CN115688989A - Carbon footprint prediction method and terminal based on Stacking ensemble learning - Google Patents
Carbon footprint prediction method and terminal based on Stacking ensemble learning Download PDFInfo
- Publication number
- CN115688989A CN115688989A CN202211287905.2A CN202211287905A CN115688989A CN 115688989 A CN115688989 A CN 115688989A CN 202211287905 A CN202211287905 A CN 202211287905A CN 115688989 A CN115688989 A CN 115688989A
- Authority
- CN
- China
- Prior art keywords
- carbon footprint
- stacking
- learner
- ensemble learning
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a carbon footprint prediction method and a terminal based on Stacking ensemble learning, which are used for acquiring carbon footprint influence factors; acquiring influence factor data corresponding to the carbon footprint influence factors; constructing a learning device for Stacking ensemble learning according to preset evaluation indexes; training a learner according to the influence factor data to obtain a target Stacking integrated learning model; using a Stacking ensemble learning model to predict the carbon footprint; the invention introduces the Stacking integrated learning model which can combine different models and integrate learning, fuses different models, makes up the shortage of single model prediction, can effectively improve the prediction precision by combining the models and the integrated learning, and further ensures the adaptability of the finally selected learner and the content to be predicted by screening through the preset evaluation index when confirming the learner in the Stacking integrated learning model, thereby improving the prediction effect, and training the Stacking integrated learning model according to the reconfirmined influence factors, thereby ensuring the quality of the model.
Description
Technical Field
The invention relates to the field of carbon footprint prediction, in particular to a Stacking ensemble learning-based carbon footprint prediction method and a Stacking ensemble learning-based carbon footprint prediction terminal.
Background
The thermal power industry is a main industry consuming fossil energy and is one of industries with the largest carbon dioxide footprint, and the scientific prediction of the carbon footprint of the thermal power industry and the determination of the carbon peak reaching time and peak value have important significance.
In the prior art, carbon dioxide equivalent calculations are typically used to represent the carbon footprint. Currently, there are two main categories of methods for carbon footprint prediction. The first method is to directly analyze the future carbon footprint trend by using the carbon footprint historical data, the method has less data demand, but has a hysteresis phenomenon in the reaction of data fluctuation, and cannot reflect the influence of the future established policy on the data during prediction. The second method is to build a carbon footprint prediction model considering multi-factors, where the machine learning method is gradually becoming a hotspot tool. But the single model has the limitations of weak learning performance and low prediction precision.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the carbon footprint prediction method and the terminal based on Stacking ensemble learning are provided, and accuracy of carbon footprint prediction is improved.
In order to solve the technical problems, the invention adopts a technical scheme that:
a carbon footprint prediction method based on Stacking ensemble learning comprises the following steps:
acquiring carbon footprint influence factors;
acquiring influence factor data corresponding to the carbon footprint influence factors;
constructing a learning device for Stacking ensemble learning according to a preset evaluation index;
training the learner according to the influence factor data to obtain a target Stacking integrated learning model;
and using the Stacking ensemble learning model to predict the carbon footprint.
In order to solve the technical problem, the invention adopts another technical scheme as follows:
a carbon footprint prediction terminal based on Stacking ensemble learning comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the following steps:
acquiring carbon footprint influence factors;
acquiring influence factor data corresponding to the carbon footprint influence factors;
constructing a learning device for Stacking ensemble learning according to preset evaluation indexes;
training the learner according to the influence factor data to obtain a target Stacking integrated learning model;
and using the Stacking ensemble learning model to predict the carbon footprint.
The invention has the beneficial effects that: the Stacking integrated learning model capable of combining different models and integrating learning is introduced, different models are fused, the defect of single model prediction is overcome, the prediction precision can be effectively improved by combining models and integrated learning, and when the learner in the Stacking integrated learning model is confirmed, screening is carried out through preset evaluation indexes, the adaptability of the finally selected learner and the content to be predicted is further guaranteed, the prediction effect is improved, the Stacking integrated learning model is trained according to the reconfirmed influence factors, and the quality of the model is guaranteed.
Drawings
FIG. 1 is a flowchart illustrating steps of a carbon footprint prediction method based on Stacking ensemble learning according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a Stacking ensemble learning model according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating the evaluation of the results of 9 learners;
FIG. 4 is a schematic diagram illustrating the estimation of the prediction result of the optimal Stacking ensemble learning model according to the embodiment of the present invention;
FIG. 5 is a schematic diagram of a prediction result under different situations based on an optimal Stacking ensemble learning model according to an embodiment of the present invention;
FIG. 6 is a block diagram of a carbon footprint prediction device based on Stacking ensemble learning according to an embodiment of the present invention;
FIG. 7 is another block diagram of a carbon footprint prediction device based on Stacking ensemble learning according to an embodiment of the present invention;
FIG. 8 is another block diagram of a carbon footprint prediction device based on Stacking ensemble learning, in accordance with an embodiment of the present invention;
fig. 9 is a schematic structural diagram of a carbon footprint prediction terminal based on Stacking ensemble learning according to an embodiment of the present invention;
Detailed Description
In order to explain technical contents, achieved objects, and effects of the present invention in detail, the following description is made with reference to the accompanying drawings in combination with the embodiments.
