CN117910657A - Prediction method, model training method, computing device, storage medium, and program product for carbon shift factor - Google Patents

Abstract

The embodiment of the invention provides a prediction method, a model training method, computing equipment, a storage medium and a program product of carbon rank factors. The prediction method of the carbon removal factor comprises the following steps: acquiring first energy consumption data generated by at least one energy source in a target area in a historical time period; predicting second energy consumption data of the at least one energy source within a predicted time period of the target area based on the first energy consumption data of the at least one energy source; acquiring a first carbon number factor released by the target area in the historical time period; and predicting a second carbon deposit factor corresponding to the target area in the prediction time period based on the second energy consumption data of the at least one energy source and at least one characteristic parameter in the first carbon deposit factor. The technical scheme provided by the embodiment of the invention realizes the prediction of the carbon rejection factor and ensures the accuracy of the carbon rejection factor.

Description

Carbon emission factor prediction method, model training method, computing equipment, storage medium and program product

Technical Field

Embodiments of the present invention relate to the field of artificial intelligence technology, and in particular to a carbon emission factor prediction method, a model training method, a computing device, a storage medium, and a program product.

Background technique

Carbon emission factor refers to the amount of carbon emissions released per unit of energy during production, transportation, use or processing. In the field of electricity consumption, it also refers to the amount of carbon emissions generated per unit of electrical energy.

The carbon emission factor can usually be used as an influencing factor to calculate the greenness of energy and to guide energy conservation and emission reduction measures. In the field of electricity consumption, since electricity may involve multiple sources of electricity, such as wind energy, thermal energy, solar energy and other energy sources, the carbon emission factor is usually calculated by an authoritative agency based on the consumption of multiple energy sources. Electricity users can only obtain the carbon emission factor currently released by the authoritative agency and perform data calculations based on the carbon emission factor to achieve energy conservation and emission reduction measures such as future electricity consumption planning.

However, the carbon emission factors published by authoritative organizations cannot accurately represent future carbon emission factors. Therefore, future energy-saving and emission reduction measures formulated based on them are not accurate.

Summary of the invention

The embodiments of the present invention provide a carbon emission factor prediction method, a model training method, a computing device, a storage medium and a program product, which are used to solve the technical problem of the accuracy of carbon emission factors in the prior art.

In a first aspect, an embodiment of the present invention provides a carbon emission factor prediction method, comprising:

Acquire first energy consumption data generated by at least one energy source in a target area during a historical time period;

Based on the first energy consumption data of the at least one energy source, predicting second energy consumption data of the at least one energy source within a predicted time period of the target area;

Obtaining the first carbon emission factor released by the target area during the historical time period;

Based on the second energy consumption data of the at least one energy source and at least one characteristic parameter in the first carbon emission factor, a second carbon emission factor corresponding to the target area within the prediction time period is predicted.

In a second aspect, an embodiment of the present invention provides a model training method, comprising:

Determine at least one sample characteristic parameter and a training label corresponding to the at least one sample characteristic parameter; the at least one sample characteristic parameter includes at least one of first energy consumption sample data, time sample data, and at least one weather sample data, and the training label includes energy consumption prediction sample data; or the at least one sample characteristic parameter includes at least one of predicted energy consumption sample data, time sample data, carbon emission factor historical sample data, and at least one weather sample data, and the training label includes carbon emission factor prediction sample data;

The at least one sample feature parameter forms a target feature set;

Using the target feature set and the training labels, training a prediction model;

Select at least one key feature parameter from the target feature set;

Perform operation on any two key feature parameters to generate candidate feature parameters;

Calculating the correlation between the candidate feature parameter and any feature sample parameter in the target feature set;

The candidate feature parameters whose correlations do not meet the correlation requirements are added to the target feature set, and the step of training the prediction model using the target feature set and the training labels is returned to continue until the prediction model meets the training conditions.

In a third aspect, an embodiment of the present invention provides a carbon emission factor prediction device, comprising:

A first acquisition module is used to acquire first energy consumption data generated by at least one energy source in a target area during a historical time period;

A first prediction module, configured to predict second energy consumption data of the at least one energy source within a prediction time period of the target area based on the first energy consumption data of the at least one energy source;

A second acquisition module is used to acquire a first carbon emission factor released by the target area within the historical time period;

The second prediction module is used to predict the second carbon emission factor corresponding to the target area within the prediction time period based on the second energy consumption data of the at least one energy source and at least one characteristic parameter in the first carbon emission factor.

In a fourth aspect, an embodiment of the present invention provides a model training device, comprising:

A first determination module is used to determine at least one sample characteristic parameter and a training label corresponding to the at least one sample characteristic parameter; the at least one sample characteristic parameter includes at least one of first energy consumption sample data, time sample data and at least one weather sample data, and the training label includes energy consumption prediction sample data; or the at least one sample characteristic parameter includes at least one of predicted energy consumption sample data, time sample data, carbon emission factor historical sample data and at least one weather sample data, and the training label includes carbon emission factor prediction sample data;

A first feature set construction module, configured to construct a target feature set from the at least one sample feature parameter;

A first training module, used to train a prediction model using the target feature set and the training labels;

A first screening module, used to screen at least one key feature parameter from the target feature set;

A first parameter generation module is used to perform an operation on any two key feature parameters to generate candidate feature parameters;

A first calculation module, used to calculate the correlation between the candidate feature parameter and any feature sample parameter in the target feature set;

The second training module is used to add candidate feature parameters whose correlations do not meet the correlation requirements into the target feature set, and trigger the first training module to continue executing until the prediction model meets the training conditions.

In a fifth aspect, an embodiment of the present invention provides a computing device, including a processing component and a storage component;

The storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the carbon emission factor prediction method provided in the embodiment of the present invention, or to implement the model training method provided in the embodiment of the present invention.

In a sixth aspect, an embodiment of the present invention provides a computer storage medium storing a computer program. When the computer program is executed by a computer, the method for predicting the carbon emission factor provided by the embodiment of the present invention is implemented, or the model training method provided by the embodiment of the present invention is implemented.

In the seventh aspect, an embodiment of the present invention provides a computer program product, which includes a computer program code. When the computer program code is executed by a computer device, the computer device executes the carbon emission factor prediction method provided by the embodiment of the present invention, or executes the model training method provided by the embodiment of the present invention.

An embodiment of the present invention obtains first energy consumption data generated by at least one energy source in a target area within a historical time period; based on the first energy consumption data of the at least one energy source, predicts second energy consumption data of the at least one energy source within a predicted time period of the target area; obtains a first carbon emission factor published by the target area within the historical time period; based on the second energy consumption data of the at least one energy source and at least one characteristic parameter in the first carbon emission factor, predicts a corresponding second carbon emission factor of the target area within the predicted time period. Since energy consumption has a certain regularity, an embodiment of the present application performs big data analysis on energy consumption and historically published carbon emission factors, adopts a two-layer prediction architecture, and predicts the carbon emission factor of the target area within the predicted time period based on the energy consumption of the target area within the historical time period, so that the predicted second carbon emission factor can accurately reflect the carbon emission level of the target area in the predicted time period, thereby more accurately guiding energy-saving and emission reduction measures in the target area.

These and other aspects of the present invention will become more apparent from the following description of the embodiments.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG1 schematically shows a flow chart of a method for predicting a carbon emission factor provided by an embodiment of the present invention;

FIG2 schematically shows a schematic diagram of a carbon emission factor prediction method provided by an embodiment of the present invention;

FIG3 schematically shows a schematic diagram of a training process of a first prediction model provided by an embodiment of the present invention;

FIG4 schematically shows a flow chart of a model training method provided by an embodiment of the present invention;

FIG5 schematically shows an application scenario of a method for predicting a carbon emission factor provided by an embodiment of the present invention;

FIG6 schematically shows a block diagram of a device for predicting a carbon emission factor provided by an embodiment of the present invention;

FIG7 schematically shows a block diagram of a carbon emission factor prediction device provided by an embodiment of the present invention;

FIG8 schematically shows a block diagram of a computing device provided by an embodiment of the present invention.

Detailed ways

In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

As can be seen from the description of the background technology, the carbon emission factor is usually calculated by an authoritative organization based on the consumption of multiple energy sources. Electricity users can only obtain the carbon emission factor currently released by the authoritative organization and perform data calculations based on the carbon emission factor to implement energy-saving and emission reduction measures such as future electricity consumption planning.

However, in the process of implementing the concept of the present invention, it was found that the carbon emission factors published by the authority cannot accurately represent the future carbon emission factors, resulting in unreliable data calculations based on the current carbon emission factors. As a result, the future energy conservation and emission reduction measures formulated based on the carbon emission factors published by the authority are inaccurate.

In order to improve the accuracy of the carbon emission factor, a technical solution of an embodiment of the present application was proposed after a series of studies. In the embodiment of the present application, first energy consumption data of at least one energy source in a target area during a historical time period is obtained; based on the first energy consumption data of the at least one energy source, second energy consumption data of the at least one energy source during a predicted time period of the target area is predicted; a first carbon emission factor published by the target area during the historical time period is obtained; based on the second energy consumption data of the at least one energy source and at least one characteristic parameter in the first carbon emission factor, a corresponding second carbon emission factor of the target area during the predicted time period is predicted. Since energy consumption has a certain regularity, the embodiment of the present application performs big data analysis on energy consumption and historically published carbon emission factors, adopts a two-layer prediction architecture, and predicts the carbon emission factor of the target area during the predicted time period based on the energy consumption of the target area during the historical time period, so that the predicted second carbon emission factor can accurately reflect the carbon emission level of the target area during the predicted time period, thereby more accurately guiding energy-saving and emission reduction measures in the target area.

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

FIG1 schematically shows a flow chart of a method for predicting a carbon emission factor provided by an embodiment of the present invention. The technical solution of this embodiment can be executed by a server. In practical applications, the server can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. The server can also be a server of a distributed system, or a server combined with a blockchain. The server can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology, etc. This application does not limit this.

The method may include the following steps:

101. Obtain first energy consumption data of at least one energy source in a target area during a historical time period.

The historical time period may refer to a time period before the current time. For example, the historical time period may be a historical time period consisting of a certain period of time from the current time and the current time.

In the embodiment of the present invention, the length of the historical time period can be flexibly selected according to actual application requirements, for example, one day, one week, one month, one year, etc., and the length of the historical time period is not limited here.

Among them, in the field of electricity consumption, energy can refer to energy that can generate electricity, including traditional energy and new energy. Traditional energy can refer to energy that will produce greenhouse gases such as carbon dioxide during use or production, such as coal, oil, natural gas, etc. New energy can refer to solar energy, wind energy, etc.

The target area can refer to a relatively independent geographical scope, such as a country, province, city, school, industrial park, etc.

The first energy consumption data of each energy source may refer to the energy consumption of the electric energy used by each energy source within the target area during the historical time period.

The first energy consumption data may be obtained based on data released by an authoritative organization. Of course, in some cases, the first energy consumption data of a certain energy source may not be available, and the corresponding historical energy data will be empty.

102 . Predict second energy consumption data of at least one energy source within a prediction time period in a target area based on first energy consumption data of at least one energy source.

The prediction time period may refer to a time period after the current time. For example, the prediction time period may be a prediction time period consisting of a prediction time of a certain market from the current time and the current time.

In the embodiment of the present invention, the length of the prediction time period can be flexibly selected according to actual application requirements, for example, one day, one week, one month, one year, etc., and the length of the prediction time period is not limited here. The length of the prediction time period can be the same as or different from the length of the historical time period, and the present invention does not limit this.

103. Obtain the first carbon emission factor released by the target area in the historical time period.

The first carbon emission factor may refer to the value of the carbon emission factor of the target area in a historical period of time published by an authoritative organization.

104. Predict a second carbon emission factor corresponding to the target area within a prediction time period based on second energy consumption data of at least one energy source and at least one characteristic parameter in the first carbon emission factor.

In practical applications, the predicted energy data or the first carbon emission factor may be empty. Therefore, when any characteristic parameter is empty, the remaining characteristic parameters may be used for prediction. Therefore, in a possible implementation of the present invention, the second energy consumption data or the first carbon emission factor may be used to predict the second carbon emission factor.

In another possible implementation of the present invention, the second energy consumption data and the first carbon emission factor may be used to predict the second carbon emission factor.

In an embodiment of the present invention, the second carbon emission factor may be used to indicate that energy conservation and emission reduction processing should be performed on the target area within the predicted time period.

This embodiment utilizes the regularity of energy consumption, conducts big data analysis on energy consumption and historically published carbon emission factors, and adopts a two-layer prediction architecture to achieve carbon emission factor prediction, so that the predicted second carbon emission factor can accurately reflect the carbon emission level of the target area during the prediction time period, thereby more accurately guiding the energy conservation and emission reduction processing of the target area.

In some embodiments, the method may further include:

Get at least one type of weather data corresponding to the target area within the forecast period:

Predicting second energy consumption data of at least one energy source within a prediction time period of a target area based on first energy consumption data of at least one energy source includes:

Based on the first energy consumption data of the at least one energy source and at least one characteristic parameter in the at least one weather data, second energy consumption data of the at least one energy source within a prediction time period of the target area is predicted.

In the process of implementing the concept of the present invention, it is found that the energy consumption of the target area is usually affected by the weather conditions in the target area. Therefore, when predicting the energy consumption of the target area, the weather conditions of the target area can be quantified to obtain weather data, and the weather data can be used as auxiliary data for energy consumption prediction.

Weather data may refer to information used to describe atmospheric conditions and meteorological elements, and may include, for example, one or more of temperature, humidity, air pressure, precipitation, wind speed, wind direction, visibility, cloud cover, and UV index.

Weather data is usually collected and recorded by meteorological equipment such as meteorological observation stations, satellites, and radars, and collated and released by meteorological agencies. Therefore, weather data can be obtained from meteorological agencies.

The following takes the target area as a computing center and the energy source as electricity as an example to schematically illustrate the impact of weather conditions on the power consumption of the computing center.

First, temperature may have an impact on the energy consumption of the computing center. The operation of the computing center equipment will generate a lot of heat, so heat dissipation is needed to maintain the normal operating temperature of the equipment. When the outdoor temperature is high, the computing center needs to increase the use of air conditioning to lower the indoor temperature to ensure the stability and performance of the equipment.

Secondly, humidity may affect the energy consumption of computing centers. Humidity affects the reliability and durability of computing equipment. Too high humidity may cause damage and corrosion to equipment, while too low humidity may cause electrostatic discharge and equipment failure. Therefore, computing centers usually need humidity control to ensure that equipment operates within an appropriate humidity range.

In addition, precipitation may have an impact on the energy consumption of the computing center. Computing centers usually need to take waterproof measures to prevent rainwater from invading equipment and computer rooms. These measures may include waterproof layers, drainage systems, etc.

Also, wind speed may have an impact on the energy consumption of a computing center. Computing centers may use air cooling systems instead of traditional air conditioning systems to reduce energy consumption. These systems use the natural force of wind to dissipate heat and maintain the temperature of equipment. Therefore, when the wind speed is higher, the air cooling system is more effective and energy consumption is reduced accordingly.

In some embodiments, since electricity consumption at different times may have certain regularity, such as more electricity consumption during the day and less electricity consumption at night, the above-mentioned prediction of the second energy consumption data of at least one energy in the predicted time period of the target area based on the first energy consumption data of at least one energy and at least one characteristic parameter in at least one weather data may include:

Based on first energy consumption data of at least one energy source, at least one weather data, and at least one characteristic parameter in time data corresponding to a prediction time period, second energy consumption data of at least one energy source within a prediction time period of a target area is predicted.

Of course, as another embodiment, second energy consumption data of at least one energy source within the predicted time period of the target area may be predicted by using first energy consumption data of at least one energy source and at least one characteristic parameter in time data corresponding to the predicted time period.

