x i Indicates the first historical period in the selected i Monthly electricity consumption data of electricity users, 1≤i≤n;
[0013] Calculate the average value of n points of historical monthly carbon emissions data of electricity users where c i Indicates the first historical period in the selected i Monthly carbon emission data of electricity users, 1≤i≤n;
[0014] The carbon-electricity correlation coefficient of power users is calculated between the historical monthly power consumption data of power users and the historical monthly carbon emissions data of power users. The calculation formula is as follows:
[0015]
[0016] Among them, P represents the carbon-electricity correlation coefficient of electricity users.
[0017] Preferably, the specific method for selecting the carbon emission prediction strategy of power users through the carbon-electricity correlation coefficient of power users is:
[0018] When P≥0.6, there is a strong correlation between the historical monthly electricity consumption data of power users and the historical monthly carbon emissions data of power users, and the carbon-electricity correlation prediction strategy is used to divide the power user carbon-electricity correlation prediction data set; when P<0.6, there is no strong correlation between the historical monthly electricity consumption data of power users and the historical monthly carbon emissions data of power users, and the time series prediction strategy is used to divide the power user carbon-electricity correlation prediction data set.
[0019] Preferably, the specific method of dividing the power user carbon-electricity correlation prediction data set according to the selected power user carbon emission prediction strategy is:
[0020] When using the carbon-electricity correlation prediction strategy for prediction, the historical monthly electricity consumption data of power users and the historical monthly carbon emissions data of power users are used to divide the data set into a training set and a validation set;
[0021] When using the time series forecasting strategy for forecasting, the historical monthly carbon emissions data of power users are used to divide the data set into training sets and validation sets;
[0022] When the data set is divided into training set and validation set, the data set is divided into training set and validation set according to the number of points of historical monthly carbon emissions data of power users. For the historical monthly carbon emissions data of power users with a total of n points, 0.8*(n-12) is rounded off as the number of points t of the training set, and C train =[c 13 , c 14 , ..., c t+12 ] is used as the true value part of the model output in the training set to obtain the training error, where C train represents the true value of carbon emissions output during the training process of the training set, c t+12 represents the carbon emission data of electricity users in the t+12th month; C val =[c t+13 , c t+14 , ..., c n ] is used as the true value part of the model output in the validation set to obtain the validation error, where C val represents the true value of carbon emissions output during the validation set validation process, c n Represents the carbon emission data of electricity users in the nth month;
[0023] When using the carbon-electricity correlation prediction strategy for prediction, the t+11 points of the n points of historical monthly power consumption data of power users are used. As part of the model input history value in the training set to obtain the training prediction value, where represents the historical value of electricity consumption data output during the training process of the training set, x t+11Represents the electricity consumption data of power users in month t+11; As part of the historical value of the model input in the validation set to obtain the validation prediction value, represents the historical value of electricity consumption data output during the verification process of the verification set, x n-1 Represents the electricity consumption data of power users in month n-1;
[0024] When using the time series forecasting strategy for forecasting, the historical monthly carbon emissions data of power users is taken as the t+11 points out of the n points. As part of the model input history value in the training set to obtain the training prediction value, where represents the historical value of carbon emissions output during the training process of the training set, c t+11 represents the carbon emission data of electricity users in month t+11; As part of the historical value of the model input in the validation set to obtain the validation prediction value, Represents the historical value of carbon emissions output during the validation process of the validation set, c n-1 Represents the carbon emission data of electricity users in month n-1.
[0025] Preferably, the specific process of training the prediction model by the long short-term memory model prediction algorithm is:
[0026] The training set data is input into the long short-term memory model. When the carbon-electricity correlation prediction strategy is used for prediction, the input training set data is the t+11 points of the n points of the historical monthly power consumption data of the power users. When using the time series forecasting strategy for forecasting, the input training set data is the t+11 points of the n historical monthly carbon emissions data. The training set data sample input once is 12 consecutive points, and the predicted carbon emission data of 1 point is calculated. After rolling input calculation, the training set prediction data C of t points is obtained. pred-train =[c' 13 , c' 14 , ..., c' t+12 ], where C pred-train Represents the predicted value of carbon emissions output during the training process of the training set, c' 13 Represents the predicted carbon emissions value output during the training process of the first input training set data sample, c' t+12 Represents the predicted carbon emissions value output during the training process of the t-th input training set data sample; used during training Calculate C pred-train =[c' 13 , c' 14 , ..., c' t+12 ] and C train =[c13 , c 14 , ..., c t+12 ] is used as the training error to adjust the model weights;
[0027] Use the long short-term memory model to load the saved weight file, import the validation set data to verify the model effect, and use the carbon-electricity correlation prediction strategy to predict the historical monthly electricity consumption data. As the historical value part of the model input in the validation set to obtain the validation prediction value; when using the time series prediction strategy for prediction, As part of the historical value input of the model in the validation set, the validation prediction value is obtained. The validation set data sample inputted at a time is 12 consecutive points, and the predicted carbon emission data of 1 point is calculated. After rolling input calculation, the validation set prediction data of nt-12 points is obtained. in Represents the predicted value of carbon emissions output during the validation set validation process, Represents the predicted carbon emissions output during the verification process of the first input verification set data sample, Represents the carbon emission prediction value output during the verification process of the validation set data sample input for the nt-12th time; used during training calculate With C val =[c t+13 , c t+14 , ..., c n ] is used as the verification error to evaluate the model effect;
[0028] The above training and verification process is repeatedly performed until both the training error and the verification error drop to a stable level, that is, the fluctuation of the training error and the verification error is no more than A%, and then the final model weight file is saved to obtain the carbon emission prediction model.
[0029] Preferably, the specific method of using the trained prediction model to predict carbon emissions for power users is:
[0030] Use the long short-term memory model to load the final model weight file, import the associated prediction data to calculate the carbon emission prediction value, and when using the carbon-electricity associated prediction strategy for prediction, the imported associated prediction data is the electricity consumption data of the selected second historical time period, a total of 12 points X history =[x n-11 , x n-10 , ..., x n ], where X history represents the electricity consumption data of the power users in the selected second historical time period, x n-11 represents the electricity consumption data of power users in the first selected historical period from month n to month 11, xn Represents the electricity consumption data of the power users in the nth month in the selected first historical time period; when using the time series prediction strategy for prediction, the imported associated prediction data is the carbon emission data of the power users in the selected second historical time period, a total of 12 points C history =[c n-11 , c n-10 , ..., c n ], where C history represents the carbon emission data of electricity users in the selected second historical period, c n-11 represents the carbon emission data of electricity users in the first selected historical period from month n to month 11, c n represents the carbon emission data of electricity users in the nth month in the selected first historical time period; the imported associated prediction data is used as the input of the long short-term memory model to obtain the carbon emission prediction data c for the prediction month pred .
[0031] A method for predicting carbon emissions of power users based on carbon-electricity correlation comprises the following steps:
[0032] The collected electricity consumption data and carbon emission data of power users in the selected first historical time period are used as a power user carbon-electricity correlation prediction data set, the power user carbon-electricity correlation coefficient is calculated through the power user carbon-electricity correlation prediction data set, and the power user carbon emission prediction strategy is selected through the power user carbon-electricity correlation coefficient;
[0033] According to the selected carbon emission prediction strategy for power users, the power user carbon-electricity correlation prediction data set is divided into a training set, and the long short-term memory model is trained by the long short-term memory model prediction algorithm to obtain a carbon emission prediction model;
[0034] The electricity consumption data and carbon emission data of the power users in the selected second historical time period are input into the carbon emission prediction model to perform carbon emission prediction.
[0035] A computer program product comprises a computer program, wherein the computer program implements the steps of the above method when executed by a processor.
[0036] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.
[0037] Beneficial effects of the present invention:
[0038] By combining the carbon-electricity correlation characteristics, the present invention compares the carbon emissions output by the prediction model with the electricity usage behavior data, which helps power users better understand the impact of their energy usage behavior on carbon emissions, so as to take more environmentally friendly and energy-saving measures. At the same time, it also provides an effective method for monitoring and evaluating carbon emissions in the power industry, providing data support for the formulation of more scientific environmental protection policies, providing reference for optimizing power users' electricity usage behavior and reducing carbon emissions, and realizing precise and intelligent carbon emission management. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 It is a structural schematic diagram of the present invention;
[0040] Figure 2 A process diagram for carbon emission prediction;
[0041] Figure 3 It is a flow chart of the present invention. DETAILED DESCRIPTION
[0042] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:
[0043] Example 1
[0044] A carbon emission prediction system for power users based on carbon-electricity correlation, such as Figure 1 , Figure 2 As shown, it includes:
[0045] The strategy module is used to collect the electricity consumption data and carbon emission data of the selected first historical time period (the first historical time period refers to 2 to 10 years before the electricity user prediction month) as the electricity user carbon-electricity correlation prediction data set, calculate the electricity user carbon-electricity correlation coefficient through the electricity user carbon-electricity correlation prediction data set, and select the electricity user carbon emission prediction strategy through the electricity user carbon-electricity correlation coefficient. This design can more accurately understand the relationship between electricity use and carbon emissions and select the most appropriate electricity user carbon emission prediction strategy by calculating the electricity user carbon-electricity correlation coefficient.
[0046] The carbon emission prediction model module is used to divide the power user carbon-electricity correlation prediction data set into a training set (and a validation set, of which the training set accounts for 80% and the validation set accounts for 20%) according to the selected power user carbon emission prediction strategy, and train the long short-term memory model through the long short-term memory model prediction algorithm (LSTM prediction algorithm model) to obtain a carbon emission prediction model. This design can improve the accuracy of carbon emission prediction for power users;
[0047] The carbon emission prediction module is used to input the power consumption data and carbon emission data of power users in the selected second historical time period (the second historical time period refers to the year before the power user prediction month) into the carbon emission prediction model for carbon emission prediction. This design can reflect the carbon emissions of power users in real time and provide timely information support for carbon emission management. Through the carbon emission prediction results, power users and relevant management departments can formulate more effective carbon emission reduction strategies, optimize energy use, and reduce carbon emission costs.
[0048] In the above technical solution, the specific method of collecting the power consumption data and carbon emission data of the power users in the selected first historical time period includes:
[0049] The data sources of the power consumption data and carbon emission data of power users in the selected first historical time period include carbon emission verification reports of power users, carbon emission monitoring service platforms, data collected by smart meters, monthly electricity bills, and statistical data of power companies;
[0050] The electricity consumption data and carbon emission data of the selected first historical time period to be collected include: the electricity user forecast month M (M is specifically 2 to 10) years ago, the electricity user historical monthly electricity consumption data, a total of n points, X = [x 1 , x 2 , ..., x n ] (the unit of measurement is kWh); the electricity user is predicted M years ago, and the electricity user’s historical monthly carbon emissions data has a total of n points, C = [c 1 , c 2 , ..., c n ](Unit of measurement is tCO 2 ); X represents the power consumption data of the power user in the selected first historical time period, x n represents the electricity consumption data of the power user in the nth month in the selected first historical time period, C represents the carbon emission data of the power user in the selected first historical time period, c n It represents the carbon emission data of electricity users in the nth month in the selected first historical time period; the above design provides strong data support for subsequent carbon emission forecasting and decision-making.
[0051] In the above technical solution, the specific method for calculating the carbon-electricity correlation coefficient of power users is:
[0052] Calculate the average value of n points of historical monthly electricity consumption data of electricity users where x i Indicates the first historical period in the selected i Monthly electricity consumption data of electricity users, 1≤i≤n;
[0053] Calculate the average value of n points of historical monthly carbon emissions data of electricity users where c i Indicates the first historical period in the selected i Monthly carbon emission data of electricity users, 1≤i≤n;
[0054] The carbon-electricity correlation coefficient of power users is calculated between the historical monthly power consumption data of power users and the historical monthly carbon emissions data of power users. The calculation formula is as follows:
[0055]
[0056] Among them, P represents the carbon-electricity correlation coefficient of electricity users; the above design is used to obtain the carbon-electricity correlation coefficient of electricity users to select the carbon emission prediction strategy of electricity users.
[0057] In the above technical solution, the specific method for selecting the carbon emission prediction strategy of power users through the carbon-electricity correlation coefficient of power users is:
[0058] When P≥0.6, there is a strong correlation between the historical monthly electricity consumption data of power users and the historical monthly carbon emissions data of power users (at this time, carbon emissions are predicted through the historical monthly electricity consumption data of power users, and the historical monthly carbon emissions data of power users will be used when verifying the results, and the predicted carbon emissions data are verified by the historical monthly carbon emissions data of power users to verify the error), and the carbon-electricity correlation prediction strategy is used to divide the power user carbon-electricity correlation prediction data set; when P<0.6, there is no strong correlation between the historical monthly electricity consumption data of power users and the historical monthly carbon emissions data of power users (at this time, carbon emissions are predicted through the historical monthly carbon emissions data of power users), and the time series prediction strategy is used to divide the power user carbon-electricity correlation prediction data set; the above design can optimize the data set division and support personalized prediction, thereby improving the accuracy of carbon emission prediction.
[0059] In the above technical solution, the specific method of dividing the power user carbon-electricity correlation prediction data set according to the selected power user carbon emission prediction strategy is:
[0060] When using the carbon-electricity correlation prediction strategy for prediction, the historical monthly electricity consumption data of power users and the historical monthly carbon emissions data of power users are used to divide the data set into a training set and a validation set;
[0061] When using the time series forecasting strategy for forecasting, the historical monthly carbon emissions data of power users are used to divide the data set into training sets and validation sets;
[0062] When the data set is divided into training set and validation set, the data set is divided into training set and validation set according to the number of points of historical monthly carbon emissions data of power users. For the historical monthly carbon emissions data of power users with a total of n points, 0.8*(n-12) is rounded off as the number of points t of the training set, and C train =[c 13 , c 14 , ..., c t+12 ] is used as the true value part of the model output in the training set to obtain the training error, where C train represents the true value of carbon emissions output during the training process of the training set, c t+12 represents the carbon emission data of electricity users in the t+12th month; C val =[c t+13 , c t+14 , ..., c n ] is used as the true value part of the model output in the validation set to obtain the validation error, where C val represents the true value of carbon emissions output during the validation set validation process, c n Represents the carbon emission data of electricity users in the nth month;
[0063] When using the carbon-electricity correlation prediction strategy for prediction, the t+11 points of the n points of historical monthly power consumption data of power users are used. As part of the model input history value in the training set to obtain the training prediction value, where represents the historical value of electricity consumption data output during the training process of the training set, x t+11 Represents the electricity consumption data of power users in month t+11; As part of the historical value of the model input in the validation set to obtain the validation prediction value, represents the historical value of electricity consumption data output during the verification process of the verification set, x n-1 Represents the electricity consumption data of power users in month n-1;
[0064] When using the time series forecasting strategy for forecasting, the historical monthly carbon emissions data of power users is taken as the t+11 points out of the n points. As part of the model input history value in the training set to obtain the training prediction value, where represents the historical value of carbon emissions output during the training process of the training set, c t+11 represents the carbon emission data of electricity users in month t+11; As part of the historical value of the model input in the validation set to obtain the validation prediction value, Represents the historical value of carbon emissions output during the validation process of the validation set, c n-1 Represents the carbon emission data of electricity users in month n-1;
[0065] The above design helps to make the data set division scientific and reasonable.