Referring to fig. 1, a carbon footprint prediction method based on Stacking ensemble learning includes the steps of:
acquiring carbon footprint influence factors;
acquiring influence factor data corresponding to the carbon footprint influence factors;
constructing a learning device for Stacking ensemble learning according to preset evaluation indexes;
training the learner according to the influence factor data to obtain a target Stacking integrated learning model;
and using the Stacking ensemble learning model to predict the carbon footprint.
From the above description, the beneficial effects of the present invention are: the Stacking integrated learning model capable of combining different models and integrating learning is introduced, different models are fused, the defect of single model prediction is overcome, the prediction precision can be effectively improved by combining models and integrated learning, and when the learner in the Stacking integrated learning model is confirmed, screening is carried out through preset evaluation indexes, the adaptability of the finally selected learner and the content to be predicted is further guaranteed, the prediction effect is improved, the Stacking integrated learning model is trained according to the reconfirmed influence factors, and the quality of the model is guaranteed.
Further, the acquiring carbon footprint influencing factors comprises:
setting the Kaya identity formula:
wherein C represents the total carbon footprint of the target industry; m i Representing a carbon footprint coefficient for the target industry; n is a radical of i Representing an energy consumption structure of a target industry; s represents the energy consumption intensity of the target industry; t represents the industry structure of the target industry; r represents economic output; p represents the resident general population; i equals 1,2 or 3, representing coal, oil and gas, respectively.
According to the description, the existing Kaya identity formula is improved, so that the existing power generation energy structure in the power industry is more consistent, and the accuracy of subsequent prediction is improved.
Further, in the Kaya identity equation:
wherein, C i Representing a total carbon footprint for the type i energy source; ei represents the total consumption amount of the ith type of energy of the target industry; g 1 Representing the total value of production for the target industry.
According to the description, the total carbon footprint is classified into the total amount corresponding to coal, oil and natural gas respectively, and the total carbon footprint is calculated respectively, so that the comprehensiveness of the influence factors is further improved, and the accuracy of the prediction result is improved.
Further, the step of constructing the learner for the Stacking ensemble learning according to the preset evaluation index comprises the following steps:
acquiring a preset evaluation index and acquiring an alternative learner;
respectively passing the influence factor data through each alternative learner to obtain a prediction result;
and calculating an evaluation value corresponding to each alternative learner according to the prediction result and the preset evaluation index, and confirming a target learner according to the evaluation values.
According to the description, the Stacking integrated learning model comprises a plurality of levels, each level can be constructed by adopting different basic models, the fusion between the models is realized, the effect of making up for deficiencies is achieved, however, if the models are determined to be capable of participating in the construction of the Stacking integrated learning model, the models are inevitably influenced by artificial subjective factors, objective evaluation indexes are introduced, a target learner capable of participating in the construction of the Stacking integrated learning model is confirmed according to evaluation values calculated by the evaluation indexes, the quality of the constructed models is objectively guaranteed, and the accuracy of model prediction results is guaranteed.
Further, the learner comprises a primary learner and a meta learner;
the step of constructing a learning machine for Stacking ensemble learning according to the preset evaluation index includes:
after the target primary learner is confirmed, acquiring an alternative element learner;
respectively enabling the influence factor data to pass through Stacking ensemble learning models with different candidate learners to obtain a prediction result, wherein each Stacking ensemble learning model comprises a target primary learner;
and calculating an evaluation value corresponding to each candidate element learner according to the prediction result and the preset evaluation index, and confirming the target element learner according to the evaluation value.
As can be seen from the above description, the Stacking ensemble learning model includes multiple levels of learners, and after the target primary learners are determined, the Stacking ensemble learning model with the target primary learners is used in the process of calculating the evaluation values, so that meta-learners can be further determined on the basis of the optimal target primary learners that have been confirmed, that is, in the process of confirming the learners, the association among learners in different levels is also considered, and the quality of the Stacking ensemble learning model is further improved.
Further, the evaluation index includes a decision coefficient and a mean absolute percentage error.