In practical applications, since the first energy consumption data of any energy source, any weather data, etc. may be empty and cannot be obtained, the technical solution of the embodiment of the present invention can also be used to predict the second energy consumption data of at least one energy source within the predicted time period of the target area when any characteristic parameter is missing. Therefore, in a possible implementation of the present invention, the first energy consumption data of any energy source, any weather data, or time number can be used as characteristic parameters to predict the second energy consumption data.

In another possible implementation of the present invention, the second energy consumption data may be obtained by using at least one first energy consumption data, at least one weather data, and multiple characteristic parameters in the time data.

As can be seen from the foregoing description, weather data will affect energy consumption, and therefore will also affect the carbon emission factor. In some embodiments, based on the second energy consumption data of at least one energy source and at least one characteristic parameter in the first carbon emission factor, the second carbon emission factor corresponding to the predicted target area within the predicted time period can be specifically implemented as follows:

Based on the second energy consumption data of at least one energy source, at least one weather data and at least one characteristic parameter of the first carbon emission factor, a second carbon emission factor corresponding to the target area within the prediction time period is predicted.

In addition, time data may also be added for prediction. Therefore, in some embodiments, the second energy consumption data based on at least one energy source, at least one weather data and at least one characteristic parameter in the first carbon emission factor may be used to predict the second carbon emission factor corresponding to the target area within the prediction time period. The prediction may include:

Based on the second energy consumption data of at least one energy source, at least one weather data, time data and at least one characteristic parameter of the first carbon emission factor, a second carbon emission factor corresponding to the target area within the prediction time period is predicted.

Of course, the second carbon emission factor corresponding to the target area within the predicted time period may also be predicted based on the second energy consumption data of at least one energy source, the time data and at least one characteristic parameter of the first carbon emission factor.

Among them, in order to further improve the prediction accuracy, in addition to using the data analysis method, a machine learning model can also be used. Therefore, in some embodiments, based on the first energy consumption data, at least one weather data, and at least one characteristic parameter in the time data corresponding to the prediction time period, predicting the second energy consumption data of at least one energy source in the prediction time period of the target area can be specifically implemented as follows:

For any type of energy, based on the first energy consumption data of the energy, at least one weather data, and at least one characteristic parameter in the time data corresponding to the prediction time period, the first prediction model is used to predict the second energy consumption data of the energy within the prediction time period.

The first prediction model may be implemented as one of a linear regression model, a decision tree model, a random forest model, a support vector machine model, and a neural network model. In a possible implementation, the first prediction model may be one of an ANN (Artificial Neutral Network) and an LSTM (Long short-term memory).

Specifically, using the first prediction model to predict the second energy consumption data can be achieved as follows:

At least one characteristic parameter is input into the first prediction model so that the first prediction model outputs second energy consumption data.

In practical applications, in order to achieve refined prediction, the above-mentioned first energy consumption data, weather data, and time data can be time series data, which is composed of multiple time step data elements, wherein each time step can be set according to actual needs, for example, it can be hourly level, and each time step represents 1 hour.

The first energy consumption data is composed of energy consumption data corresponding to the current time step and multiple time steps before the current time step; the weather data is composed of weather data corresponding to multiple time steps;

The above-mentioned time data is composed of time information corresponding to multiple time steps, which may refer to calendar variables corresponding to the next L hours. Optionally, the time data may include at least one time series, such as time series corresponding to months, dates and hours. Assuming that the current time is 14:00 and the predicted time period is the next 4 hours, the time series corresponding to the hours is (10, 11, 12, 13). Since the month and date corresponding to each time step are the same, the time series corresponding to the month can be a value such as 5, and the time series corresponding to the date can be a value such as 26. Then the time data is: 5, 26, (10, 11, 12, 13).

The second energy consumption data is composed of prediction data corresponding to a plurality of time steps after the current time step.

In a practical application, the first prediction model is used to predict the second energy consumption data, for example, as shown in the following formula (1).

;(1)

in, , e can represent the first energy consumption data, /> Can represent hourly data, /> Can represent month data, /> Can represent date data, /> Can represent temperature data, /> Can represent humidity data,/> Can represent wind speed data, /> Can represent dew point data, /> The network parameters of the first prediction model may be represented; /> Indicates the second energy consumption data.

According to an embodiment of the present invention, based on the second energy consumption data of at least one energy source, at least one weather data, time data and at least one characteristic parameter of the first carbon emission factor, the second carbon emission factor corresponding to the target area in the prediction time period can be predicted as follows:

Based on the second energy consumption data of at least one energy source, at least one weather data, time data and at least one characteristic parameter of the first carbon emission factor, a second prediction model is used to predict the second carbon emission factor of the target area within the prediction time period.

The second prediction model can be implemented as one of a linear regression model, a decision tree model, a random forest model, a support vector machine model, and a neural network model. In a possible implementation of the present invention, the second prediction model can be implemented as XGBoost (Extreme Gradient Boosting).

Specifically, predicting the second energy consumption data using the second prediction model can be implemented as follows:

At least one characteristic parameter is input into the second prediction model so that the second prediction model outputs a second carbon emission factor.

The second carbon emission factor predicted by the second prediction model can be expressed by the following formula (2).

in, The second energy consumption data may be represented by Can represent hourly data, /> Can represent month data, /> Can represent date data, /> Can represent temperature data, /> Can represent humidity data, Can represent wind speed data, /> Dew point data can be displayed.

The second prediction model can be generated by training using a loss function. In an embodiment of the present invention, taking XGBoost as the second prediction model, the loss function can be a joint loss of multiple weak learners, specifically, the following formula (3).

in, Can refer to the loss value, /> It can refer to the output value of the second prediction model, /> Can refer to the label value of the second prediction model, /> Can refer to the model complexity of the weak learner decision tree.

For ease of understanding, FIG2 schematically shows a schematic diagram of a carbon emission factor prediction method provided in an embodiment of the present invention.

As shown in FIG2 , the carbon emission factor prediction method proposed in the embodiment of the present invention adopts a two-layer architecture. When the first energy consumption data is used to predict the carbon emission factor value within the prediction time period, the first energy consumption data, weather data and time data are first input into the first prediction model to predict the second energy consumption data within the prediction time period. Then, the second energy consumption data, weather data and time data are input into the second prediction model to predict the second carbon emission factor within the prediction time period.

It should be noted that when the first prediction model is used to generate the second energy consumption data, the same or different first prediction models may be used to predict the first energy consumption data of different types of energy. The multiple first prediction models may all be first prediction models of the same type or first prediction models of different types.

According to an embodiment of the present invention, the first prediction model can be pre-trained based on the first energy consumption sample data, time sample data, at least one sample characteristic parameter in at least one weather sample data, and energy consumption prediction sample data corresponding to at least one sample characteristic parameter.

Among them, the first energy consumption sample data, time sample data, weather sample data and the data content referred to by the above-mentioned first energy consumption data, time data and weather data are similar and will not be repeated here.

In practical applications, since the first energy consumption sample data, time sample data, weather sample data, etc. of any energy source may be empty and cannot be obtained, when any characteristic parameter is empty, the remaining characteristic parameters can also be used for prediction. The sample characteristic parameters can be obtained by using any one or more of the first energy consumption sample data, time sample data, and weather sample data, and the sample characteristic parameters are used as the training data set of the first prediction model to train the first prediction model.

Specifically, the training data set may be input into the first prediction model to be trained, and the first prediction model may predict the energy consumption according to the input training data set, and output the predicted energy consumption value.

The energy consumption prediction sample may be the expected output of the first prediction model during the current round of training, that is, the value expected to be predicted by the first prediction model based on the training data set.

After obtaining the energy consumption prediction value output by the first prediction model, the network parameters of the first prediction model can be adjusted based on the deviation between the energy consumption prediction value and the energy consumption prediction sample data to implement the training of the first prediction model.

In a possible implementation of the present invention, the deviation between the energy consumption prediction value and the energy consumption prediction sample data may be calculated using the following formula (4).

As can be seen from the foregoing description, some sample feature parameters may be empty and cannot be obtained. In order to ensure that the first prediction model still achieves accurate prediction when the feature parameters are missing, in some embodiments, the first prediction model is specifically trained in the following manner:

Determine at least one sample characteristic parameter of the first energy consumption sample data, the time sample data, and at least one weather sample data;

A target feature set is formed by at least one sample feature parameter;

The target feature set is used as the model input, and the energy consumption prediction sample data is used as the training label to train the first prediction model;

Select at least one key feature parameter from the target feature set;

Perform operation on any two key feature parameters to generate candidate feature parameters;

Calculate the correlation between the candidate feature parameters and any feature sample parameters in the target feature set;

The candidate feature parameters whose correlations do not meet the correlation requirements are added to the target feature set, and the target feature set is returned as the model input, and the energy consumption prediction sample data is used as the training label. The step of training the first prediction model continues until the first prediction model meets the training conditions.

In practical applications, in order to achieve refined prediction, the above-mentioned first energy consumption sample data, time sample data, and at least one weather sample data can be time series data, composed of multiple time step data elements, wherein each time step can be set according to actual needs, for example, it can be at the hourly level, and each time step represents 1 hour.

The first energy consumption sample data may be composed of a plurality of time steps and an energy consumption value corresponding to each time step; the weather sample data may be composed of a plurality of time steps and a weather data value corresponding to each time step.

For the target feature set, the sample feature parameters contained in the target feature set can be analyzed to determine the importance of each sample feature parameter to the model prediction, and the sample feature parameters with higher importance can be determined as key feature parameters.

In a possible implementation of the present invention, analysis methods such as residual decision tree and Shapley Additive exPlanations (SHAP) can be used to analyze the sample feature parameters included in the target feature set to measure the contribution of each sample feature parameter to the model prediction.

After the key characteristic parameters are obtained through screening, more characteristic parameters may be generated based on at least one key characteristic parameter obtained through screening.

In another implementation of the present invention, when multiple key feature parameters are screened from the target feature set, two key feature parameters can be randomly selected from the multiple key feature parameters to generate a data pair, and calculation operations are performed on the data pair to generate candidate feature parameters.

The operation may include, for example, addition, subtraction, multiplication, division, etc. The specific operation method performed on each pair of data may be randomly determined.

In order to improve the accuracy of model prediction, it is often necessary to ensure that the correlation between various parameters in the training data set is low, so as to reduce the impact of multicollinearity on the model and avoid excessive linear dependence between parameters.

Therefore, after obtaining the candidate feature parameters, the correlation between the candidate feature parameters and any feature sample parameters in the target feature set can be determined first, and the correlation result corresponding to each candidate feature parameter can be obtained. In an embodiment of the present invention, the correlation between the candidate feature parameters and any feature sample parameters in the target feature set can be obtained using the Pearson correlation coefficient and the Spearman's rank correlation coefficient.

The correlation requirement may include, for example, that the correlation is greater than a preset threshold. That is, only candidate feature parameters whose correlation results are less than the preset threshold, that is, those with a low correlation with any feature sample parameter in the target feature set, can be added to the target feature set.

The target feature set with the newly added candidate feature parameters can be used for the next round of model training.

Furthermore, after the next round of model training is completed, the key feature parameters can be screened from the target feature set and the candidate feature parameters can be generated based on the key feature parameters until the first prediction model training is completed.

The training conditions may include, for example, that the number of training times reaches the number of prediction times, or that the deviation between the energy consumption prediction value output by the first prediction model and the energy consumption prediction sample data is less than a preset deviation threshold.

According to an embodiment of the present invention, screening at least one key feature parameter from the target feature set may be implemented as follows:

Determine a first prediction result generated by a first prediction model based on a target feature set;

Based on the first prediction result, at least one key feature parameter is selected from the target feature set.

Among them, feature parameters with a high correlation with the first prediction result in the target feature set can be screened as key feature parameters.

The correlation between the feature parameter and the first prediction result can be achieved by calculating the feature importance index based on, for example, the feature importance of a decision tree, the coefficient size of LASSO (The Least Absolute Shrinkage and Selection Operator) regression, or by using a feature selection algorithm.

According to an embodiment of the present invention, the carbon emission factor prediction method further includes:

From the historical production data of the target area, first energy consumption data generated in a first time period before the target historical time is obtained as first energy consumption sample data, and first energy consumption data generated in a second time period after the target historical time is obtained as energy consumption prediction sample data;

The weather data occurring in the second time period is used as weather sample data, and the time data corresponding to the second time period is used as time sample data.

In some embodiments of the present invention, the carbon emission factor prediction method further includes:

Training a plurality of first candidate models based on the first energy consumption sample data, the time sample data, at least one sample characteristic parameter in at least one weather sample data, and energy consumption prediction sample data corresponding to the at least one sample characteristic parameter;

performing model evaluation on a plurality of first candidate models;

The first candidate model whose model evaluation result meets the performance requirements is selected as the first prediction model.

In the process of implementing the concept of the present invention, it is found that different models have different dependencies and sensitivities on the characteristic parameters in the data set, so different models may have different accuracies when predicting energy consumption. Based on this, the embodiment of the present application can select multiple first candidate models and uniformly train the multiple first candidate models. After the training is completed, the first prediction model that meets the new performance requirements is determined from the multiple candidate models by evaluating the multiple first candidate models.

The performance requirement may include, for example, that the prediction accuracy is higher than a preset threshold.

Among them, the multiple first candidate models can be models constructed using different ideas and methods, such as a model constructed based on the idea of linear regression, a model constructed based on a neural network, a model constructed based on a support vector machine, etc.

According to an embodiment of the present invention, the second prediction model is obtained by training the carbon emission factor prediction sample data based on the predicted energy consumption sample data, the second time sample data, the carbon emission factor sample data, and at least one sample characteristic parameter in at least one first weather sample data, and at least one sample characteristic parameter corresponding to the carbon emission factor prediction sample data.

According to an embodiment of the present invention, the second prediction model is trained and obtained in the following manner:

Determine at least one sample characteristic parameter among the predicted energy consumption sample data, the time sample data, the carbon emission factor sample data, and at least one first weather sample data;

A target feature set is formed by at least one sample feature parameter;

The target feature set is used as the model input, and the carbon emission factor prediction sample data is used as the training label to train the second prediction model;

Select at least one key feature parameter from the target feature set;

Perform operation on any two key feature parameters to generate candidate feature parameters;

Calculate the correlation between the candidate feature parameters and any feature sample parameters in the target feature set;

The candidate feature parameters whose correlations do not meet the correlation requirements are added to the target feature set, and the target feature set is returned as the model input, and the carbon emission factor prediction sample data is used as the training label. The step of training the second prediction model continues until the second prediction model meets the training conditions.

In practical applications, in order to achieve refined prediction, the above-mentioned predicted energy consumption sample data, time sample data, and carbon emission factor sample data can be time series data, consisting of multiple time step data elements, where each time step can be set according to actual needs, for example, it can be at the hourly level, and each time step represents 1 hour.

The above-mentioned predicted energy consumption sample data may be composed of multiple time steps and the energy consumption value corresponding to each time step; the above-mentioned weather sample data may be composed of multiple time steps and the weather data value corresponding to each time step.

For the target feature set, the sample feature parameters contained in the target feature set can be analyzed to determine the importance of each sample feature parameter to the model prediction, and the sample feature parameters with higher importance can be determined as key feature parameters.

In a possible implementation of the present invention, the sample feature parameters contained in the target feature set can be analyzed using analysis methods such as residual decision tree, Shapley and interpretation to measure the contribution of each sample feature parameter to the model prediction. A larger index value can indicate that the feature has a greater impact on the prediction result.

After the key characteristic parameters are obtained through screening, more characteristic parameters may be generated based on at least one key characteristic parameter obtained through screening.

In one implementation of the present invention, when a key feature parameter is screened from a target feature set, the key feature parameter can be preprocessed to generate a first key feature parameter, and then the key feature parameter and the first key feature parameter are combined into a data pair, and an operation is performed on the data pair to generate a candidate feature parameter.

The preprocessing of the key characteristic parameter may include, for example, obtaining a weight factor, and performing an operation on the weight factor and the key characteristic parameter to obtain a first key parameter.

In another implementation of the present invention, when multiple key feature parameters are screened from the target feature set, two key feature parameters can be randomly selected from the multiple key feature parameters to generate a data pair, and calculation operations are performed on the data pair to generate candidate feature parameters.

The calculation operation may include, for example, addition, subtraction, multiplication, division, etc. The specific calculation method performed on each pair of data may be randomly determined.

Since the key feature parameters are parameters that contribute more to the model prediction, the candidate feature parameters generated using the key feature parameters may have a higher contribution to the model prediction.

In order to improve the accuracy of model prediction, it is often necessary to ensure that the correlation between various parameters in the training data set is low, so as to reduce the impact of multicollinearity on the model and avoid excessive linear dependence between parameters.

Therefore, after obtaining the candidate feature parameters, the correlation between the candidate feature parameters and any feature sample parameter in the target feature set can be determined first, and the correlation result corresponding to each candidate feature parameter can be obtained.

The correlation requirement may include, for example, that the correlation is greater than a preset threshold. That is, only candidate feature parameters whose correlation results are less than the preset threshold, that is, those with a low correlation with any feature sample parameter in the target feature set, can be added to the target feature set.

The target feature set with the newly added candidate feature parameters can be used for the next round of model training.

Furthermore, after the next round of model training is completed, the key feature parameters can be screened from the target feature set and the candidate feature parameters can be generated based on the key feature parameters until the first prediction model training is completed.

The training conditions may include, for example, that the number of training times reaches the number of prediction times, or that the deviation between the energy consumption prediction value output by the first prediction model and the energy consumption prediction sample data is less than a preset deviation threshold.

For ease of understanding, the training process of the first prediction model is introduced below by taking the first prediction model as an example in conjunction with the model training schematic diagram shown in FIG3 .

As shown in FIG3 , input data x such as first energy consumption data of at least one energy source in a historical time period, weather data in a predicted time period, time data, etc. may be input into a first prediction model to obtain a first prediction result.

After a round of training of the first prediction model is completed, a first prediction result is obtained, and then key feature extraction 301 can be performed based on the first prediction result to obtain at least one key feature parameter.

After the key feature parameters are obtained, a feature generation operation 302 may be performed. For example, a data pair consisting of any two key feature parameters may be processed to generate candidate feature parameters.

Furthermore, feature selection processing 303 may be performed on the candidate feature parameters. For example, the correlation between the candidate feature parameters and the feature parameters in the target feature set may be determined, and the candidate feature parameters with lower correlation may be added to the input data x for the next round of training of the first prediction model until the first prediction model meets the training conditions.

According to an embodiment of the present invention, the first energy consumption data, the weather data, the time data and the second energy consumption data are respectively time series data;

According to an embodiment of the present invention, the method may further include:

According to the second carbon emission factor corresponding to the target area, the corresponding carbon emission amount within any time range within the forecast time period is calculated.

The carbon emission amount can be calculated by multiplying the predicted value of the carbon emission factor by the energy consumption of the target area within the selected time range.

According to an embodiment of the present invention, a data center for providing computing services is deployed in the target area, and the carbon emission factor prediction method further includes:

Generate recommended information for data centers based on carbon emissions;

Send recommendation prompt information to target users.

The recommended prompt information may include a time range with the least carbon emissions among multiple time ranges in the target area. The prompt information is used to prompt the user to schedule the computing task of the data center to be performed within the time range.

According to an embodiment of the present invention, a data center is deployed in the target area, and the carbon emission factor prediction method further includes:

According to the second carbon emission factor corresponding to the target area, the corresponding carbon emission amount within any time range within the forecast time period is calculated;

Combined with the second carbon emission factor corresponding to the target area, calculate the corresponding computing cost of the data center within any time range within the forecast time period;

Computing tasks are allocated based on the computing costs corresponding to data centers in different target areas in different time ranges.

By calculating the computing costs of different time ranges through the predicted value of the carbon emission factor, the computing tasks can be allocated to the time range with lower computing costs, thereby reducing the operating costs of the data center.

According to an embodiment of the present invention, based on the first energy consumption data, at least one weather data, and at least one characteristic parameter in the time data representing the prediction time period, predicting the second energy consumption data corresponding to at least one energy source in the target area within the prediction time period can be specifically implemented as follows:

determining at least one initial characteristic parameter consisting of the first energy consumption data, the at least one weather data, and time data representing the predicted time period;

forming a first feature set by the at least one initial feature parameter;

Performing an operation on any two feature parameters in the first feature set to generate a target feature parameter;

Calculating the correlation between the target feature parameter and any feature parameter in the initial feature set;

Adding target feature parameters whose correlations do not meet the correlation requirements to the first feature set, and returning to perform an operation on any two feature parameters in the first feature set to generate target feature parameters, the step continues until the first feature set meets the feature requirements, thereby obtaining a second feature set;

The second feature set is used to predict second energy consumption data corresponding to the at least one energy source in the target area within a prediction time period.

In the actual application process of carbon emission factor prediction, the first energy consumption data, at least one weather data, or time data representing the prediction time period may be missing. Missing data may affect the accuracy of carbon emission factor prediction.

Therefore, before performing the prediction of the carbon emission factor, a feature generation operation can be performed first.

For the first feature set, the initial feature parameters contained in the first feature set can be analyzed to determine the importance of each initial feature parameter to the model prediction, and the sample feature parameters with higher importance can be determined as key feature parameters.

In a possible implementation of the present invention, the initial feature parameters contained in the first feature set can be analyzed using analysis methods such as residual decision tree, Shapley and interpretation to measure the contribution of each initial feature parameter to the model prediction. A larger index value can indicate that the feature has a greater impact on the prediction result.

After the initial key characteristic parameters are obtained through screening, more characteristic parameters may be generated based on at least one initial key characteristic parameter obtained through screening.

In one implementation of the present invention, when multiple initial key feature parameters are screened from the first feature set, two initial key feature parameters can be randomly selected from the multiple initial key feature parameters to generate a data pair, and an operation is performed on the data pair to generate a target feature parameter.

The operation may include, for example, addition, subtraction, multiplication, division, etc. The specific operation method performed on each pair of data may be randomly determined.

Since the initial key feature parameters are parameters that contribute more to the model prediction, the target feature parameters generated using the initial key feature parameters may have a higher contribution to the model prediction.

In order to improve the accuracy of model prediction, it is often necessary to ensure that the correlation between various parameters in the training data set is low, so as to reduce the impact of multicollinearity on the model and avoid excessive linear dependence between parameters.

Therefore, after obtaining the target feature parameter, the correlation between the target feature parameter and any initial feature sample parameter in the initial feature set can be determined first, and the correlation result corresponding to each target feature parameter can be obtained. In an embodiment of the present invention, the correlation between the candidate feature parameter and any feature sample parameter in the target feature set can be obtained using the Pearson correlation coefficient and the Spearman's rank correlation coefficient.

The correlation requirement may include, for example, that the correlation is greater than a preset threshold. That is, only target feature parameters with a correlation result less than the preset threshold, that is, with a low correlation with any initial feature sample parameter in the initial feature set, can be added to the initial feature set.

FIG4 schematically shows a flow chart of a model training method provided by an embodiment of the present invention. As shown in FIG4 , the model training method may specifically include the following steps:

401, determining at least one sample feature parameter and a training label corresponding to the at least one sample feature parameter;

402, forming a target feature set from at least one sample feature parameter;

403, training a prediction model using the target feature set and the training labels;

As an optional method, at least one sample characteristic parameter includes at least one of first energy consumption sample data, time sample data and at least one weather sample data, and the training label includes energy consumption prediction sample data; the prediction model is also the above-mentioned first prediction model.

As another optional method, at least one sample characteristic parameter includes at least one of predicted energy consumption sample data, time sample data, carbon emission factor historical sample data and at least one weather sample data, and the training label includes carbon emission factor predicted sample data; the prediction model is also the second prediction model mentioned above.

404, selecting at least one key feature parameter from the target feature set;

405, performing an operation on any two key feature parameters to generate candidate feature parameters;

406, calculating the correlation between the candidate feature parameter and any feature sample parameter in the target feature set;

407, adding the candidate feature parameters whose correlations do not meet the correlation requirements to the target feature set, and returning to the step of training the prediction model using the target feature set and the training labels until the prediction model meets the training conditions.

In practical applications, in order to achieve refined prediction, the above-mentioned first energy consumption sample data, weather sample data, and time sample data can be time series data, composed of multiple time step data elements, wherein each time step can be set according to actual needs, for example, it can be at the hourly level, and each time step represents 1 hour.

The above historical sample energy consumption data may be composed of multiple time steps and the energy consumption value corresponding to each time step; the above weather sample data may be composed of multiple time steps and the weather data value corresponding to each time step.

The first energy consumption sample data may include first energy consumption data generated in a first time period before a target historical time obtained from historical production data of the target area. The first energy consumption data generated in a second time period after the target historical time may be used as the energy consumption sample data.

The weather sample data may be weather data occurring in the second time period; and the time sample data may be time data corresponding to the second time period.

Among them, the target historical time can refer to the time before the current time point. For example, the target historical time can be the historical time period composed of the historical time point before the current time point and the current time point, or it can be the historical time period composed of the first historical time point before the current time point and the second historical time point before the current time point.

In the embodiment of the present invention, the length of the historical time period can be flexibly selected according to actual application requirements, for example, one day, one week, one month, one year, etc., and the length of the historical time period is not limited here.

Among them, in the field of electricity consumption, energy can refer to energy that can generate electricity, including traditional energy and new energy. Traditional energy can refer to energy that will produce greenhouse gases such as carbon dioxide during use or production, such as coal, oil, natural gas, etc. New energy can refer to solar energy, wind energy, etc.

The target area can refer to a relatively independent geographical scope, such as a country, province, city, school, industrial park, etc.

For the target feature set, the sample feature parameters contained in the target feature set can be analyzed to determine the importance of each sample feature parameter to the model prediction, and the sample feature parameters with higher importance can be determined as key feature parameters.

In a possible implementation of the present invention, the sample feature parameters contained in the target feature set can be analyzed using analysis methods such as residual decision tree, Shapley and interpretation to measure the contribution of each sample feature parameter to the model prediction. A larger index value can indicate that the feature has a greater impact on the prediction result.

After the key characteristic parameters are obtained through screening, more characteristic parameters may be generated based on at least one key characteristic parameter obtained through screening.

In one implementation of the present invention, when a key feature parameter is screened from a target feature set, the key feature parameter can be preprocessed to generate a first key feature parameter, and then the key feature parameter and the first key feature parameter are combined into a data pair, and an operation is performed on the data pair to generate a candidate feature parameter.

The preprocessing of the key characteristic parameter may include, for example, obtaining a weight factor, and performing an operation on the weight factor and the key characteristic parameter to obtain a first key parameter.

In another implementation of the present invention, when multiple key feature parameters are screened from the target feature set, two key feature parameters can be randomly selected from the multiple key feature parameters to generate a data pair, and calculation operations are performed on the data pair to generate candidate feature parameters.

The calculation operation may include, for example, addition, subtraction, multiplication, division, etc. The specific calculation method performed on each pair of data may be randomly determined.

Since the key feature parameters are parameters that contribute more to the model prediction, the candidate feature parameters generated using the key feature parameters may have a higher contribution to the model prediction.

In order to improve the accuracy of model prediction, it is often necessary to ensure that the correlation between various parameters in the training data set is low, so as to reduce the impact of multicollinearity on the model and avoid excessive linear dependence between parameters.

Therefore, after obtaining the candidate feature parameters, the correlation between the candidate feature parameters and any feature sample parameter in the target feature set can be determined first, and the correlation result corresponding to each candidate feature parameter can be obtained.

The correlation requirement may include, for example, that the correlation is greater than a preset threshold. That is, only candidate feature parameters whose correlation results are less than the preset threshold, that is, those with a low correlation with any feature sample parameter in the target feature set, can be added to the target feature set.

The target feature set with the newly added candidate feature parameters can be used for the next round of model training.

Furthermore, after the next round of model training is completed, the key feature parameters can be screened from the target feature set and the generation operation of candidate feature parameters based on the key feature parameters can be continued until the prediction model training is completed.

The training conditions may include, for example, that the number of training times reaches the number of prediction times, or that the deviation between the energy consumption prediction value output by the prediction model and the energy consumption prediction sample data is less than a preset deviation threshold.

The following takes the computing task allocation scenario of a data center as an example, combined with the scenario diagram shown in Figure 5, to introduce the technical solution of the embodiment of the present application.

Assuming that data centers 500 are deployed in multiple regions, the server 501 can obtain the first energy consumption data of the target region in the historical time period T~T-L for any region (target region), and use the first prediction model to predict the second energy consumption data of the target region in the prediction time period based on the first energy consumption data. Then, the second prediction model can be used to predict the second carbon emission factor of the target region in the prediction time period based on the second energy consumption data and the first carbon emission factor of the target region in the historical time period published by the authoritative organization.

Among them, T represents the current time step, L can be any natural number, and T-L can represent L time steps before the current time step.

Based on the carbon emission factor prediction sequence of the time period T+1~T+L, the server 501 assumes that it wants to schedule tasks for M hours from T+1~T+M, where M is less than L. The power of the time period T+1~T+M can be combined to calculate the carbon emission amount for each hour in the M hours from T+1~T+M. T+L can represent L time steps after the current time step, and T+M can represent M time steps after the current time step.

The server 501 can migrate the computing tasks of a first data center with a larger carbon emission amount to a second data center with a smaller carbon emission amount, for example, according to the carbon emission amounts corresponding to the multiple regions in the time period T+1 to T+M.

Alternatively, the computing tasks to be processed are allocated to the first data center for computing and processing during the T+N hours when the carbon emissions are relatively small.

FIG6 is a schematic diagram of the structure of an embodiment of a device for predicting a carbon emission factor provided in an embodiment of the present application. The device for predicting a carbon emission factor 600 may include:

A first acquisition module 601 is used to acquire first energy consumption data generated by at least one energy source in a target area during a historical period;

A first prediction module 602, configured to predict second energy consumption data of the at least one energy source within a prediction time period of the target area based on the first energy consumption data of the at least one energy source;

The second acquisition module 603 is used to acquire the first carbon emission factor released by the target area within the historical time period;

The second prediction module 604 is used to predict the second carbon emission factor corresponding to the target area within the prediction time period based on the second energy consumption data of the at least one energy source and at least one characteristic parameter in the first carbon emission factor.

According to an embodiment of the present invention, the device for predicting the carbon emission factor further includes:

The third acquisition module is used to acquire at least one weather data corresponding to the target area within the predicted time period:

According to an embodiment of the present invention, the first prediction module 602 includes:

The first prediction submodule is used to predict the second energy consumption data of the at least one energy within the prediction time period of the target area based on the first energy consumption data of the at least one energy, the at least one weather data, and at least one characteristic parameter in the time data corresponding to the prediction time period.

According to an embodiment of the present invention, the second prediction module 604 includes:

The second prediction submodule is used to predict the second carbon emission factor corresponding to the target area within the prediction time period based on the second energy consumption data of the at least one energy source, the at least one weather data, the time data and at least one characteristic parameter of the first carbon emission factor.

According to an embodiment of the present invention, the first prediction submodule includes:

A first prediction unit is used for predicting, for any energy source, second energy consumption data of the energy source within the prediction time period using a first prediction model based on the first energy consumption data of the energy source, the at least one weather data, and at least one characteristic parameter in the time data corresponding to the prediction time period;

According to an embodiment of the present invention, the second prediction submodule includes:

The second prediction unit is used to predict the second carbon emission factor of the target area within the prediction time period using a second prediction model based on the second energy consumption data of the at least one energy source, the at least one weather data, the time data and at least one characteristic parameter of the first carbon emission factor.

According to an embodiment of the present invention, the device for predicting the carbon emission factor further includes:

A second determination module, used to determine at least one sample characteristic parameter among the first energy consumption sample data, the time sample data, and at least one weather sample data;

A second feature set construction module, configured to construct a target feature set from the at least one sample feature parameter;

A third training module is used to train the first prediction model by using the target feature set as a model input and the energy consumption prediction sample data as a training label;

A second screening module, used to screen at least one key feature parameter from the target feature set;

The second parameter generation module is used to perform an operation on any two key feature parameters to generate candidate feature parameters;

A second calculation module, used for calculating the correlation between the candidate feature parameter and any feature sample parameter in the target feature set;

The fourth training module is used to add candidate feature parameters whose correlations do not meet the correlation requirements into the target feature set, and trigger the third training module to continue executing until the first prediction model meets the training conditions.

According to an embodiment of the present invention, the second screening module includes:

A first result determination unit, configured to determine a first prediction result generated by the first prediction model based on the target feature set;

A screening unit is used to screen at least one key feature parameter from the target feature set based on the first prediction result.

According to an embodiment of the present invention, the device for predicting the carbon emission factor further includes:

A first data determination module is used to obtain, from the historical production data of the target area, first energy consumption data generated in a first time period before the target historical time as first energy consumption sample data, and first energy consumption data generated in a second time period after the target historical time as energy consumption prediction sample data;

The second data determination module is used to use the weather data occurring in the second time period as weather sample data, and the time data corresponding to the second time period as time sample data.

According to an embodiment of the present invention, the device for predicting the carbon emission factor further includes:

A model training module, used for training a plurality of first candidate models based on the first energy consumption sample data, the time sample data, at least one sample characteristic parameter in at least one weather sample data, and energy consumption prediction sample data corresponding to the at least one sample characteristic parameter;

A model evaluation module, used for performing model evaluation on the plurality of first candidate models;

The model determination module is used to select a first candidate model whose model evaluation result meets the performance requirements as the first prediction model.

According to an embodiment of the present invention, the device for predicting the carbon emission factor further includes:

A third determination module is used to determine at least one sample characteristic parameter among the second energy consumption sample data, the time sample data, the carbon emission factor historical sample data and at least one weather sample data;

A third feature set construction module, configured to construct a target feature set from the at least one sample feature parameter;

A fifth training module, used to train the second prediction model by taking the target feature set as a model input and the carbon emission factor prediction sample data as a training label;

A third screening module, used to screen at least one key feature parameter from the target feature set;

The third parameter generation module is used to perform an operation on any two key feature parameters to generate candidate feature parameters;

A third calculation module, used to calculate the correlation between the candidate feature parameter and any feature sample parameter in the target feature set;

The sixth training module is used to add candidate feature parameters whose correlations do not meet the correlation requirements into the target feature set, and trigger the fifth training module to continue executing until the second prediction model meets the training conditions.

According to an embodiment of the present invention, the device for predicting the carbon emission factor further includes:

The first carbon emission calculation module is used to calculate the corresponding carbon emission quantity within any time range within the predicted time period according to the second carbon emission factor corresponding to the target area.

According to an embodiment of the present invention, a data center for providing computing services is deployed in the target area.

According to an embodiment of the present invention, the device for predicting the carbon emission factor further includes:

A prompt information generating module, used for generating recommended prompt information of the data center based on the carbon emission amount;

The prompt information sending module is used to send the recommendation prompt information to the target user.

According to an embodiment of the present invention, a data center is deployed in the target area.

According to an embodiment of the present invention, the device for predicting the carbon emission factor further includes:

A second carbon emission calculation module, used to calculate the corresponding carbon emission amount within any time range within the predicted time period according to the second carbon emission factor corresponding to the target area;

A cost calculation module, used to calculate the corresponding computing cost of the data center within any time range within the forecast time period in combination with the second carbon emission factor corresponding to the target area;

The task allocation module is used to allocate computing tasks according to the computing costs corresponding to data centers in different target areas in different time ranges.

According to an embodiment of the present invention, the first prediction submodule includes:

a fourth determination module, configured to determine at least one initial characteristic parameter consisting of the first energy consumption data, the at least one weather data, and time data representing the predicted time period;

a fifth feature set construction module, configured to construct a first feature set from the at least one initial feature parameter;

A fourth parameter generating module, used for performing an operation on any two feature parameters in the first feature set to generate a target feature parameter;

A fourth calculation module, used to calculate the correlation between the target feature parameter and any feature parameter in the initial feature set;

A feature merging module, used for adding target feature parameters whose correlations do not meet the correlation requirements into the first feature set, and returning to perform an operation on any two feature parameters in the first feature set to generate target feature parameters, until the first feature set meets the feature requirements, thereby obtaining a second feature set;

The energy consumption prediction module is used to use the second feature set to predict the second energy consumption data corresponding to the at least one energy source in the target area within a prediction time period.

The model training device described in FIG6 can execute the carbon emission factor prediction method described in the embodiment shown in FIG1, and its implementation principle and technical effect are not described in detail. The specific manner in which each module and unit performs the operation of the carbon emission factor prediction device in the above embodiment has been described in detail in the embodiment of the method, and will not be described in detail here.

FIG. 7 is a schematic diagram of a structure of an embodiment of a model training device provided in an embodiment of the present application. The model training device 700 may include:

The first determination module 701 is used to determine at least one sample characteristic parameter and a training label corresponding to the at least one sample characteristic parameter; the at least one sample characteristic parameter includes at least one of first energy consumption sample data, time sample data and at least one weather sample data, and the training label includes energy consumption prediction sample data; or the at least one sample characteristic parameter includes at least one of predicted energy consumption sample data, time sample data, carbon emission factor historical sample data and at least one weather sample data, and the training label includes carbon emission factor prediction sample data;

A first feature set construction module 702, configured to construct a target feature set from the at least one sample feature parameter;

A first training module 703, used to train a prediction model using the target feature set and the training labels;

A first screening module 704, configured to screen at least one key feature parameter from the target feature set;

The first parameter generation module 705 is used to perform an operation on any two key feature parameters to generate candidate feature parameters;

A first calculation module 706, configured to calculate the correlation between the candidate feature parameter and any feature sample parameter in the target feature set;

The second training module 707 is used to add candidate feature parameters whose correlations do not meet the correlation requirements to the target feature set, and trigger the first training module to continue executing until the prediction model meets the training conditions.

The model training device described in FIG7 can execute the model training method described in the embodiment shown in FIG5, and its implementation principle and technical effect are not described in detail. The specific manner in which each module and unit performs operations in the model training device in the above embodiment has been described in detail in the embodiment of the method, and will not be described in detail here.

In a possible design, the carbon emission factor prediction device and model training device provided in the embodiment of the present invention may be implemented as a computing device. As shown in FIG8 , the computing device may include a storage component 801 and a processing component 802;

The storage component 801 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 802 to implement the carbon emission factor prediction method and model training method provided in the embodiment of the present invention.

Of course, the computing device may also include other components, such as input/output interfaces, communication components, etc. The input/output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc. The communication component is configured to facilitate wired or wireless communication between the computing device and other devices.

Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

When the computing device is a physical device, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device.

An embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a computer, the carbon emission factor prediction method and model training method provided in the embodiment of the present invention can be implemented.

The embodiment of the present invention further provides a computer program product, including a computer program. When the computer program is executed by a computer, the carbon emission factor prediction method and model training method provided in the embodiment of the present invention can be implemented.

The processing components in the above corresponding embodiments may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing components may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

The storage component is configured to store various types of data to support operations in the device. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM/RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (17)

1. A method for predicting a carbon emission factor, comprising: Acquire first energy consumption data generated by at least one energy source in a target area during a historical time period; Based on the first energy consumption data of the at least one energy source, predicting second energy consumption data of the at least one energy source within a predicted time period of the target area; Obtaining the first carbon emission factor released by the target area during the historical time period; Based on the second energy consumption data of the at least one energy source and at least one characteristic parameter in the first carbon emission factor, a second carbon emission factor corresponding to the target area within the prediction time period is predicted. 2. The method according to claim 1, further comprising: Obtain at least one weather data corresponding to the target area within the forecast time period: The predicting, based on the first energy consumption data of the at least one energy source, second energy consumption data of the at least one energy source within a predicted time period of the target area comprises: Based on the first energy consumption data of the at least one energy source, the at least one weather data, and at least one characteristic parameter in the time data corresponding to the predicted time period, second energy consumption data of the at least one energy source within the predicted time period of the target area is predicted. 3. The method according to claim 2, characterized in that the predicting of the second carbon emission factor corresponding to the target area within the prediction time period based on the second energy consumption data of the at least one energy source and at least one characteristic parameter in the first carbon emission factor comprises: Based on the second energy consumption data of the at least one energy source, the at least one weather data, the time data and at least one characteristic parameter of the first carbon emission factor, a second carbon emission factor corresponding to the target area within the prediction time period is predicted. 4. The method according to claim 3, characterized in that the predicting the second energy consumption data of the at least one energy source within the predicted time period of the target area based on the first energy consumption data, the at least one weather data, and at least one characteristic parameter in the time data corresponding to the predicted time period comprises: For any energy source, based on the first energy consumption data of the energy source, the at least one weather data, and at least one characteristic parameter in the time data corresponding to the prediction time period, using the first prediction model, predict the second energy consumption data of the energy source within the prediction time period; The predicting of the second carbon emission factor corresponding to the target area within the prediction time period based on the second energy consumption data of the at least one energy source, the at least one weather data, the time data and at least one characteristic parameter of the first carbon emission factor includes: Based on the second energy consumption data of the at least one energy source, the at least one weather data, the time data and at least one characteristic parameter of the first carbon emission factor, a second prediction model is used to predict the second carbon emission factor of the target area within the prediction time period. 5. The method according to claim 4, characterized in that the first prediction model is trained and obtained in the following manner: Determine at least one sample characteristic parameter of the first energy consumption sample data, the time sample data, and at least one weather sample data; The at least one sample feature parameter forms a target feature set; Using the target feature set as a model input and the energy consumption prediction sample data as training labels to train the first prediction model; Select at least one key feature parameter from the target feature set; Perform operation on any two key feature parameters to generate candidate feature parameters; Calculating the correlation between the candidate feature parameter and any feature sample parameter in the target feature set; The candidate feature parameters whose correlations do not meet the correlation requirements are added to the target feature set, and the target feature set is returned as the model input, and the energy consumption prediction sample data is used as the training label. The step of training the first prediction model continues until the first prediction model meets the training conditions. 6. The method according to claim 5, characterized in that the step of selecting at least one key feature parameter from the target feature set comprises: Determine a first prediction result generated by the first prediction model based on the target feature set; Based on the first prediction result, at least one key feature parameter is selected from the target feature set. 7. The method according to claim 4, further comprising: From the historical production data of the target area, first energy consumption data generated in a first time period before the target historical time is obtained as first energy consumption sample data, and first energy consumption data generated in a second time period after the target historical time is obtained as energy consumption prediction sample data; The weather data occurring in the second time period is used as weather sample data, and the time data corresponding to the second time period is used as time sample data. 8. The method according to claim 7, further comprising: Training a plurality of first candidate models based on the first energy consumption sample data, the time sample data, at least one sample characteristic parameter in at least one weather sample data, and energy consumption prediction sample data corresponding to the at least one sample characteristic parameter; Performing model evaluation on the multiple first candidate models; The first candidate model whose model evaluation result meets the performance requirements is selected as the first prediction model. 9. The method according to claim 4, characterized in that the second prediction model is trained and obtained in the following manner: Determine at least one sample characteristic parameter among the second energy consumption sample data, the time sample data, the carbon emission factor historical sample data, and at least one weather sample data; The at least one sample feature parameter forms a target feature set; Using the target feature set as model input and the carbon emission factor prediction sample data as training labels, training the second prediction model; Select at least one key feature parameter from the target feature set; Perform operation on any two key feature parameters to generate candidate feature parameters; Calculating the correlation between the candidate feature parameter and any feature sample parameter in the target feature set; The candidate feature parameters whose correlations do not meet the correlation requirements are added to the target feature set, and the target feature set is returned as the model input, and the carbon emission factor prediction sample data is used as the training label. The step of training the second prediction model continues until the second prediction model meets the training conditions. 10. The method according to claim 1, further comprising: According to the second carbon emission factor corresponding to the target area, the corresponding carbon emission quantity within any time range within the predicted time period is calculated. 11. The method according to claim 10, wherein a data center for providing computing services is deployed in the target area, and the method further comprises: Based on the carbon emission amount, generating recommendation prompt information of the data center; The recommendation prompt information is sent to the target user. 12. The method according to claim 1, wherein a data center is deployed in the target area, and the method further comprises: Calculate the corresponding carbon emission amount within any time range within the forecast time period according to the second carbon emission factor corresponding to the target area; In combination with the second carbon emission factor corresponding to the target area, the corresponding computing cost of the data center within any time range within the forecast time period is calculated; Computing tasks are allocated based on the computing costs corresponding to data centers in different target areas in different time ranges. 13. The method according to claim 1, characterized in that the predicting the second energy consumption data corresponding to the at least one energy source in the target area within the prediction time period based on the first energy consumption data, the at least one weather data, and at least one characteristic parameter in the time data representing the prediction time period comprises: determining at least one initial characteristic parameter consisting of the first energy consumption data, the at least one weather data, and time data representing the predicted time period; forming a first feature set by the at least one initial feature parameter; Performing an operation on any two feature parameters in the first feature set to generate a target feature parameter; Calculating the correlation between the target feature parameter and any feature parameter in the initial feature set; Adding target feature parameters whose correlations do not meet the correlation requirements to the first feature set, and returning to perform an operation on any two feature parameters in the first feature set to generate target feature parameters, the step continues until the first feature set meets the feature requirements, thereby obtaining a second feature set; The second feature set is used to predict second energy consumption data corresponding to the at least one energy source in the target area within a prediction time period. 14. A model training method, characterized by comprising: Determine at least one sample characteristic parameter and a training label corresponding to the at least one sample characteristic parameter; the at least one sample characteristic parameter includes at least one of first energy consumption sample data, time sample data, and at least one weather sample data, and the training label includes energy consumption prediction sample data; or the at least one sample characteristic parameter includes at least one of predicted energy consumption sample data, time sample data, carbon emission factor historical sample data, and at least one weather sample data, and the training label includes carbon emission factor prediction sample data; The at least one sample feature parameter forms a target feature set; Using the target feature set and the training labels, training a prediction model; Select at least one key feature parameter from the target feature set; Perform operation on any two key feature parameters to generate candidate feature parameters; Calculating the correlation between the candidate feature parameter and any feature sample parameter in the target feature set; The candidate feature parameters whose correlations do not meet the correlation requirements are added to the target feature set, and the step of training the prediction model using the target feature set and the training labels is returned to continue until the prediction model meets the training conditions. 15. A computing device, comprising a processing component and a storage component; The storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the carbon emission factor prediction method as described in any one of claims 1 to 13, or to implement the model training method as described in claim 14. 16. A computer storage medium, characterized in that a computer program is stored therein, and when the computer program is executed by a computer, the method for predicting the carbon emission factor as described in any one of claims 1 to 13 is implemented, or the model training method as described in claim 14 is implemented. 17. A computer program product, characterized in that the computer program product comprises a computer program code, and when the computer program code is executed by a computer device, the computer device executes the carbon emission factor prediction method described in any one of claims 1 to 13, or executes the model training method described in claim 14.