[0066] In the above technical solution, the specific process of training the prediction model through the long short-term memory model prediction algorithm is as follows:
[0067] The training set data is input into the long short-term memory model. When the carbon-electricity correlation prediction strategy is used for prediction, the input training set data is the t+11 points of the n points of the historical monthly power consumption data of the power users. When using the time series forecasting strategy for forecasting, the input training set data is the t+11 points of the n historical monthly carbon emissions data. The training set data sample input once is 12 consecutive points, and the predicted carbon emission data of 1 point is calculated. After rolling input calculation, the training set prediction data C of t points is obtained. pred-train =[c' 13 , c' 14 , ..., c' t+12 ], where C pred-train Represents the predicted value of carbon emissions output during the training process of the training set, c' 13 Represents the predicted carbon emissions value output during the training process of the first input training set data sample, c' t+12 Represents the predicted carbon emissions value output during the training process of the t-th input training set data sample; used during training Calculate C pred-train =[c' 13 , c' 14 , ..., c' t+12 ] and C train =[c 13 , c 14 , ..., c t+12 ] is used as the training error to adjust the model weights;
[0068] Use the long short-term memory model to load the saved weight file, import the validation set data to verify the model effect, and use the carbon-electricity correlation prediction strategy to predict the historical monthly electricity consumption data. As the historical value part of the model input in the validation set to obtain the validation prediction value; when using the time series prediction strategy for prediction, As part of the historical value input of the model in the validation set, the validation prediction value is obtained. The validation set data sample inputted at a time is 12 consecutive points, and the predicted carbon emission data of 1 point is calculated. After rolling input calculation, the validation set prediction data of nt-12 points is obtained. in Represents the predicted value of carbon emissions output during the validation set validation process, Represents the predicted carbon emissions output during the verification process of the first input verification set data sample, Represents the carbon emission prediction value output during the verification process of the validation set data sample input for the nt-12th time; used during training calculate With C val =[c t+13 , c t+14 , ..., c n ] is used as the verification error to evaluate the model effect;
[0069] The above training and verification process is repeatedly performed until the training error and the verification error both drop to a stable level, that is, the fluctuation of the training error and the verification error is no more than A% (A% is 5%), then the final model weight file is saved to obtain the carbon emission prediction model; the above design optimizes and improves the long short-term memory model through training sets and verification sets to obtain the carbon emission prediction model, which can improve the training efficiency of the carbon emission prediction model and improve the accuracy of carbon emission prediction.
[0070] In the above technical solution, the specific method of using the trained prediction model to predict carbon emissions for power users is:
[0071] Use the long short-term memory model to load the final model weight file, import the associated prediction data to calculate the carbon emission prediction value, and when using the carbon-electricity associated prediction strategy for prediction, the imported associated prediction data is the electricity consumption data of the selected second historical time period, a total of 12 points X history =[x n-11 , x n-10 , ..., x n ], where X history represents the electricity consumption data of the power users in the selected second historical time period, x n-11 represents the electricity consumption data of power users in the first selected historical period from month n to month 11, x n Represents the electricity consumption data of the power users in the nth month in the selected first historical time period; when using the time series prediction strategy for prediction, the imported associated prediction data is the carbon emission data of the power users in the selected second historical time period, a total of 12 points C history =[c n-11 , c n-10 , ..., c n ], where C history represents the carbon emission data of electricity users in the selected second historical period, c n-11 represents the carbon emission data of electricity users in the first selected historical period from month n to month 11, c nrepresents the carbon emission data of electricity users in the nth month in the selected first historical time period; the imported associated prediction data is used as the input of the long short-term memory model to obtain the carbon emission prediction data c for the prediction month pred The above design can improve the accuracy and reliability of carbon emission prediction results and provide data support for energy conservation and emission reduction decisions.
[0072] Example 2
[0073] A method for predicting carbon emissions of power users based on carbon-electricity correlation, such as Figure 3 As shown, the electricity consumption data and carbon emission data of power users in the selected first historical time period are collected as a data set for carbon-electricity correlation prediction of power users to calculate the carbon-electricity correlation coefficient, and the carbon emission prediction strategy for power users is selected according to the carbon-electricity correlation coefficient; the data set for carbon-electricity correlation prediction of power users is divided into training sets according to the selected carbon emission prediction strategy for power users, and the training set is used to train the long short-term memory model through the long short-term memory model prediction algorithm to obtain a carbon emission prediction model; the electricity consumption data and carbon emission data of power users in the selected second historical time period are input into the carbon emission prediction model for carbon emission prediction.
[0074] The specific method for predicting carbon emissions of power users includes the following steps:
[0075] The collected electricity consumption data and carbon emission data of power users in the selected first historical time period are used as a power user carbon-electricity correlation prediction data set, the power user carbon-electricity correlation coefficient is calculated through the power user carbon-electricity correlation prediction data set, and the power user carbon emission prediction strategy is selected through the power user carbon-electricity correlation coefficient;
[0076] According to the selected carbon emission prediction strategy for power users, the power user carbon-electricity correlation prediction data set is divided into a training set, and the long short-term memory model is trained by the long short-term memory model prediction algorithm to obtain a carbon emission prediction model;
[0077] The electricity consumption data and carbon emission data of the power users in the selected second historical time period are input into the carbon emission prediction model to perform carbon emission prediction.
[0078] Example 3
[0079] A computer program product includes a computer program, characterized in that when the computer program is executed by a processor, the steps of the method described in Example 2 are implemented.
[0080] Example 4
[0081] A computer-readable storage medium stores a computer program, wherein the computer program implements the steps of the method described in Example 2 when executed by a processor.
[0082] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.
Claims
1. A carbon emission prediction system for power users based on carbon-electricity correlation, characterized in that: It includes: The strategy module is used to use the collected electricity consumption data and carbon emission data of the power users in the selected first historical time period as the power user carbon-electricity correlation prediction data set, calculate the power user carbon-electricity correlation coefficient through the power user carbon-electricity correlation prediction data set, and select the power user carbon emission prediction strategy through the power user carbon-electricity correlation coefficient; The carbon emission prediction model module is used to divide the power user carbon-electricity correlation prediction data set into a training set according to the selected power user carbon emission prediction strategy, and train the long-short-term memory model through the long-short-term memory model prediction algorithm to obtain a carbon emission prediction model. The carbon emission prediction module is used to input the power consumption data and carbon emission data of the power users in the selected second historical time period into the carbon emission prediction model to perform carbon emission prediction.
2. The carbon emission prediction system for power users based on carbon-electricity correlation according to claim 1 is characterized by: The specific method of collecting the electricity consumption data and carbon emission data of the power users in the selected first historical time period includes: The data sources of the power consumption data and carbon emission data of power users in the selected first historical time period include carbon emission verification reports of power users, carbon emission monitoring service platforms, data collected by smart meters, monthly electricity bills, and statistical data of power companies; The electricity consumption data and carbon emission data of the selected first historical time period to be collected include: the electricity consumption data of the electricity users M years before the forecast month, the historical monthly electricity consumption data of the electricity users, a total of n points, X = [x1, x2, ..., x n ]; The historical monthly carbon emission data of power users in the M years before the power user prediction month has a total of n points, C = [c1, c2, ..., c n ]; X represents the power consumption data of the power users in the selected first historical time period, x n represents the electricity consumption data of the power user in the nth month in the selected first historical time period, C represents the carbon emission data of the power user in the selected first historical time period, and c n Represents the carbon emission data of electricity users in the nth month in the selected first historical time period.
3. The carbon emission prediction system for power users based on carbon-electricity correlation according to claim 2 is characterized by: The specific method for calculating the carbon-electricity correlation coefficient of power users is: Calculate the average value of n points of historical monthly electricity consumption data of electricity users where x i Indicates the first historical period in the selected i Monthly electricity consumption data of electricity users, 1≤i≤n; Calculate the average value of n points of historical monthly carbon emissions data of electricity users where c i Indicates the first historical period in the selected i Monthly carbon emission data of electricity users, 1≤i≤n; The carbon-electricity correlation coefficient of power users is calculated between the historical monthly power consumption data of power users and the historical monthly carbon emissions data of power users. The calculation formula is as follows: Among them, P represents the carbon-electricity correlation coefficient of electricity users.
4. The carbon emission prediction system for power users based on carbon-electricity correlation according to claim 1 is characterized by: The specific method for selecting the carbon emission prediction strategy of power users through the carbon-electricity correlation coefficient of power users is as follows: When P≥0.6, there is a strong correlation between the historical monthly electricity consumption data of power users and the historical monthly carbon emissions data of power users, and the carbon-electricity correlation prediction strategy is used to divide the power user carbon-electricity correlation prediction data set; when P<0.6, there is no strong correlation between the historical monthly electricity consumption data of power users and the historical monthly carbon emissions data of power users, and the time series prediction strategy is used to divide the power user carbon-electricity correlation prediction data set.
5. The carbon emission prediction system for power users based on carbon-electricity correlation according to claim 4 is characterized by: The specific method for dividing the power user carbon-electricity correlation prediction data set according to the selected power user carbon emission prediction strategy is: When using the carbon-electricity correlation prediction strategy for prediction, the historical monthly electricity consumption data of power users and the historical monthly carbon emissions data of power users are used to divide the data set into a training set and a validation set; When using the time series forecasting strategy for forecasting, the historical monthly carbon emissions data of power users are used to divide the data set into training sets and validation sets; When the data set is divided into training set and validation set, the data set is divided into training set and validation set according to the number of points of historical monthly carbon emissions data of power users. For the historical monthly carbon emissions data of power users with a total of n points, 0.8*(n-12) is rounded off as the number of points t of the training set, and C train =[c 13 , c 14 , ..., c t+12 ] is used as the true value part of the model output in the training set to obtain the training error, where C train represents the true value of carbon emissions output during the training process of the training set, c t+12 represents the carbon emission data of electricity users in the t+12th month; C val =[c t+13 , c t+14 , ..., c n ] is used as the true value part of the model output in the validation set to obtain the validation error, where C val represents the true value of carbon emissions output during the validation set validation process, c n Represents the carbon emission data of electricity users in the nth month; When using the carbon-electricity correlation prediction strategy for prediction, the t+11 points of the n points of historical monthly power consumption data of power users are used. As part of the model input history value in the training set to obtain the training prediction value, where represents the historical value of electricity consumption data output during the training process of the training set, x t+11 Represents the electricity consumption data of power users in month t+11; As part of the historical value of the model input in the validation set to obtain the validation prediction value, represents the historical value of electricity consumption data output during the verification process of the verification set, x n-1 Represents the electricity consumption data of power users in month n-1; When using the time series forecasting strategy for forecasting, the historical monthly carbon emissions data of power users is taken as the t+11 points out of the n points. As part of the model input history value in the training set to obtain the training prediction value, where represents the historical value of carbon emissions output during the training process of the training set, c t+11 represents the carbon emission data of electricity users in month t+11; As part of the historical value of the model input in the validation set to obtain the validation prediction value, represents the historical value of carbon emissions output during the validation set validation process, c n-1 Represents the carbon emission data of electricity users in month n-1.
6. The carbon emission prediction system for power users based on carbon-electricity correlation according to claim 1 is characterized by: The specific process of training the prediction model through the long short-term memory model prediction algorithm is as follows: The training set data is input into the long short-term memory model. When the carbon-electricity correlation prediction strategy is used for prediction, the input training set data is the t+11 points of the n points of the historical monthly power consumption data of the power users. When using the time series forecasting strategy for forecasting, the input training set data is the t+11 points of the n historical monthly carbon emissions data. The training set data sample input once is 12 consecutive points, and the predicted carbon emission data of 1 point is calculated. After rolling input calculation, the training set prediction data C of t points is obtained. pred-train =[c' 13 , c' 14 , ..., c' t+12 ], where C pred-train Represents the predicted value of carbon emissions output during the training process of the training set, c' 13 Represents the predicted carbon emissions value output during the training process of the first input training set data sample, c' t+12 Represents the predicted carbon emissions value output during the training process of the t-th input training set data sample; used during training Calculate C pred-train =[c' 13 , c' 14 , ..., c' t+12 ]、C train =[c 13 , c 14 , ..., c t+12 ] is used as the training error to adjust the model weights; Use the long short-term memory model to load the saved weight file, import the validation set data to verify the model effect, and use the carbon-electricity correlation prediction strategy to predict the historical monthly electricity consumption data. As the historical value part of the model input in the validation set to obtain the validation prediction value; when using the time series prediction strategy for prediction, As part of the historical value input of the model in the validation set, the validation prediction value is obtained. The validation set data sample inputted at a time is 12 consecutive points, and the predicted carbon emission data of 1 point is calculated. After rolling input calculation, the validation set prediction data of nt-12 points is obtained. in Represents the predicted carbon emissions output during the validation set validation process. Represents the predicted carbon emissions output during the verification process of the first input verification set data sample, Represents the carbon emission prediction value output during the verification process of the validation set data sample input for the nt-12th time; used during training calculate With C val =[c t+13 , c t+14 , ..., c n ] is used as the verification error to evaluate the model effect; The above training and verification process is repeatedly performed until both the training error and the verification error drop to a stable level, that is, the fluctuation of the training error and the verification error is no more than A%, and then the final model weight file is saved to obtain the carbon emission prediction model.
7. The carbon emission prediction system for power users based on carbon-electricity correlation according to claim 1 is characterized by: The specific method of using the trained prediction model to predict carbon emissions for power users is: Use the long short-term memory model to load the final model weight file, import the associated prediction data to calculate the carbon emission prediction value, and when using the carbon-electricity associated prediction strategy for prediction, the imported associated prediction data is the electricity consumption data of the selected second historical time period, a total of 12 points X history =[x n-11 , x n-10 , ..., x n ], where X history represents the electricity consumption data of the power users in the selected second historical period, x n-11 represents the electricity consumption data of power users in the first selected historical period from month n to month 11, x n Represents the electricity consumption data of the power users in the nth month in the selected first historical time period; when using the time series prediction strategy for prediction, the imported associated prediction data is the carbon emission data of the power users in the selected second historical time period, a total of 12 points C history =[c n-11 , c n-10 , ..., c n ], where C history represents the carbon emission data of electricity users in the selected second historical period, c n-11 represents the carbon emission data of electricity users in the first selected historical period from month n to month 11, c n represents the carbon emission data of electricity users in the nth month in the selected first historical time period; the imported associated prediction data is used as the input of the long short-term memory model to obtain the carbon emission prediction data c for the prediction month pred .
8. A method for predicting carbon emissions of power users based on carbon-electricity correlation, characterized in that: It follows the steps: The collected electricity consumption data and carbon emission data of power users in the selected first historical time period are used as a power user carbon-electricity correlation prediction data set, the power user carbon-electricity correlation coefficient is calculated through the power user carbon-electricity correlation prediction data set, and the power user carbon emission prediction strategy is selected through the power user carbon-electricity correlation coefficient; According to the selected carbon emission prediction strategy for power users, the power user carbon-electricity correlation prediction data set is divided into a training set, and the long short-term memory model is trained by the long short-term memory model prediction algorithm to obtain a carbon emission prediction model; The electricity consumption data and carbon emission data of the power users in the selected second historical time period are input into the carbon emission prediction model to perform carbon emission prediction.
9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method described in claim 8 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to claim 8 are implemented.