From the above description, the decision coefficient and the average absolute percentage error can both evaluate the difference between the predicted value and the actual value, thereby realizing the evaluation of the model prediction effect and knowing the difference between the prediction accuracies of different models.
Further, the calculating an evaluation value corresponding to each candidate learner according to the prediction result and the preset evaluation index, and confirming the target learner according to the evaluation value includes:
in the formula: y is i Representing the real value of the test sample in the ith data set in the influence factor data;
the predicted value of the ith sample obtained by representing the influence factor data; y is mean Mean values representing the true values; m represents the number of test samples in the influence factor data.
From the above description, it can be known that the difference of the prediction accuracy of different models is calculated to be a numerical value for comparison, and the prediction accuracy of different models can be objectively evaluated, so that a proper target learner is selected, and the quality of the finally formed Stacking ensemble learning model is ensured.
Further, using the Stacking ensemble learning model for carbon footprint prediction comprises:
receiving scene setting information;
adjusting model parameters according to the scene setting information;
and (5) using the Stacking ensemble learning model after the model parameters are adjusted to predict the carbon footprint.
As can be seen from the above description, in the actual prediction process, since the future situation has different changes due to different development trends compared with the current situation, the situation setting is introduced in the prediction process of the future trend, and the model parameters are adjusted according to the received situation setting information, so that the future prediction data is more suitable for the actual situation.
Further, the scenario setting information includes reference development scenario setting information, industry optimization scenario setting information, technical breakthrough scenario setting information, and low-carbon development scenario setting information.
According to the description, different scenes are set according to the prediction of different future development trends, the scenes can be directly called to obtain input values when the future is predicted, and the operation cost of the user in the prediction process is simplified.
Referring to the drawings, a carbon footprint prediction terminal based on Stacking ensemble learning comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the steps of the carbon footprint prediction method based on Stacking ensemble learning.
The carbon footprint prediction method based on Stacking ensemble learning can be suitable for scenes needing to predict future carbon footprints, is particularly suitable for predicting the carbon footprints in the thermal power industry, and is explained through a specific implementation mode.
Referring to fig. 1, a first embodiment of the present invention is:
a carbon footprint prediction method based on Stacking ensemble learning specifically comprises the following steps:
s1, acquiring target industry carbon footprint influence factors, including:
s11, setting a Kaya identity formula:
wherein C represents the total carbon footprint of the target industry, namely the total carbon dioxide emission of the target industry; m i Representing a carbon footprint coefficient for the target industry; n is a radical of i Representing an energy consumption structure of a target industry; s represents the energy consumption intensity of the target industry; t represents the industry structure of the target industry; r represents economic output; p represents a resident general population, and is specifically explained in the following table 1 by taking the thermal power industry as an example;
wherein, C i Representing the total carbon footprint amount of the ith type of energy in the target industry, wherein i is 1,2 or 3, and respectively representing coal, oil and natural gas; ei represents the total consumption amount of ith type energy of the target industry; e is the total energy consumption of the target industry, such as the total energy consumption of thermal power; g 1 Representing a total production value of a target industry, such as a total thermal power production value; g is the total production value of all industries; according to different specific analysis objects, a Kaya identity formula is subjected to extended decomposition by combining the characteristics of thermal power enterprises, fossil energy coal, petroleum and natural gas which are mainly consumed by the thermal power enterprises are introduced, and the carbon footprint of the thermal power enterprises can be analyzed more accurately;
TABLE 1
S2, acquiring influence factor data corresponding to the carbon footprint influence factors;
in an alternative embodiment, referring to table 1, in the case that influence factor data can be supported, the urbanization level L is preferentially obtained as the number of population for calculation, and town has a greater influence on carbon footprint and can reflect the carbon footprint level more accurately;
in an optional implementation manner, dimensionless normalization processing is performed on the influence factor data, influences of different measurement units of different types of influence factor data on the prediction result are eliminated, and the problem that different influence factors have different dimensions and are not comparable is solved, for example, an average normalization method can be adopted:
wherein, the first and the second end of the pipe are connected with each other,
is a normalized numerical value; xij is a sample characteristic value; m is the number of samples;
the average value of the sample characteristics is taken; wherein the sample characteristics are the influencing factor data mentioned above;
s3, building a Stacking integrated learning learner according to preset evaluation indexes, wherein the Stacking integrated learning model can build a multi-layer network model, each layer is composed of one or more different learners, and the idea is that a plurality of weak supervision models are combined to obtain a strong supervision model, so that the advantages of different learners are fully exerted, and the prediction accuracy of the whole model is improved; the Stacking of the two-layer structure can not only strengthen the learning effect but also not cause the model to be complicated; therefore, in the embodiment, a two-layer Stacking ensemble learning prediction model is constructed, the structure of which is shown in fig. 2, the model is divided into two levels, a learner in the first layer is called a primary learner, and a learner in the second layer is called a meta-learner; the method comprises the following steps:
s31, acquiring a preset evaluation index and acquiring an alternative learner;
in an alternative embodiment, KNN (K-nearest neighbor), LR (Logistic Regression), DT (Decision Tree algorithm), BP (back propagation, back propagation algorithm), SVM (support vector machine), and RF (Random Forest), adaBoost (Adaptive Boosting, step-by-step Boosting), GBDT (Gradient Boosting Decision Tree), and XGBoost (Extreme Gradient Boosting ) based on Bagging parallel integration and Boosting serial integration as representatives in machine learning are used as candidate learners;
in an optional embodiment, the preset evaluation index includes a decision coefficient and an average absolute percentage error;
s32, respectively passing the influence factor data through each alternative learner to obtain a prediction result;
for example, known influence factor data corresponding to 6 carbon footprint influence factors in table 1 are used as characteristic variables, and known carbon footprint quantities are used as target variables to form a data set for model training; the known influence factor data and the carbon footprint amount are the influence factor data and the carbon footprint amount in the historical data; dividing a data set into a training set and a testing set according to the proportion of 7:3, and respectively establishing carbon footprint prediction models of the 9 learners by adopting a K-fold cross validation and grid parameter optimization method;
s33, calculating an evaluation value corresponding to each alternative learner according to the prediction result and the preset evaluation index, and confirming a target learner according to the evaluation values;
calculating a decision coefficient evaluation value:
calculate mean absolute percentage error estimate:
wherein, y i Representing the true value of the test sample in the ith data set in the training model;
representing the predicted value of the ith sample obtained by training the model; y is mean Mean values representing the true values; m represents the number of test samples; the closer the evaluation value of the decision coefficient is to 1, the closer the evaluation value of the average absolute percentage error is to 0, and the higher the prediction precision of the model is;
here, the data set used to train the model may be obtained directly from the influencing factor data;
in an optional implementation mode, a Stacking ensemble learning model is constructed according to the target learners with the preset number of confirmed evaluation values;
in an alternative embodiment, the learner comprises a primary learner and a meta learner;
after executing steps S31-S33 to construct the primary learner, the method further includes:
repeatedly executing S31-S33 to construct the meta learner, and replacing the step S32 with: respectively enabling the influence factor data to pass through Stacking ensemble learning models with different candidate learners to obtain a prediction result, wherein each Stacking ensemble learning model comprises a target primary learner;
for example, after training a model from a data set is completed, a predicted value of the model is obtained, and a decision coefficient (R) is used 2 ) The evaluation value evaluates the model predictive ability, and referring to fig. 3, the evaluation results are: LR (R2 = 0.9886)>GBDT(R2＝0.9877)>BP(R2＝0.9839)>XGBoost(R2＝0.9720)>RF (R2 = 0.9570); evaluating the model prediction ability by using the MAPE index, wherein the evaluation result is as follows: GBDT (MAPE = 0.0089)>BP(MAPE＝0.0116)>LR(MAPE＝0.0121)>XGBoost (MAPE＝0.0186)>AdaBoost (MAPE = 0.0247); whether or not with R 2 Indexes or MAPE indexes can be obtained, and the carbon footprint prediction effect independently established by the LR, GBDT, BP and XGboost in the 9 optional learners is the best, and the indexes or MAPE indexes can be used as candidate objects of the primary learner; in addition, the method just belongs to four categories of symbolic meaning, connection meaning, bagging parallel and boosting serial of machine learning, so that a final Stacking model can be more comprehensive;
in StackiOn the premise that the primary learning layer of ng ensemble learning is determined, training of a prediction model is performed by using the first 9 learners as meta learners, and R is used as well 2 The results are shown in table 2, in comparison with the MAPE index:
TABLE 2
| Meta learning device | R2 | MAPE |
| Stacking-AdaBoost | 0.9710 | 0.0185 |
| Stacking-BP | 0.9884 | 0.0111 |
| Stacking-DT | 0.9792 | 0.0154 |
| Stacking-GBDT | 0.9947 | 0.0058 |
| Stacking-KNN | 0.9709 | 0.0298 |
| Stacking-LR | 0.9839 | 0.0099 |
| Stacking-RF | 0.9746 | 0.0276 |
| Stacking-SVM | 0.9476 | 0.0437 |
| Stacking-XGBoost | 0.9717 | 0.0178 |
As can be seen from Table 2, when GBDT is selected as the meta-learner, R 2 Can reach 0.9947, MAPE can reach 0.0058, and the prediction performance is superior to other combinations; therefore, determining the GBDT as a meta-learner of the Stacking ensemble learning model; the schematic diagram of the carbon footprint prediction result of the Stacking-GBDT is shown in FIG. 4;
s4, training the learner according to the influence factor data to obtain a target Stacking integrated learning model;
s5, predicting carbon footprints by using the Stacking ensemble learning model;
in an alternative embodiment, the method comprises the following steps: .
S51, receiving scene setting information, wherein the scene setting information comprises benchmark development scene setting information, industry optimization scene setting information, technical breakthrough scene setting information and low-carbon development scene setting information;
s52, adjusting model parameters according to the scene setting information;
in an alternative embodiment, for each scenario, the corresponding model parameters are matched, as shown in tables 3-4 below:
TABLE 3
TABLE 4
Referring to table 4, wherein the rate is a change rate compared to the previous year data, since the prediction of the future is based on the data corresponding to the influence factor data acquired in step S2, i.e. historically collected data, for example, the historical data is to 2020, the population resident in the predicted data 2021-2025 is increased at a rate of 0.60%, the economic output is increased at a rate of 6.3%, the industrial structure is decreased at a rate of-1.5%, the energy consumption intensity is decreased at a rate of-0.73%, and the energy consumption structure is decreased at a rate of-0.04% when predicted in the baseline development scenario, corresponding to table 4; the data changes in other years and under other conditions are the same on the basis of table 4, and are not described again here;
s53, using the Stacking integrated learning model after the model parameters are adjusted to predict the carbon footprint, wherein the carbon footprint prediction comprises carbon peak reaching time and a peak value;
the future carbon peak reaching situation prediction analysis of the 4 low-carbon development modes is based on an established Stacking-GBDT ensemble learning carbon footprint prediction model, and can predict the carbon footprint in 2021-2035 years; a schematic diagram of the carbon footprint prediction results for 4 different scenarios is shown in fig. 5.
From the schematic diagram of fig. 5, the following conclusions can be drawn:
the standard development scenario: the setting of the scene is that the scene is approximately developed according to the planning of government-related documents on the premise of not considering other factors to change suddenly; as can be seen from the schematic diagram of fig. 5, under this development scenario, the thermal power industry cannot reach a carbon peak by 2030 years, the carbon peak is expected to be realized in 2032, the peak value of the highest carbon footprint will reach 298 million tons (t), and then the peak value shows a weak descending trend year by year; therefore, on the premise of the situation, certain measures are still needed to be taken to control the development rate of the factors related to the carbon footprint of the thermal power enterprise;
and (3) industrial optimization scenarios: the setting of the scene is to reduce the development mode of the second industry share by optimizing the industry structure; as can be seen from the schematic diagram in fig. 5, under the development scenario, the thermal power industry can reach a carbon peak in 2029, the peak value of the maximum carbon footprint is 296 million t, and then the trend is slightly decreased year by year; compared with the standard development scene, the difference of the carbon footprint peak value is not large, but the carbon peak can be reached 3 years ahead; therefore, the low-carbon and carbon-free energy can be developed by further depending on the advantages of nuclear power, wind power and the like, the energy storage industry, the digital energy industry, the advanced manufacturing industry and the modern service industry can be cultivated, and the industrial structure can be optimized;
the technical breakthrough situation is as follows: the setting of the scene is a development mode for reducing energy consumption by improving energy technology; as can be seen from the schematic diagram in fig. 5, under the development scenario, the thermal power industry can reach a carbon peak in 2028, the peak value of the maximum carbon footprint is 288 million t, and then the thermal power industry is in a weak descending trend year by year; compared with the industrial optimization situation, the mode of improving the energy technology can effectively reduce the carbon footprint and can realize the carbon peak reaching earlier; therefore, a novel power system taking new energy as a main body can be further constructed, and the development of low-carbon, carbon-free and negative-carbon technologies is promoted;
the low-carbon development scenario is as follows: the setting of the scene is a development mode of intervention control in all aspects from the aspects of population, economy, industry and energy; as can be seen from the schematic diagram in fig. 5, under the development scenario, the thermal power industry can reach a carbon peak in 2027, the peak value of the maximum carbon footprint is 278 million t, and then the peak value shows a weak descending trend year by year; compared with the technical breakthrough situation, the comprehensive intervention mode can effectively reduce the carbon footprint and promote the realization of carbon peak-reaching earlier; however, it is also necessary to consider whether the overall intervention mode would bring about additional negative effects, such as affecting the economic benefits of the thermal power industry.
Referring to fig. 6, the second embodiment of the present invention is:
a carbon footprint prediction device based on Stacking ensemble learning can realize the carbon footprint prediction method based on Stacking ensemble learning, and comprises an influence factor determination module 601, a data processing module 602, an ensemble learning learner selection module 603 and a carbon scene prediction module 604.
The influence factor determining module 601 is used for identifying influence factors of the carbon footprint of the thermal power plant based on the improved Kaya identity; the relevant factors identified by Kaya identity may continue to be expanded as desired, and are not limited to the existing influencing factors in table 1.
The data processing module 602 is configured to input and store influence factor data corresponding to the carbon footprint influence factor, and perform dimensionless normalization processing on the data, so as to facilitate training of the following prediction model.
The ensemble learning learner selection module 603 is used for training a carbon footprint prediction model of the Stacking ensemble learning model; as shown in fig. 7, the ensemble learning learner selection module 603 includes: an index unit 701, a primary learner selection unit 702, a meta learner selection unit 703, and a prediction result display unit 704.
Further, the index unit 701 is configured to select an evaluation index for evaluating the quality of the prediction model, and may select only one of the evaluation index R2 (decision coefficient) and the evaluation index MAPE (mean absolute percentage error), or may select both the evaluation index and the evaluation index simultaneously. Meanwhile, more other evaluation indexes can be embedded.
Further, the primary learner selecting unit 702 is configured to select several learners from the multiple learners to complete the establishment of the Stacking ensemble learning primary layer. The number of the primary learners can be increased or decreased according to the requirement. The unit provides default primary layers for reference by the user.
Further, the meta-learner selecting unit 703 is configured to select one learner from the multiple learners to complete the establishment of the Stacking ensemble learning meta-level layer. The unit provides default primary layers for reference by the user.
Furthermore, the prediction structure display unit is used for outputting the prediction model result. The result output has two aspects: on one hand, the index value is output; another aspect is the output of the "year-carbon footprint value".
The carbon scenario prediction module 604 is configured to select different carbon footprint development routes for carbon scenario prediction. As shown in fig. 8, the module designs four development mode units by default, including: a reference development scenario unit 801, an industry optimization scenario unit 802, a technical breakthrough scenario unit 803, and a low-carbon development scenario unit 804; the development mode can also adjust the rate of the development influencing factors to change according to the user requirements so as to meet timeliness, and a custom unit 803 is formed.
Further, the reference development scenario unit 801 assumes that the current economic and technical environment is unchanged and a government cannot take new emission reduction measures according to the characteristics of the past thermal power industry development, and the main driving force of the development is still economic production, so that the development of society and technology is promoted.
Further, the industry optimization scenario unit 802 is based on the existing policy, further optimizes and modernizes the industry structure, adjusts the industry structure, reduces the development participation of the second industry, reduces the proportion of the traditional industry, and enables the high and new technology industry and the service industry to become the leading industry.
Further, the technology breakthrough scenario unit 803 enhances scientific and technological research and development through system design, technical progress and structure transformation on the basis of the reference development scenario, promotes resource circulation and efficient utilization in the thermal power industry, and adjusts the value of the energy consumption intensity and the energy consumption structure change rate.
Furthermore, the industry optimization scenario unit 804 strengthens an energy policy, implements a series of energy conservation and emission reduction measures, and actively adjusts the energy structure of a power generation end on the basis of the existing policy in the thermal power industry, so that a low-carbon footprint is realized.
Further, the self-defined scenario is to redefine the development rate of each influence factor according to the user requirement.
Referring to fig. 9, a third embodiment of the present invention is:
a carbon footprint prediction terminal 900 based on Stacking ensemble learning, comprising a processor 901, a memory 902 and a computer program stored on the memory 902 and operable on the processor 901, wherein the processor 2 implements the steps of the first embodiment when executing the computer program;
specifically, the carbon footprint prediction terminal 900 based on Stacking ensemble learning may include one or more of the following components: processing components 901, memory 902, power components 903, input/output (I/O) interfaces 904, and communication components 905.
Further, the processing component 901 generally controls overall operation of the predictive device 900, such as operations associated with display, data communication, and recording operations. The processing component 901 may include one or more processors 920 to execute instructions to perform all or part of the steps of the methods described above. The processor may be a central processing unit, but may also be other general purpose processors, digital signal processors, application specific integrated circuits, etc.
Further, the memory 902 is configured to store various types of data to support operations at the predictive device 900. Examples of such data include instructions, data, messages, pictures, etc. for any application or method operating on terminal device 900. The memory 902 may be implemented by any type of volatile or non-volatile storage device or combination thereof.
Further, a power supply component 903 provides power to the various components of the terminal device 900. The power components 903 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the predictive device 900.
Further, I/O interface 904 provides an interface between processing component 901 and peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to: a home button, a volume button, a start button, and a lock button.
Further, the communication component 905 is configured to facilitate communication between the predictive device 900 and other devices in a wired or wireless manner. The predictive device 900 may access a wireless network based on a communication standard, such as WiFi, 4G or 5G, or a combination thereof.
In conclusion, the invention provides a carbon footprint prediction method and a terminal based on Stacking ensemble learning, wherein a Stacking ensemble learning model is introduced, a plurality of machine learning models can be integrated for prediction, the problems of bias and incompleteness of single model prediction are avoided, two objective evaluation indexes, namely a decision coefficient and an average absolute percentage error, are introduced when a model is selected as a learner in the Stacking ensemble learning model for judging the quality of model prediction, a model which shows relatively more excellent performance is added into the Stacking ensemble learning model, the quality of the Stacking ensemble learning model is further ensured, and meanwhile, the Stacking ensemble learning model can have multiple layers; in addition, in the process of prediction, a set contextual model is provided for selection, different results can be predicted according to different future trends, comprehensiveness of the predicted results is improved, accuracy is guaranteed, meanwhile, a user can adjust contextual parameters according to self understanding of the trends, and finally the Stacking ensemble learning model is more flexibly used.
The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all equivalent changes made by using the contents of the present specification and the drawings, or applied directly or indirectly to the related technical fields, are included in the scope of the present invention.
Claims (10)
1. A carbon footprint prediction method based on Stacking ensemble learning is characterized by comprising the following steps:
acquiring target industry carbon footprint influence factors;
acquiring influence factor data corresponding to the carbon footprint influence factors;
constructing a learning device for Stacking ensemble learning according to preset evaluation indexes;
training the learner according to the influence factor data to obtain a target Stacking integrated learning model;
and using the Stacking ensemble learning model to predict the carbon footprint.
2. The carbon footprint prediction method based on Stacking ensemble learning of claim 1, wherein said obtaining carbon footprint influencing factors comprises:
setting the Kaya identity formula:
wherein C represents the total carbon footprint of the target industry; i is equal to 1,2 or 3, representing coal, oil and gas, respectively; m is a group of i Representing a carbon footprint coefficient for the target industry; n is a radical of i Representing an energy consumption structure of a target industry; s represents the energy consumption intensity of the target industry; t represents the industry structure of the target industry; r represents economic output; p represents the resident general population.
3. The carbon footprint prediction method based on Stacking ensemble learning of claim 2, wherein in the Kaya identity formula:
wherein, C i Representing a total carbon footprint for the type i energy source; ei represents the total consumption amount of the ith type of energy of the target industry; g 1 Representing the total production value of the target industry.
4. The carbon footprint prediction method based on Stacking ensemble learning as claimed in claim 1, wherein the step of constructing a learner of the Stacking ensemble learning according to the preset evaluation index comprises:
acquiring a preset evaluation index and acquiring an alternative learner;
respectively passing the influence factor data through each alternative learner to obtain a prediction result;
and calculating an evaluation value corresponding to each alternative learner according to the prediction result and the preset evaluation index, and confirming a target learner according to the evaluation values.
5. The carbon footprint prediction method based on Stacking ensemble learning of claim 4, wherein said learner comprises a primary learner and a meta-learner;
the step of constructing a learning machine for Stacking ensemble learning according to the preset evaluation index includes:
after the target primary learner is confirmed, the alternative meta-learner is obtained;
respectively enabling the influence factor data to pass through Stacking ensemble learning models with different candidate learners to obtain a prediction result, wherein each Stacking ensemble learning model comprises a target primary learner;
and calculating an evaluation value corresponding to each alternative element learner according to the prediction result and the preset evaluation index, and confirming a target element learner according to the evaluation value.
6. The carbon footprint prediction method based on Stacking ensemble learning of claim 4, wherein said evaluation index comprises a decision coefficient and a mean absolute percentage error.
7. The carbon footprint prediction method based on Stacking ensemble learning according to claim 6, wherein the calculating an evaluation value corresponding to each candidate learner according to the prediction result and the preset evaluation index, and confirming the target learner according to the evaluation value comprises:
in the formula: y is i Representing the true values of the test samples in the ith data set in the influence factor data;
the predicted value of the ith sample obtained by representing the influence factor data; y is mean Mean values representing the true values; m represents the number of test samples in the influence factor data.
8. The carbon footprint prediction method based on Stacking ensemble learning of claim 1, wherein using said Stacking ensemble learning model for carbon footprint prediction comprises:
receiving scene setting information;
adjusting model parameters according to the scene setting information;
and (5) using the Stacking ensemble learning model after the model parameters are adjusted to predict the carbon footprint.
9. The carbon footprint prediction method based on Stacking ensemble learning of claim 8, wherein the scenario setting information comprises benchmark development scenario setting information, industry optimization scenario setting information, technology breakthrough scenario setting information, and low carbon development scenario setting information.
10. A carbon footprint prediction terminal based on Stacking ensemble learning, comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any one of claims 1 to 9 in the carbon footprint prediction based on Stacking ensemble learning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211287905.2A CN115688989A (en) | 2022-10-20 | 2022-10-20 | Carbon footprint prediction method and terminal based on Stacking ensemble learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211287905.2A CN115688989A (en) | 2022-10-20 | 2022-10-20 | Carbon footprint prediction method and terminal based on Stacking ensemble learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115688989A true CN115688989A (en) | 2023-02-03 |
Family
ID=85067350
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211287905.2A Pending CN115688989A (en) | 2022-10-20 | 2022-10-20 | Carbon footprint prediction method and terminal based on Stacking ensemble learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115688989A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116542380A (en) * | 2023-05-09 | 2023-08-04 | 武汉智网兴电科技开发有限公司 | Power plant supply chain carbon footprint optimization method and device based on natural language |
| CN116843080A (en) * | 2023-07-12 | 2023-10-03 | 济南明泉数字商务有限公司 | Machine learning-based carbon element footprint prediction method and system for urea production |
-
2022
- 2022-10-20 CN CN202211287905.2A patent/CN115688989A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116542380A (en) * | 2023-05-09 | 2023-08-04 | 武汉智网兴电科技开发有限公司 | Power plant supply chain carbon footprint optimization method and device based on natural language |
| CN116542380B (en) * | 2023-05-09 | 2023-11-14 | 武汉智网兴电科技开发有限公司 | Power plant supply chain carbon footprint optimization method and device based on natural language |
| CN116843080A (en) * | 2023-07-12 | 2023-10-03 | 济南明泉数字商务有限公司 | Machine learning-based carbon element footprint prediction method and system for urea production |
| CN116843080B (en) * | 2023-07-12 | 2024-03-19 | 济南明泉数字商务有限公司 | Machine learning-based carbon element footprint prediction method and system for urea production |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115688989A (en) | Carbon footprint prediction method and terminal based on Stacking ensemble learning | |
| CN111695793A (en) | Method and system for evaluating energy utilization flexibility of comprehensive energy system | |
| CN106097055A (en) | Enterprise order processing method under personalized customization demand | |
| CN110598923A (en) | Air conditioner load prediction method based on support vector regression optimization and error correction | |
| CN110837915B (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
| CN113887916A (en) | Dynamic quantitative evaluation method and system for line loss of power distribution network | |
| CN113112090B (en) | Space load prediction method based on principal component analysis of comprehensive mutual informativity | |
| CN109858798B (en) | Power grid investment decision modeling method and device for correlating transformation measures with voltage indexes | |
| CN109697566A (en) | Electronic product processing technology evaluation system and its evaluation method | |
| CN112329969A (en) | Building intelligent engineering investment prediction method based on support vector machine | |
| CN111680452A (en) | Power grid engineering accurate investment decision simulation method based on full-factor data mining | |
| CN103700030A (en) | Grey rough set-based power grid construction project post-evaluation index weight assignment method | |
| CN115345297A (en) | Platform area sample generation method and system based on generation countermeasure network | |
| CN110880044A (en) | Markov chain-based load prediction method | |
| WO2024082836A1 (en) | Optimization method for comprehensive benefit evaluation scheme for water-wind-photovoltaic energy storage multi-energy complementary system | |
| CN110837952A (en) | Game theory-based power grid new technology equipment selection method and system | |
| CN116823008A (en) | Park energy utilization efficiency evaluation method, system, equipment and storage medium | |
| CN115471098A (en) | Carbon emission acquisition method and device, computer equipment and storage medium | |
| CN114238102A (en) | Staged workload prediction method and system based on composite variable characteristics | |
| CN114004065A (en) | Transformer substation engineering multi-objective optimization method based on intelligent algorithm and environmental constraints | |
| CN109598375B (en) | Energy demand structure prediction method based on energy demand total quantity improvement | |
| Shiyun et al. | Data analysis and cloud computing of power grid infrastructure projects under the background of informatization | |
| Chen et al. | The Application of Adaptive Generalized NGBM (1, 1) To Sales Forecasting: A Case Study of an Underwear Shop. | |
| CN110807599A (en) | Method, device, server and storage medium for deciding electrochemical energy storage scheme | |
| Hinova et al. | Risk Analysis and Assessment when Investing in Energy from Natural Gas and Photovoltaics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |