CN116029466B - Carbon emission prediction method, device, storage medium and equipment - Google Patents
Carbon emission prediction method, device, storage medium and equipment Download PDFInfo
- Publication number
- CN116029466B CN116029466B CN202310315526.8A CN202310315526A CN116029466B CN 116029466 B CN116029466 B CN 116029466B CN 202310315526 A CN202310315526 A CN 202310315526A CN 116029466 B CN116029466 B CN 116029466B
- Authority
- CN
- China
- Prior art keywords
- carbon emission
- model
- influence factor
- moment
- 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.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/80—Management or planning
- Y02P90/84—Greenhouse gas [GHG] management systems
Abstract
The embodiment of the application discloses a carbon emission prediction method, a device, a storage medium and equipment, wherein the method comprises the following steps: each carbon emission influence factor predictor model is operated for n2 times in an iterative mode, and each iterative process comprises the following steps: based on each carbon emission influence factor predictor model and input data thereof, obtaining the characteristics of the carbon emission influence factors from the i time to the i+n1-1 time and the predicted values of the carbon emission influence factors from the i+n1 time, wherein the input data comprise the original data of the carbon emission influence factors from the i time to the i+n1-1 time, and each iteration is performed once from i=1; inputting the carbon emission influence factor characteristics obtained by each iteration of the N carbon emission influence factor predictor models into a fully-connected neural network for operation to obtain multi-carbon emission influence factor combination characteristics corresponding to each iteration; based on the carbon emission prediction sub-model and input data thereof, a carbon emission prediction value at the time of n1+n2 is obtained, and the input data comprises a sequence formed by n2 multi-carbon emission influence factor combination characteristics.
Description
Technical Field
The embodiment of the application belongs to the technical field of artificial intelligence, and particularly relates to a carbon emission prediction method, a device, a storage medium and equipment.
Background
Carbon emissions are one of the important factors affecting the air environment and global warming, and thus, it is important to predict the carbon emissions. In order to rapidly predict the future carbon emission, model training may be generally performed based on the carbon emission at a plurality of historic times to obtain a carbon emission prediction model capable of predicting the carbon emission at the next time. However, there are many factors affecting the carbon emission amount, and the accuracy is low by predicting the carbon emission amount at the next time only from the carbon emission amounts at a plurality of historic times.
Disclosure of Invention
The application provides a carbon emission prediction method, a device, a storage medium and equipment, which can improve the accuracy of carbon emission prediction.
The specific technical scheme is as follows:
in a first aspect, an embodiment of the present application provides a carbon emission prediction method, where the method is applied to a carbon emission prediction model, where the carbon emission prediction model includes N carbon emission influencing factor predictor models, a fully connected neural network, and a carbon emission prediction submodel, where different carbon emission influencing factor predictor models in the N carbon emission influencing factor predictor models are respectively used to predict values of different carbon emission influencing factors, and the method includes:
And respectively carrying out iterative operation on each carbon emission influence factor predictor model n2 times, wherein each iterative process comprises the following steps: obtaining a carbon emission influence factor characteristic from an ith moment to an (i+n 1) -1 moment and a carbon emission influence factor predicted value from the (i+n 1) -1 moment based on input data of each carbon emission influence factor predicted sub-model and the carbon emission influence factor predicted sub-model, wherein the input data of the carbon emission influence factor predicted sub-model comprises carbon emission influence factor raw data from the ith moment to the (i+n 1) -1 moment, the carbon emission influence factor raw data at a target moment comprises carbon emission influence factor basic data at the target moment and a carbon emission measurement value corresponding to the target moment, the target moment is any moment, i=i+1 is iterated each time, the N, the i, the N1 and the N2 are positive integers, the N1 is a step length of the carbon emission influence factor predicted sub-model, and the N2 is a step length of the carbon emission amount predicted sub-model;
inputting the carbon emission influence factor characteristics obtained by N carbon emission influence factor predictor models into the fully-connected neural network for operation aiming at the carbon emission influence factor characteristics obtained by each iteration to obtain multi-carbon emission influence factor combination characteristics corresponding to each iteration;
And obtaining a carbon emission predicted value at the time of n < 1+ > n < 2 >, based on the carbon emission predicted sub-model and the input data of the carbon emission predicted sub-model, wherein the input data of the carbon emission predicted sub-model comprises a sequence formed by n < 2 > multi-carbon emission influence factor combination characteristics obtained by n < 2 > iterations.
In one embodiment, the carbon emission value corresponding to the target time includes any one of a carbon emission true value at the target time, a carbon emission predicted value at the target time, and a carbon emission average value at m times before the target time, where m is a positive integer; and/or the number of the groups of groups,
the carbon emission influencing factor basic data of the target moment comprises a carbon emission influencing factor true value of the target moment or a carbon emission influencing factor predicted value of the target moment.
In one embodiment, the carbon emission value at the target time is regarded as the priority of the carbon emission value corresponding to the target time, the carbon emission predicted value at the target time is regarded as the priority of the carbon emission value corresponding to the target time, and the carbon emission average value at m times before the target time is regarded as the priority of the carbon emission value corresponding to the target time.
In one embodiment, during the training phase of the carbon emission influencing factor predictor model, the input data of the carbon emission influencing factor predictor model further comprises a carbon emission influencing factor true value at the i+n1 time; and/or the number of the groups of groups,
in the training stage of the carbon emission quantity prediction sub-model, the input data of the carbon emission quantity prediction sub-model further comprises a carbon emission quantity true value at the n < 1 > +n2 time, or a carbon emission quantity true value after Gaussian filtering is carried out on the carbon emission quantity true value at the n < 1 > +n2 time.
In one embodiment, the carbon emission influencing factor predictor model is a long-short term memory artificial neural network LSTM model and/or the carbon emission predictor model is a LSTM model.
In a second aspect, an embodiment of the present application provides a carbon emission prediction apparatus, where the apparatus is applied to a carbon emission prediction model, where the carbon emission prediction model includes N carbon emission influencing factor prediction sub-models, a fully connected neural network, and a carbon emission prediction sub-model, where different carbon emission influencing factor prediction sub-models in the N carbon emission influencing factor prediction sub-models are respectively used to predict values of different carbon emission influencing factors, and the apparatus includes:
The carbon emission influence factor processing unit is used for respectively and iteratively operating each carbon emission influence factor prediction sub-model n2 times, and each iteration process comprises the following steps: obtaining a carbon emission influence factor characteristic from an ith moment to an (i+n 1) -1 moment and a carbon emission influence factor predicted value from the (i+n 1) -1 moment based on input data of each carbon emission influence factor predicted sub-model and the carbon emission influence factor predicted sub-model, wherein the input data of the carbon emission influence factor predicted sub-model comprises carbon emission influence factor raw data from the ith moment to the (i+n 1) -1 moment, the carbon emission influence factor raw data at a target moment comprises carbon emission influence factor basic data at the target moment and a carbon emission measurement value corresponding to the target moment, the target moment is any moment, i=i+1 is iterated each time, the N, the i, the N1 and the N2 are positive integers, the N1 is a step length of the carbon emission influence factor predicted sub-model, and the N2 is a step length of the carbon emission amount predicted sub-model;
the fully-connected processing unit is used for inputting the carbon emission influence factor characteristics obtained by the N carbon emission influence factor predictor models into the fully-connected neural network for operation aiming at the carbon emission influence factor characteristics obtained by each iteration to obtain multi-carbon emission influence factor combination characteristics corresponding to each iteration;
And the carbon emission prediction unit is used for obtaining a carbon emission prediction value at the time of n < 1+ > n < 2 >, based on the carbon emission prediction sub-model and the input data of the carbon emission prediction sub-model, wherein the input data of the carbon emission prediction sub-model comprises a sequence formed by n < 2 > multi-carbon emission influence factor combination characteristics obtained by n < 2 > iterations.
In one embodiment, the carbon emission value corresponding to the target time includes any one of a carbon emission true value at the target time, a carbon emission predicted value at the target time, and a carbon emission average value at m times before the target time, where m is a positive integer; and/or the number of the groups of groups,
the carbon emission influencing factor basic data of the target moment comprises a carbon emission influencing factor true value of the target moment or a carbon emission influencing factor predicted value of the target moment.
In one embodiment, the carbon emission value at the target time is regarded as the priority of the carbon emission value corresponding to the target time, the carbon emission predicted value at the target time is regarded as the priority of the carbon emission value corresponding to the target time, and the carbon emission average value at m times before the target time is regarded as the priority of the carbon emission value corresponding to the target time.
In one embodiment, during the training phase of the carbon emission influencing factor predictor model, the input data of the carbon emission influencing factor predictor model further comprises a carbon emission influencing factor true value at the i+n1 time; and/or the number of the groups of groups,
in the training stage of the carbon emission quantity prediction sub-model, the input data of the carbon emission quantity prediction sub-model further comprises a carbon emission quantity true value at the n < 1 > +n2 time, or a carbon emission quantity true value after Gaussian filtering is carried out on the carbon emission quantity true value at the n < 1 > +n2 time.
In one embodiment, the carbon emission influencing factor predictor model is a long-short term memory artificial neural network LSTM model and/or the carbon emission predictor model is a LSTM model.
In a third aspect, embodiments of the present application provide a storage medium having stored thereon executable instructions that when executed by a processor cause the processor to implement a method according to any embodiment of the first aspect.
In a fourth aspect, an embodiment of the present application provides an electronic device, including:
one or more processors;
storage means for storing one or more programs,
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of the embodiments of the first aspect.
As can be seen from the foregoing, the carbon emission amount prediction method, apparatus, storage medium and device provided in the embodiments of the present application can obtain the carbon emission influence factor characteristics from the i-th time to the i+n1-1-th time and the carbon emission influence factor predicted value from the i+n1-th time based on each carbon emission influence factor prediction sub-model of the N carbon emission influence factor prediction sub-models and the input data thereof, where the input data of the carbon emission influence factor prediction sub-models includes the carbon emission influence factor raw data from the i=1 to the i+n1-th time, iterate each carbon emission influence factor prediction sub-model N2 times according to the foregoing method, then input the carbon emission influence factor characteristics obtained by the N carbon emission influence factor prediction sub-models into the fully connected neural network for operation with respect to the carbon emission influence factor characteristics obtained by each iteration, obtain the multi-carbon emission influence factor combined characteristics corresponding to each iteration, and finally obtain the carbon emission amount predicted value at the N1+n2 time based on the carbon emission amount prediction sub-model and the input data thereof, where the carbon emission amount prediction sub-model includes the N2-th iteration combined characteristics. Therefore, according to the embodiment of the application, not only can the carbon emission quantity be predicted by combining various carbon emission influencing factors, but also the prediction of the carbon emission influencing factors can be realized, and when the carbon emission influencing factors are predicted, the influence of the carbon emission quantity on the carbon emission influencing factors is considered, so that the accuracy of the prediction of the carbon emission influencing factors is improved, and in addition, the continuous prediction of the carbon emission influencing factors and the continuous prediction of the carbon emission quantity can be realized in an iterative mode. Of course, not all of the above-described advantages need be achieved simultaneously in practicing any one of the products or methods of the present application.
The innovative points of the embodiments of the present application include, but are not limited to, the following:
1. according to the embodiment of the application, after the carbon emission prediction model comprising N carbon emission influence factor prediction sub-models, the fully-connected neural network and the carbon emission prediction sub-models is constructed, not only can the carbon emission be predicted by combining various carbon emission influence factors, but also the prediction of the carbon emission influence factors can be realized, and when the carbon emission influence factors are predicted, the influence of the carbon emission on the carbon emission influence factors is considered, so that the accuracy of the prediction of the carbon emission influence factors is improved, and in addition, the continuous prediction of the carbon emission influence factors and the continuous prediction of the carbon emission can be realized in an iterative mode.
2. By selecting any one of the carbon emission truth value at the target moment, the carbon emission predicted value at the target moment and the carbon emission average value at m moments before the target moment as the carbon emission value corresponding to the target moment, the influence of the carbon emission on the carbon emission influencing factors can be integrated, so that the accuracy of the carbon emission influencing factor prediction is improved. Especially when the carbon emission value at the target time is taken as the priority of the carbon emission value corresponding to the target time, the carbon emission predicted value at the target time is taken as the priority of the carbon emission value corresponding to the target time, and the carbon emission average value at m times before the target time is taken as the priority of the carbon emission value corresponding to the target time, the actual carbon emission amount can be ensured as much as possible, or the carbon emission amount close to the actual carbon emission amount is added to the prediction of the carbon emission influencing factor, so that the accuracy of the prediction of the carbon emission influencing factor can be further improved.
3. In the training stage of the carbon emission prediction sub-model, the carbon emission true value at the nth 1+n2 moment after Gaussian filtering is used as input data of the carbon emission prediction sub-model to calculate a loss function, so that the influence of unpredictable parts, which are influenced by severe fluctuation factors such as policies, environmental transformation and the like, on the training of the carbon emission prediction sub-model can be reduced, and the quality of the finally trained carbon emission prediction sub-model can be improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following description will briefly introduce the drawings that are required to be used in the description of the embodiments or the prior art. It is apparent that the drawings in the following description are only some of the embodiments of the present application. Other figures may be derived from these figures without inventive effort for a person of ordinary skill in the art.
Fig. 1 is a schematic flow chart of a carbon emission prediction method according to an embodiment of the present application;
FIG. 2 is an exemplary diagram of a carbon emission prediction model structure provided in an embodiment of the present application;
fig. 3 is a block diagram showing the composition of a carbon emission prediction apparatus according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without undue burden, are within the scope of the present application.
It should be noted that the terms "comprising" and "having" and any variations thereof in the embodiments and figures herein are intended to cover a non-exclusive inclusion. A process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those listed but may alternatively include other steps or elements not listed or inherent to such process, method, article, or apparatus.
Fig. 1 is a schematic flow chart of a carbon emission prediction method provided in an embodiment of the present application, where the method may be applied to a terminal or a server, and the method is applied to a carbon emission prediction model, where the carbon emission prediction model includes N carbon emission influencing factor prediction sub-models, a fully connected neural network, and a carbon emission prediction sub-model, and different carbon emission influencing factor prediction sub-models in the N carbon emission influencing factor prediction sub-models are respectively used to predict values of different carbon emission influencing factors, and the method may include the following steps:
S110: each carbon emission influence factor predictor model is operated for n2 times in an iterative mode, and each iterative process comprises the following steps: and obtaining the characteristics of the carbon emission influence factors from the i time to the i+n1-1 time and the predicted values of the carbon emission influence factors from the i+n1 time based on the input data of each carbon emission influence factor predictive sub-model and each carbon emission influence factor predictive sub-model, wherein the input data of the carbon emission influence factor predictive sub-model comprises the original data of the carbon emission influence factors from the i time to the i+n1-1 time, and each iteration starts from i=1, i=i+1.
Carbon emission influencing factors include, but are not limited to, population count, abundance, skill level, urbanization level, industry architecture, total energy consumption, fossil fuel consumption duty, real estate, automotive industry, and the like. The carbon emission influence factor raw data at the target moment comprises carbon emission influence factor basic data at the target moment and a carbon emission value corresponding to the target moment, wherein the target moment is any moment, N, i, N1 and N2 are positive integers, N1 is the step length of the carbon emission influence factor prediction sub-model, N2 is the step length of the carbon emission amount prediction sub-model, and N1 and N2 can be the same or different.
The carbon emission value corresponding to the target moment comprises any one of a carbon emission true value at the target moment, a carbon emission predicted value at the target moment and a carbon emission average value at m moments before the target moment, wherein m is a positive integer; and/or the number of the groups of groups,
the carbon emission influencing factor basic data at the target time includes a carbon emission influencing factor true value at the target time or a carbon emission influencing factor predicted value at the target time.
In order to ensure the actual carbon emission amount as much as possible, or to approximate the actual carbon emission amount, the prediction of the carbon emission influencing factor is added, thereby further improving the accuracy of the carbon emission influencing factor prediction, the carbon emission amount true value at the target time may be regarded as the priority of the carbon emission value corresponding to the target time > the carbon emission amount predicted value at the target time is regarded as the priority of the carbon emission value corresponding to the target time > the average value of the carbon emission amounts at m times before the target time is regarded as the priority of the carbon emission value corresponding to the target time. That is, in the case where the carbon emission amount true value at the target timing can be obtained, the carbon emission amount true value at the target timing is taken as the carbon emission amount value corresponding to the target timing; under the condition that the carbon emission value at the target moment cannot be obtained, but the carbon emission predicted value at the target moment can be obtained, taking the carbon emission predicted value at the target moment as the carbon emission value corresponding to the target moment; when the actual value and the predicted value of the carbon emission amount at the target time cannot be obtained, the average value of the carbon emission amounts at m times before the target time is taken as the corresponding carbon emission amount value at the target time. The m times before the target time include m adjacent times that are nearest to and before the target time. Under the condition that the carbon emission truth values of m times before the target time can be obtained, the average value of the carbon emission values of m times before the target time is the average value of the carbon emission truth values of m times before the target time; in the case where only the carbon emission amount true value at a part of the m times before the target time and the carbon emission amount predicted value at another part of the m times before the target time can be obtained, the carbon emission amount average value at the m times before the target time is the average value of the carbon emission amount true value at the part of the time and the carbon emission amount predicted value at the another part of the time.
Under the condition that the true value of the carbon emission influencing factor at the target moment can be obtained, the carbon emission influencing factor basic data at the target moment comprises the true value of the carbon emission influencing factor at the target moment; in the case where only the predicted value of the carbon emission influencing factor at the target time can be obtained, the carbon emission influencing factor basic data at the target time includes the predicted value of the carbon emission influencing factor at the target time.
The carbon emission influencing factor base data for the target time also includes other data for the target time related to influencing the carbon emission influencing factor. For example, when the carbon emission influencing factor is the population number, the carbon emission influencing factor base data at the target time may include other data regarding the influencing population number, such as the total domestic production value at the target time, the male-female ratio, and the like, in addition to the population number true/predicted value at the target time.
It should be added that, in the training stage of the carbon emission influencing factor predictor model, the input data of the carbon emission influencing factor predictor model further comprises a true value of the carbon emission influencing factor at the i+n1 time. In the training stage of the carbon emission influence factor predictor model, the mean square error can be calculated as a loss value according to the true value and the predicted value of each carbon emission influence factor. And when the mean square error is smaller than or equal to the first error threshold, stopping training the carbon emission influence factor predictor model, and when the mean square error is larger than the first error threshold, continuing training the carbon emission influence factor predictor model. Wherein each influence factor predictor model is independently trained, each influence factor predictor model loss1The calculation method is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,mto train the number of samplesmSamples at each time),
predicted value of carbon emission influence factor at j-th moment,/->
And the true value of the carbon emission influencing factor at the j-th moment is obtained.
Wherein the prediction is performed by means of linear regression, i.epAnd (3) withy,xThe relationship of (2) can be expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,w、b、
、/>
parameters of the predictor model for carbon emission influencing factors, which can be obtained by training the model,/->
Is an equivalent function representation of the network role.
In addition, the carbon emission influencing factor predictor model is an LSTM (Long Short-Term Memory artificial neural network) model, and/or the carbon emission predictor model is an LSTM model.
Taking the application stage of the carbon emission influence factor predictor model as an example, the following describes n2 iterative processes:
when i=1, performing first iteration, respectively inputting the original data of the carbon emission influence factors from the 1 st moment to the n1 st moment into the corresponding carbon emission influence factor prediction sub-model for each carbon emission influence factor, performing operation, outputting the carbon emission influence factor characteristics from the 1 st moment to the n1 st moment at the topmost layer of the carbon emission influence factor prediction sub-model, and obtaining a carbon emission influence factor predicted value at the n1+1 th moment in a linear regression mode;
When i=2, performing a second iteration, inputting the original data of the carbon emission influence factors from the 2 nd moment to the n1+1 th moment into the corresponding carbon emission influence factor prediction sub-model for each carbon emission influence factor, performing operation, outputting the carbon emission influence factor characteristics from the 2 nd moment to the n1+1 th moment at the topmost layer of the carbon emission influence factor prediction sub-model, and obtaining the predicted value of the carbon emission influence factor from the n1+2 th moment in a linear regression mode;
and so on …
And (3) when i=n2, carrying out n < 2 > iteration, respectively inputting the original data of the carbon emission influence factors from the n < 2 > time to the n < 1 > +n2-1 time into the corresponding carbon emission influence factor prediction submodel for each carbon emission influence factor, carrying out operation, outputting the carbon emission influence factor characteristics from the n < 2 > time to the n < 1 > +n2-1 time at the topmost layer of the carbon emission influence factor prediction submodel, and obtaining the carbon emission influence factor predicted value at the n < 1 > +n2 time in a linear regression mode.
S120: and inputting the carbon emission influence factor characteristics obtained by the N carbon emission influence factor prediction submodels into a fully-connected neural network for operation aiming at the carbon emission influence factor characteristics obtained by each iteration, and obtaining multi-carbon emission influence factor combination characteristics corresponding to each iteration.
The fully-connected neural network is also called a deep neural network (Deep Neural Networks, abbreviated as DNN), which may be multi-layered, with N-dimensional input and M-dimensional output, where M is less than N, for example, n= 9,M =4.
Theory shows that the DNN network can approach any function under the condition that the parameter nodes are sufficient. In the method, the predicted effect of influencing factors on the carbon emission is learned by using a DNN network through a big data learning mode.
The final output of the DNN network may be represented as a nested function
The activation function in the DNN network may employ a sigmoid function or the like.
When i=1, the N carbon emission influence factor characteristics at the 1 st to N1 st moments obtained by the N carbon emission influence factor predictor models can be input into a DNN network to obtain multi-carbon emission influence factor combination characteristics at the 1 st to N1 st moments;
when i=2, the N carbon emission influence factor characteristics at the 2 nd to n1+1 th moments obtained by the N carbon emission influence factor predictor models can be input into a DNN network to obtain multi-carbon emission influence factor combination characteristics at the 2 nd to n1+1 th moments;
and so on …
When i=n2, the N carbon emission influence factor characteristics at the N2 nd to n1+n2-1 times obtained by the N carbon emission influence factor predictor models can be input into the DNN network to obtain multi-carbon emission influence factor combination characteristics at the N2 nd to n1+n2-1 times.
S130: and obtaining a carbon emission predicted value at the time of n < 1+ > n < 2 >, based on the carbon emission predicted sub-model and the input data of the carbon emission predicted sub-model, wherein the input data of the carbon emission predicted sub-model comprises a sequence formed by n < 2 > multi-carbon emission influence factor combination characteristics obtained by n < 2 > iterations.
The carbon emission prediction value at the time of n < 1 > +n2-1 is obtained by inputting the input data of the carbon emission prediction sub-model into the carbon emission prediction sub-model for carbon emission prediction, namely, the carbon emission prediction is carried out by inputting the carbon emission prediction sub-model into a sequence formed by n < 2 > multi-carbon emission influence factor combination characteristics obtained through n < 2 > iterations, so that the carbon emission prediction value at the time of n < 1 > +n2-1 is obtained.
Taking DNN output 4 node as an example, the input data at a certain moment of the carbon emission prediction sub-model comprises
Z=[z1,z2,z3,z4]Wherein, the method comprises the steps of, wherein,
，/>
，/>
，
y is a combination of multiple carbon emission influencing factors, i.e. +.>
，/>
To->
The outputs of the carbon emission influencing factor predictor models 1 through N, respectively. To sum up, the input data of the carbon emission predictor model includes a sequence of n2 multi-carbon emission influencing factor combination features [Z1,Z2,...,Zn2]。
In the training stage of the carbon emission prediction sub-model, the input data of the carbon emission prediction sub-model further comprises a carbon emission true value at the time of n1+n2, or a carbon emission true value after Gaussian filtering is carried out on the carbon emission true value at the time of n 1+n2.
Because the carbon emission is affected by severe fluctuation factors such as policies, environmental changes, etc., the carbon emission contains unpredictable portions, and it is not reasonable to predict the unpredictable portions with the carbon emission predictor model. In order to reduce the influence of unpredictable parts, the actual carbon emission is subjected to Gaussian filtering treatment, noise is removed, and the carbon emission after denoising is used for prediction. Thus, during the training phase of the carbon emission prediction sub-model, the input data of the carbon emission prediction sub-model may further include a carbon emission truth value after Gaussian filtering of the carbon emission truth value at time n1+n2.
The implementation of the method can carry out joint training on the DNN network and the carbon emission predictor model, and can calculate the mean square error as a loss value according to the carbon emission true value and the carbon emission predicted value after Gaussian filtering in the training process. And stopping training the carbon emission quantity predictor model when the mean square error is smaller than or equal to the second error threshold value, and continuing training the carbon emission quantity predictor model when the mean square error is larger than the second error threshold value. The loss function loss2 of the carbon emission prediction can be expressed as
Wherein, the liquid crystal display device comprises a liquid crystal display device,
is the truth value of carbon emission after Gaussian filtration, +. >
The predicted value of the carbon emission amount is a predicted sub-model of the carbon emission amount, and k is the number of samples. Record->
For the original carbon emission value, there is +.>
Wherein->
Is a discrete Gaussian filter,>
the dimension of the first order gaussian filter may be, for example, an odd number of 3, 5, 7, etc.
The predicted value of the carbon emission at the time n1+n2 outputted by the carbon emission prediction sub-model is used as one of the input data required in the next set of n2 iterations by the carbon emission influence factor prediction sub-model, the next set of n2 iterations is performed by adding other input data, and the iterative result is passed through the fully connected neural network and the carbon emission prediction sub-model to output the predicted value of the carbon emission at the time n1+n2+1. That is, as a whole, the carbon emission prediction sub-model is also iterated, and by iterating constantly, the predicted values of the carbon emission at a plurality of times in the future can be predicted.
In addition, the "time" mentioned in the embodiment of the present application may be understood as a certain time point, and may also be understood as a time period, for example, a time is one month, then time 1 is 2023 years 1 month 1 day to 1 month 31 day, time 2 is 2023 years 2 months 1 day to 2 months 28 day, and when the time represents the time period, the data corresponding to a certain time point may be understood as a mean value of the data corresponding to the time point, for example, the carbon emission influence factor base data of the target time point includes a mean value of the carbon emission influence factor base data in the target time point.
As shown in fig. 2, the following describes the above-described carbon emission prediction method, taking n=9, where the carbon emission influencing factor predictor model and the carbon emission predictor model are both LSTM models, and the step sizes of both models are 5 as an example:
(1) LSTM1-9 represent carbon emission influencing factor predictor models corresponding to different carbon emission influencing factors, respectively, and each model in LSTM1-9 is iterated for 5 times respectively:
first iteration: inputting the initial data of the carbon emission influencing factors at the 1 st moment to the 5 th moment into the corresponding model in the LSTM1-9 to obtain the characteristics of the carbon emission influencing factors at the 1 st moment to the 5 th moment and the predicted value of the carbon emission influencing factors at the 6 th moment;
second iteration: inputting the original data of the carbon emission influencing factors at the 2 nd-6 th moments into corresponding models in LSTM1-9 to obtain the characteristics of the carbon emission influencing factors at the 2 nd-6 th moments and the predicted values of the carbon emission influencing factors at the 7 th moments;
third iteration: inputting the initial data of the carbon emission influencing factors at the 3 rd moment to the 7 th moment into the corresponding model in the LSTM1-9 to obtain the characteristics of the carbon emission influencing factors at the 3 rd moment to the 7 th moment and the predicted value of the carbon emission influencing factors at the 8 th moment;
fourth iteration: inputting the original data of the carbon emission influencing factors at the 4 th-8 th moments into corresponding models in LSTM1-9 to obtain the characteristics of the carbon emission influencing factors at the 4 th-8 th moments and the predicted values of the carbon emission influencing factors at the 9 th moments;
Fifth iteration: and (3) inputting the initial data of the carbon emission influencing factors at the 5 th to 9 th moments into corresponding models in the LSTM1-9, and obtaining the characteristics of the carbon emission influencing factors at the 5 th to 9 th moments and the predicted values of the carbon emission influencing factors at the 10 th moments.
(2) Inputting the carbon emission influence factor characteristics at the 1 st to 5 th moments obtained by LSTM1-9 into DNN to obtain multi-carbon emission influence factor combination characteristics corresponding to the first iteration (which can be regarded as input data at the 1 st moment of the carbon emission quantity prediction sub-model LSTM 10); inputting the carbon emission influence factor characteristics at the 2 nd-6 th moments obtained by LSTM1-9 into DNN to obtain multi-carbon emission influence factor combination characteristics corresponding to the second iteration (which can be regarded as input data at the 2 nd moment of LSTM 10); inputting the carbon emission influence factor characteristics at the 3 rd to 7 th moments obtained by LSTM1-9 into DNN to obtain multi-carbon emission influence factor combination characteristics corresponding to the third iteration (which can be regarded as input data at the 3 rd moment of LSTM 10); inputting the carbon emission influence factor characteristics at the 4 th to 8 th moments obtained by LSTM1-9 into DNN to obtain multi-carbon emission influence factor combination characteristics corresponding to the fourth iteration (which can be regarded as input data at the 4 th moment of LSTM 10); inputting the carbon emission influence factor characteristics at the 5 th to 9 th moments obtained by LSTM1-9 into DNN to obtain multi-carbon emission influence factor combination characteristics corresponding to the fifth iteration (which can be regarded as input data at the 5 th moment of LSTM 10).
The DNN has a 3-layer structure and comprises an input layer, a hidden layer and an output layer, wherein the input layer is 9 nodes, and the output layer is 4 nodes.
(3) Inputting the sequence formed by the combination characteristics of the 5 multi-carbon emission influencing factors obtained in the step (2) into LSTM10, and outputting the predicted value of the carbon emission at the 10 th moment.
The output result of LSTM10 may be used as the input result of LSTM1-9 for the next set of iterative operations.
According to the carbon emission prediction method provided by the embodiment of the invention, the carbon emission influence factor characteristics of the ith moment to the (i+n1) -1 moment and the carbon emission influence factor predicted value of the (i+n1) -1 moment can be obtained firstly based on each carbon emission influence factor prediction sub-model in the N carbon emission influence factor prediction sub-models and the input data thereof, wherein the input data of the carbon emission influence factor prediction sub-models comprise the carbon emission influence factor original data of the (i) moment to the (i+n1) -1 moment, each carbon emission influence factor prediction sub-model is iterated N2 times according to the method, then the carbon emission influence factor characteristics obtained by the N carbon emission influence factor prediction sub-models are input into the fully connected neural network for operation according to the carbon emission influence factor characteristics obtained by each iteration, the multi-carbon emission influence factor combined characteristics corresponding to each iteration are obtained, and finally the carbon emission influence factor predicted value of the (N1+n2) moment is obtained based on the carbon emission influence factor prediction sub-models and the input data thereof, wherein the input data of the carbon emission quantity prediction sub-models comprise N2 carbon emission influence factor combined factors of N. Therefore, according to the embodiment of the application, not only can the carbon emission quantity be predicted by combining various carbon emission influencing factors, but also the prediction of the carbon emission influencing factors can be realized, and when the carbon emission influencing factors are predicted, the influence of the carbon emission quantity on the carbon emission influencing factors is considered, so that the accuracy of the prediction of the carbon emission influencing factors is improved, and in addition, the continuous prediction of the carbon emission influencing factors and the continuous prediction of the carbon emission quantity can be realized in an iterative mode.
Corresponding to the above method embodiment, an embodiment of the present application provides a carbon emission prediction apparatus, as shown in fig. 3, where the apparatus is applied to a carbon emission prediction model, the carbon emission prediction model includes N carbon emission influencing factor prediction sub-models, a fully connected neural network, and a carbon emission prediction sub-model, and different carbon emission influencing factor prediction sub-models in the N carbon emission influencing factor prediction sub-models are respectively used to predict values of different carbon emission influencing factors, where the apparatus includes:
a carbon emission influencing factor processing unit 210, configured to iteratively run each of the carbon emission influencing factor predictor models n2 times, where each iterative process includes: obtaining a carbon emission influence factor characteristic from an i-th moment to an i+n1-1-th moment and a carbon emission influence factor predicted value from the i+n1-th moment based on input data of each carbon emission influence factor prediction sub-model and the carbon emission influence factor prediction sub-model, wherein the input data of the carbon emission influence factor prediction sub-model comprises carbon emission influence factor raw data from the i-th moment to the i+n1-1-th moment, the carbon emission influence factor raw data at a target moment comprises carbon emission influence factor basic data at the target moment and a carbon emission measurement value corresponding to the target moment, the target moment is any moment, i=i+1 from i=1 each time, and the N, the i, the N1 and the N2 are positive integers;
A fully connected processing unit 220, configured to input the carbon emission influencing factor characteristics obtained by the N carbon emission influencing factor predictor models into the fully connected neural network for operation with respect to the carbon emission influencing factor characteristics obtained by each iteration, so as to obtain multi-carbon emission influencing factor combination characteristics corresponding to each iteration;
a carbon emission amount prediction unit 230, configured to obtain a carbon emission amount predicted value at the time n1+n2 based on the carbon emission amount predicted sub-model and input data of the carbon emission amount predicted sub-model, where the input data of the carbon emission amount predicted sub-model includes a sequence of n2 multi-carbon emission influence factor combination features obtained by n2 iterations.
In one embodiment, the carbon emission value corresponding to the target time includes any one of a carbon emission true value at the target time, a carbon emission predicted value at the target time, and a carbon emission average value at m times before the target time, where m is a positive integer; and/or the number of the groups of groups,
the carbon emission influencing factor basic data of the target moment comprises a carbon emission influencing factor true value of the target moment or a carbon emission influencing factor predicted value of the target moment.
In one embodiment, the carbon emission value at the target time is regarded as the priority of the carbon emission value corresponding to the target time, the carbon emission predicted value at the target time is regarded as the priority of the carbon emission value corresponding to the target time, and the carbon emission average value at m times before the target time is regarded as the priority of the carbon emission value corresponding to the target time.
In one embodiment, during the training phase of the carbon emission influencing factor predictor model, the input data of the carbon emission influencing factor predictor model further comprises a carbon emission influencing factor true value at the i+n1 time; and/or the number of the groups of groups,
in the training stage of the carbon emission quantity prediction sub-model, the input data of the carbon emission quantity prediction sub-model further comprises a carbon emission quantity true value at the n < 1 > +n2 time, or a carbon emission quantity true value after Gaussian filtering is carried out on the carbon emission quantity true value at the n < 1 > +n2 time.
In one embodiment, the carbon emission influencing factor predictor model is a long-short term memory artificial neural network LSTM model and/or the carbon emission predictor model is a LSTM model.
According to the carbon emission quantity prediction device provided by the embodiment of the invention, based on each carbon emission influence factor prediction sub-model in N carbon emission influence factor prediction sub-models and input data thereof, the carbon emission influence factor characteristics from the ith moment to the (i+n1) -1 moment and the carbon emission influence factor predicted value from the (i+n1) -1 moment are obtained, wherein the input data of the carbon emission influence factor prediction sub-model comprises the carbon emission influence factor original data from the (i) moment to the (i+n1) -1 moment, each carbon emission influence factor prediction sub-model is iterated N2 times according to the method, then the carbon emission influence factor characteristics obtained by the N carbon emission influence factor prediction sub-models are input into a fully connected neural network for operation according to the carbon emission influence factor characteristics obtained by each iteration, the multi-carbon emission influence factor combined characteristics corresponding to each iteration are obtained, and finally the carbon emission influence factor predicted value at the (N1+n2) -2 moment is obtained based on the carbon emission quantity prediction sub-model and the input data thereof, wherein the input data of the carbon emission quantity prediction sub-model comprises N2 carbon emission influence factor combined factors obtained by N2 iterations. Therefore, according to the embodiment of the application, not only can the carbon emission quantity be predicted by combining various carbon emission influencing factors, but also the prediction of the carbon emission influencing factors can be realized, and when the carbon emission influencing factors are predicted, the influence of the carbon emission quantity on the carbon emission influencing factors is considered, so that the accuracy of the prediction of the carbon emission influencing factors is improved, and in addition, the continuous prediction of the carbon emission influencing factors and the continuous prediction of the carbon emission quantity can be realized in an iterative mode.
Based on the above method embodiments, another embodiment of the present application provides a storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to implement the method as described above.
Based on the above method embodiments, another embodiment of the present application provides an electronic device, including:
one or more processors;
storage means for storing one or more programs,
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the methods as described above.
The device embodiment corresponds to the method embodiment, and has the same technical effects as the method embodiment, and the specific description refers to the method embodiment. The apparatus embodiments are based on the method embodiments, and specific descriptions may be referred to in the method embodiment section, which is not repeated herein. Those of ordinary skill in the art will appreciate that: the figures are schematic representations of one embodiment only and the modules or flows in the figures are not necessarily required to practice the present application.
Those of ordinary skill in the art will appreciate that: the modules in the apparatus of the embodiments may be distributed in the apparatus of the embodiments according to the description of the embodiments, or may be located in one or more apparatuses different from the present embodiments with corresponding changes. The modules of the above embodiments may be combined into one module, or may be further split into a plurality of sub-modules.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application, and are not limiting thereof; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (10)
1. The method is applied to a carbon emission prediction model, the carbon emission prediction model comprises N carbon emission influence factor prediction sub-models, a fully-connected neural network and a carbon emission prediction sub-model, and different carbon emission influence factor prediction sub-models in the N carbon emission influence factor prediction sub-models are respectively used for predicting different carbon emission influence factor values, and the method comprises the following steps:
and respectively carrying out iterative operation on each carbon emission influence factor predictor model n2 times, wherein each iterative process comprises the following steps: obtaining a carbon emission influence factor characteristic from an ith moment to an (i+n 1) -1 moment and a carbon emission influence factor predicted value from the (i+n 1) -1 moment based on input data of each carbon emission influence factor predicted sub-model and the carbon emission influence factor predicted sub-model, wherein the input data of the carbon emission influence factor predicted sub-model comprises carbon emission influence factor raw data from the ith moment to the (i+n 1) -1 moment, the carbon emission influence factor raw data at a target moment comprises carbon emission influence factor basic data at the target moment and a carbon emission measurement value corresponding to the target moment, the target moment is any moment, i=i+1 is iterated each time, the N, the i, the N1 and the N2 are positive integers, the N1 is a step length of the carbon emission influence factor predicted sub-model, and the N2 is a step length of the carbon emission amount predicted sub-model;
Inputting the carbon emission influence factor characteristics obtained by N carbon emission influence factor predictor models into the fully-connected neural network for operation aiming at the carbon emission influence factor characteristics obtained by each iteration to obtain multi-carbon emission influence factor combination characteristics corresponding to each iteration;
and obtaining a carbon emission predicted value at the time of n < 1+ > n < 2 >, based on the carbon emission predicted sub-model and the input data of the carbon emission predicted sub-model, wherein the input data of the carbon emission predicted sub-model comprises a sequence formed by n < 2 > multi-carbon emission influence factor combination characteristics obtained by n < 2 > iterations.
2. The method according to claim 1, wherein the carbon emission value corresponding to the target time includes any one of a carbon emission true value at the target time, a carbon emission predicted value at the target time, and a carbon emission average value at m times before the target time, the m being a positive integer; and/or the number of the groups of groups,
the carbon emission influencing factor basic data of the target moment comprises a carbon emission influencing factor true value of the target moment or a carbon emission influencing factor predicted value of the target moment.
3. The method according to claim 2, characterized in that the priority of the carbon emission amount true value at the target time as the carbon emission amount value corresponding to the target time > the priority of the carbon emission amount predicted value at the target time as the carbon emission amount value corresponding to the target time > the priority of the carbon emission amount average value at m times before the target time as the carbon emission amount value corresponding to the target time.
4. The method of claim 1, wherein during a training phase of the carbon emission influencing factor predictor model, the input data of the carbon emission influencing factor predictor model further comprises a carbon emission influencing factor true value at time i+n1; and/or the number of the groups of groups,
in the training stage of the carbon emission quantity prediction sub-model, the input data of the carbon emission quantity prediction sub-model further comprises a carbon emission quantity true value at the n < 1 > +n2 time, or a carbon emission quantity true value after Gaussian filtering is carried out on the carbon emission quantity true value at the n < 1 > +n2 time.
5. The method of any one of claims 1-4, wherein the carbon emission influencing factor predictor model is a long-term memory artificial neural network LSTM model and/or the carbon emission predictor model is a LSTM model.
6. A carbon emission prediction apparatus, wherein the apparatus is applied to a carbon emission prediction model, the carbon emission prediction model includes N carbon emission influencing factor prediction sub-models, a fully connected neural network, and a carbon emission prediction sub-model, different carbon emission influencing factor prediction sub-models in the N carbon emission influencing factor prediction sub-models are respectively used for predicting different carbon emission influencing factor values, the apparatus includes:
The carbon emission influence factor processing unit is used for respectively and iteratively operating each carbon emission influence factor prediction sub-model n2 times, and each iteration process comprises the following steps: obtaining a carbon emission influence factor characteristic from an ith moment to an (i+n 1) -1 moment and a carbon emission influence factor predicted value from the (i+n 1) -1 moment based on input data of each carbon emission influence factor predicted sub-model and the carbon emission influence factor predicted sub-model, wherein the input data of the carbon emission influence factor predicted sub-model comprises carbon emission influence factor raw data from the ith moment to the (i+n 1) -1 moment, the carbon emission influence factor raw data at a target moment comprises carbon emission influence factor basic data at the target moment and a carbon emission measurement value corresponding to the target moment, the target moment is any moment, i=i+1 is iterated each time, the N, the i, the N1 and the N2 are positive integers, the N1 is a step length of the carbon emission influence factor predicted sub-model, and the N2 is a step length of the carbon emission amount predicted sub-model;
the fully-connected processing unit is used for inputting the carbon emission influence factor characteristics obtained by the N carbon emission influence factor predictor models into the fully-connected neural network for operation aiming at the carbon emission influence factor characteristics obtained by each iteration to obtain multi-carbon emission influence factor combination characteristics corresponding to each iteration;
And the carbon emission prediction unit is used for obtaining a carbon emission prediction value at the time of n < 1+ > n < 2 >, based on the carbon emission prediction sub-model and the input data of the carbon emission prediction sub-model, wherein the input data of the carbon emission prediction sub-model comprises a sequence formed by n < 2 > multi-carbon emission influence factor combination characteristics obtained by n < 2 > iterations.
7. The apparatus according to claim 6, wherein the carbon emission amount value corresponding to the target time includes any one of a carbon emission amount true value at the target time, a carbon emission amount predicted value at the target time, and a carbon emission amount average value at m times before the target time, the m being a positive integer; and/or the number of the groups of groups,
the carbon emission influencing factor basic data of the target moment comprises a carbon emission influencing factor true value of the target moment or a carbon emission influencing factor predicted value of the target moment.
8. The apparatus according to any one of claims 6 to 7, wherein during a training phase of the carbon emission influencing factor predictor model, the input data of the carbon emission influencing factor predictor model further comprises a carbon emission influencing factor true value at time i+n1; and/or the number of the groups of groups,
In the training stage of the carbon emission quantity prediction sub-model, the input data of the carbon emission quantity prediction sub-model further comprises a carbon emission quantity true value at the n < 1 > +n2 time, or a carbon emission quantity true value after Gaussian filtering is carried out on the carbon emission quantity true value at the n < 1 > +n2 time.
9. A storage medium having stored thereon executable instructions which, when executed by a processor, cause the processor to implement the method of any of claims 1-5.
10. An electronic device, comprising:
one or more processors;
storage means for storing one or more programs,
wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310315526.8A CN116029466B (en) | 2023-03-29 | 2023-03-29 | Carbon emission prediction method, device, storage medium and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310315526.8A CN116029466B (en) | 2023-03-29 | 2023-03-29 | Carbon emission prediction method, device, storage medium and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116029466A CN116029466A (en) | 2023-04-28 |
| CN116029466B true CN116029466B (en) | 2023-05-23 |
Family
ID=86077874
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310315526.8A Active CN116029466B (en) | 2023-03-29 | 2023-03-29 | Carbon emission prediction method, device, storage medium and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116029466B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114707120A (en) * | 2022-04-25 | 2022-07-05 | 昆明理工大学 | Statistical modeling method for measuring and calculating carbon emission scale |
| CN114778767A (en) * | 2022-04-06 | 2022-07-22 | 中国电力科学研究院有限公司 | Method, device, equipment and medium for continuously measuring direct carbon emission in carbon emission park |
| CN115293418A (en) * | 2022-07-29 | 2022-11-04 | 暨南大学 | Regional carbon emission prediction method, system and medium based on LMDI model and GRU model |
| CN115343958A (en) * | 2022-09-16 | 2022-11-15 | 中邮科通信技术股份有限公司 | Carbon analysis method based on intelligent building VMD-LSTM-GRU |
| CN115759346A (en) * | 2022-10-21 | 2023-03-07 | 中山大学 | AO algorithm based carbon emission prediction method, device and equipment |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20220358515A1 (en) * | 2021-05-07 | 2022-11-10 | The Boston Consulting Group, Inc. | Carbon emissions management system |
-
2023
- 2023-03-29 CN CN202310315526.8A patent/CN116029466B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114778767A (en) * | 2022-04-06 | 2022-07-22 | 中国电力科学研究院有限公司 | Method, device, equipment and medium for continuously measuring direct carbon emission in carbon emission park |
| CN114707120A (en) * | 2022-04-25 | 2022-07-05 | 昆明理工大学 | Statistical modeling method for measuring and calculating carbon emission scale |
| CN115293418A (en) * | 2022-07-29 | 2022-11-04 | 暨南大学 | Regional carbon emission prediction method, system and medium based on LMDI model and GRU model |
| CN115343958A (en) * | 2022-09-16 | 2022-11-15 | 中邮科通信技术股份有限公司 | Carbon analysis method based on intelligent building VMD-LSTM-GRU |
| CN115759346A (en) * | 2022-10-21 | 2023-03-07 | 中山大学 | AO algorithm based carbon emission prediction method, device and equipment |
Non-Patent Citations (2)
| Title |
|---|
| Prediction on Shanghai’s Energy Consumption Trend and Carbon Emission Peak;Rui Huang等;2011 Third International Conference on Measuring Technology and Mechatronics Automation;第898-901页 * |
| 基于CNN-LSTM模型的中国碳排放量实时预测研究;张学清等;中阿科技论坛(中英文);第71-75页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116029466A (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112364975B (en) | Terminal running state prediction method and system based on graph neural network | |
| JP2006522376A (en) | System and method for automatic experience rate setting and / or loss accumulation | |
| CN111260124A (en) | Chaos time sequence prediction method based on attention mechanism deep learning | |
| Gao et al. | On the asymptotic normality of estimating the affine preferential attachment network models with random initial degrees | |
| CN112578089B (en) | Air pollutant concentration prediction method based on improved TCN | |
| CN115983485A (en) | Prediction method and system for carbon emission of power plant | |
| CN116451093A (en) | Training method of circuit fault analysis model and circuit fault analysis method | |
| CN111353534A (en) | Graph data category prediction method based on adaptive fractional order gradient | |
| CN116029466B (en) | Carbon emission prediction method, device, storage medium and equipment | |
| CN113380340A (en) | Training method and device of fly ash concentration prediction model and computer equipment | |
| CN113435595A (en) | Two-stage optimization method for extreme learning machine network parameters based on natural evolution strategy | |
| CN112232565A (en) | Two-stage time sequence prediction method, prediction system, terminal and medium | |
| CN115860232A (en) | Steam load prediction method, system, electronic device and medium | |
| CN114066081B (en) | Enterprise risk prediction method and device based on graph attention network and electronic equipment | |
| CN113919871A (en) | Carbon transaction price prediction method based on improved LSTM model | |
| CN114694379B (en) | Traffic flow prediction method and system based on self-adaptive dynamic graph convolution | |
| CN116303786B (en) | Block chain financial big data management system based on multidimensional data fusion algorithm | |
| CN115793472B (en) | Modeling method, modeling system, control method and control system of heavy-duty train | |
| CN114267422B (en) | Method and system for predicting surface water quality parameters, computer equipment and storage medium | |
| CN115619563A (en) | Stock price analysis method based on neural network | |
| CN117575007B (en) | Large model knowledge completion method and system based on post-decoding credibility enhancement | |
| US20240028016A1 (en) | Real-time prediction and regulation methods and systems of product quality based on process dynamic pattern | |
| CN115587629B (en) | Covariance expansion coefficient estimation method, model training method and storage medium terminal | |
| CN116757716A (en) | Electricity price prediction method and device based on improved time-varying depth feedforward neural network | |
| CN117526294A (en) | Short-term building load prediction method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |