CN116485036A - Multi-energy-flow carbon emission short-term prediction method based on multi-task learning mechanism - Google Patents
Multi-energy-flow carbon emission short-term prediction method based on multi-task learning mechanism Download PDFInfo
- Publication number
- CN116485036A CN116485036A CN202310570606.8A CN202310570606A CN116485036A CN 116485036 A CN116485036 A CN 116485036A CN 202310570606 A CN202310570606 A CN 202310570606A CN 116485036 A CN116485036 A CN 116485036A
- Authority
- CN
- China
- Prior art keywords
- data
- predicted
- task learning
- learning model
- energy
- 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
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 title claims abstract description 68
- 229910052799 carbon Inorganic materials 0.000 title claims abstract description 68
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000007246 mechanism Effects 0.000 title claims abstract description 24
- 230000006870 function Effects 0.000 claims abstract description 36
- 238000012360 testing method Methods 0.000 claims abstract description 29
- 238000000605 extraction Methods 0.000 claims abstract description 20
- 238000013527 convolutional neural network Methods 0.000 claims description 7
- 238000002790 cross-validation Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 2
- 230000008878 coupling Effects 0.000 abstract description 7
- 238000010168 coupling process Methods 0.000 abstract description 7
- 238000005859 coupling reaction Methods 0.000 abstract description 7
- 238000013135 deep learning Methods 0.000 abstract description 5
- 230000006872 improvement Effects 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 7
- 230000009471 action Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004146 energy storage Methods 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Engineering & Computer Science (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Development Economics (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Operations Research (AREA)
- Primary Health Care (AREA)
- Educational Administration (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Game Theory and Decision Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a multi-energy flow carbon emission short-term prediction method based on a multi-task learning mechanism, which relates to the technical field of multi-energy flow integrated management and comprises the following steps: acquiring historical load data, and performing feature extraction operation on the historical load data to obtain historical data to be predicted; the historical data to be predicted is sent to a multi-task learning model for training, the trained multi-task learning model is tested, and the predicted data output after the test is obtained; constructing an overall loss function according to the predicted data, and judging whether the trained multi-task learning model converges or not based on the overall loss function; and if the trained multi-task learning model converges, predicting the carbon emission amount of the predicted data to obtain the predicted multi-energy carbon emission amount. The method and the device can realize improvement of a multi-task learning mechanism in deep learning so as to fully mine implicit coupling relations among various energy sources, so that accurate prediction is carried out on short-term carbon emission conditions of the multi-energy flow network, and accuracy of multi-energy flow carbon emission prediction is improved.
Description
Technical Field
The invention relates to the technical field of comprehensive management of multi-energy flows, in particular to a multi-energy flow carbon emission short-term prediction method based on a multi-task learning mechanism.
Background
In recent years, along with clean energy equipment such as photovoltaic, wind power and the like in China and various distributed energy storage and massive grid connection, the energy network of the traditional city is more and more complex, and an effective energy regulation strategy is urgently needed. The accurate prediction of the multi-type energy demand can obtain a more reasonable, scientific and effective multi-energy flow interactive planning scheme. Most of the existing prediction schemes track and predict the original electric, cold and heat load demands, the attention to the structure proportion of clean energy and the overall carbon emission condition is not high enough, the utilization of the urban area multiple energy sources is unreasonable, the investment and construction pressure of the power grid are large, and the annual cost of multi-energy planning is increased.
From the above, how to realize the improvement of the multi-task learning mechanism in deep learning to fully mine the implicit coupling relation among various energy sources, so as to accurately predict the short-term carbon emission condition of the multi-energy flow network, and improve the accuracy of the multi-energy flow carbon emission prediction is a problem to be solved in the field.
Disclosure of Invention
Therefore, the invention aims to provide a multi-energy flow carbon emission short-term prediction method based on a multi-task learning mechanism, which can realize improvement of the multi-task learning mechanism in deep learning so as to fully mine implicit coupling relations among various energy sources, thereby accurately predicting the short-term carbon emission condition of a multi-energy flow network and improving the accuracy of multi-energy flow carbon emission prediction. The specific scheme is as follows:
the application discloses a multi-energy flow carbon emission short-term prediction method based on a multi-task learning mechanism, which comprises the following steps:
acquiring multi-energy historical load data, determining duration information to be predicted, and performing feature extraction operation on the historical load data according to the duration information to be predicted to obtain historical data to be predicted;
transmitting the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain predicted data output after testing;
constructing an overall loss function according to the prediction data, and judging whether the trained multi-task learning model converges or not based on the overall loss function;
and if the trained multi-task learning model converges, predicting the carbon emission amount of the predicted data to obtain the predicted multi-energy carbon emission amount.
Optionally, the sending the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain the predicted data output after testing, including:
and sending the historical data to be predicted to a preset multi-task learning model for training, so that the trained multi-task learning model tests the historical data to be predicted by utilizing a self shallow convolutional neural network to obtain the predicted data output after the test.
Optionally, the determining the duration information to be predicted, performing feature extraction operation on the historical load data according to the duration information to be predicted to obtain the historical data to be predicted, including:
determining duration information to be predicted and a data feature extraction range according to service requirements, screening historical load data to be extracted from the historical load data according to the duration information to be predicted and the data feature extraction range, and performing feature extraction operation on the historical load data to be extracted to obtain historical data to be predicted;
and constructing a multi-energy historical data sample base according to the historical data to be predicted.
Optionally, the sending the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain the predicted data output after testing, including:
the historical data to be predicted is sent to a preset multi-task learning model to be trained, so that the trained multi-task learning model adopts a cross-validation training method to determine data sets from the multi-energy historical data sample library, the number of the data sets is determined, then a predicted data set is screened out from the data sets, other data sets except the predicted data set are used as training data sets, the training data sets are used for training the multi-task learning model to obtain the trained multi-task learning model, the trained multi-task learning model is tested by the predicted data sets to obtain predicted sub-data, then the step of determining the predicted data sets from the data sets is skipped until the number of the obtained predicted sub-data sets is the same as the number of the data sets, the predicted sub-data is obtained, and the predicted data is determined based on the predicted sub-data.
Optionally, the determining the prediction data based on each of the predictor data includes:
acquiring all the predicted sub-data, calculating the average value of all the predicted sub-data, and taking the average value as the predicted data;
and outputting the predicted data by using the trained neural network in the multi-task learning model to obtain the predicted data.
Optionally, after the determining whether the trained multi-task learning model converges based on the overall loss function, the method further includes:
if the trained multi-task learning model is not converged, the step of transmitting the historical data to be predicted to a preset multi-task learning model for training is skipped to obtain a new overall loss function.
Optionally, the constructing an overall loss function according to the prediction data includes:
acquiring multi-energy load actual data, and calculating likelihood probability between the multi-energy load actual data and the predicted data;
an overall loss function is constructed based on the likelihood probabilities.
As can be seen, the application provides a multi-energy flow carbon emission short-term prediction method based on a multi-task learning mechanism, which comprises the steps of obtaining historical load data of multiple energy sources, determining duration information to be predicted, and performing feature extraction operation on the historical load data according to the duration information to be predicted to obtain historical data to be predicted; transmitting the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain predicted data output after testing; constructing an overall loss function according to the prediction data, and judging whether the trained multi-task learning model converges or not based on the overall loss function; and if the trained multi-task learning model converges, predicting the carbon emission amount of the predicted data to obtain the predicted multi-energy carbon emission amount. The method and the device realize accurate prediction of different loads of the urban area with multiple energy flows by constructing the multi-task learning model, calculate the total carbon emission amount of the urban area through energy regulation and control, and finally realize short-term prediction of the carbon emission of the urban area. The multi-task learning can effectively reduce the prediction cost of multi-type energy sources, and can mine the high-dimensional coupling characteristic among different tasks so as to improve the prediction precision. The accurate prediction of the carbon emission can promote the regulation trend of the subsequent energy planning and improve the regulation potential of the energy-saving scheme.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present invention, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a method for short-term prediction of carbon emissions in a multi-energy stream based on a multi-task learning mechanism as disclosed herein;
FIG. 2 is a flowchart of a method for short-term prediction of carbon emissions in a multi-energy stream based on a multi-task learning mechanism as disclosed herein;
FIG. 3 is a diagram of a multi-task learning network according to the present disclosure;
fig. 4 is a specific flowchart of a method for short-term prediction of carbon emissions in a multi-energy stream based on a multi-task learning mechanism disclosed in the present application.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In recent years, along with clean energy equipment such as photovoltaic, wind power and the like in China and various distributed energy storage and massive grid connection, the energy network of the traditional city is more and more complex, and an effective energy regulation strategy is urgently needed. The accurate prediction of the multi-type energy demand can obtain a more reasonable, scientific and effective multi-energy flow interactive planning scheme. Most of the existing prediction schemes track and predict the original electric, cold and heat load demands, the attention to the structure proportion of clean energy and the overall carbon emission condition is not high enough, the utilization of the urban area multiple energy sources is unreasonable, the investment and construction pressure of the power grid are large, and the annual cost of multi-energy planning is increased. From the above, how to realize the improvement of the multi-task learning mechanism in deep learning to fully mine the implicit coupling relation among various energy sources, so as to accurately predict the short-term carbon emission condition of the multi-energy flow network, and improve the accuracy of the multi-energy flow carbon emission prediction is a problem to be solved in the field.
Referring to fig. 1, the embodiment of the invention discloses a multi-energy flow carbon emission short-term prediction method based on a multi-task learning mechanism, which specifically comprises the following steps:
step S11: and acquiring multi-energy historical load data, determining duration information to be predicted, and performing feature extraction operation on the historical load data according to the duration information to be predicted to obtain the historical data to be predicted.
In the embodiment, historical load data of multiple energy sources are obtained, then time length information to be predicted and a data feature extraction range are determined according to service requirements, historical load data to be extracted are screened out from the historical load data according to the time length information to be predicted and the data feature extraction range, and feature extraction operation is carried out on the historical load data to be extracted to obtain the historical data to be predicted; and constructing a multi-energy historical data sample base according to the historical data to be predicted. Specifically, all historical electric, cold and heat loads, clean energy output and meteorological data (namely multi-energy historical load data) of the urban area to be predicted are obtained, and the duration to be predicted is set to be n pt (n pt =1, 2,..12) hours. In order to construct a multi-energy historical data sample library, the time length to be predicted is 24 hours before and n hours before hd (n hd =1, 2,..12) electric load, heat load, cold load, clean energy load for the corresponding period of day, and date type of day of prediction (whether working day), weather, n before prediction ht (n ht =1, 2..12) temperature, humidity, wind speed, atmospheric pressure, time information of a period to be predicted as input sample features of a subsequent recognition model, and dividing urban area history data hour by hour based on a set input feature format to obtain a multi-energy history data sample library of the urban area.
Step S12: and sending the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain the predicted data output after testing.
In this embodiment, the historical data to be predicted is sent to a preset multi-task learning model to perform training, so that the trained multi-task learning model adopts a cross-validation training method to determine each data set from the multi-energy historical data sample library, determine the number of each data set, then screen out predicted data sets from each data set, use other data sets except for the predicted data sets as training data sets, then use the training data sets to train the multi-task learning model, so as to obtain the trained multi-task learning model, use the predicted data sets to test the trained multi-task learning model, so as to obtain predicted sub-data, and then skip to the step of determining the predicted data sets from each data set until the number of the obtained predicted sub-data is the same as the number of each data set, so as to obtain each predicted sub-data, and determine the predicted data based on each predicted sub-data. The step of determining the prediction data based on each prediction sub-data specifically comprises the following steps: and obtaining all the predicted sub-data, calculating the average value of all the predicted sub-data, taking the average value as the predicted data, and outputting the predicted data by utilizing a neural network in the trained multi-task learning model to obtain the predicted data.
In this embodiment, in order to optimize the training process of the multi-task learning, promote the acceleration convergence of the model, the weight ratio of the loss function in the model needs to be reasonably set. Firstly, selecting root mean square error (Root Mean Squared Error, RMSE) as a loss function of each task, wherein compared with average absolute error, the RMSE can better reflect the actual situation of the predicted value error, and the calculation formula is as follows;
wherein y is trueThe sequence of real values is used to determine,n is the total length of the sequence, and the loss weight distribution of different tasks is required to be dynamically adjusted when the overall loss function is designed because the stage, difficulty and training effect of different tasks in the training process are different. The invention adjusts the loss weight in real time based on the difference of the descending trend of the loss function of different tasks, so as to improve the effective representation of the total loss function to different tasks and accelerate the convergence rate of multi-task training. In the present invention, the output of the predictive model is set to f in accordance with the predicted values (i.e., predicted data) of the electric load, the cold load, the heat load, and the clean energy load W (x)，f W A deep learning model representing shared weights, W is a weight parameter of the network, x represents selected features input to the network, and the true value of various loads is assumed to be y at the moment 1 ，y 2 ，y 3 ，y 4 All obey gaussian distributions, the likelihood probability of the model predictor versus the true value can be expressed as:
by log likelihood expansion, the overall loss function can be defined as:
wherein the sigma parameter can realize the weight ratio self-adaptive adjustment of different tasks, and moreover, the regularization term log sigma of the latter half part 1 σ 2 σ 3 σ 4 The magnitude of the different weights can be effectively weighed against certain tasks being overstretched or ignored.
Step S13: and constructing an overall loss function according to the predicted data, and judging whether the trained multi-task learning model converges or not based on the overall loss function.
In this embodiment, after determining whether the trained multi-task learning model converges, the method further includes: if the trained multi-task learning model is not converged, the step of transmitting the historical data to be predicted to a preset multi-task learning model for training is skipped to obtain a new overall loss function.
Furthermore, to increase the generalization ability of the model, a gradient penalty is added to the overall loss function:
where λ is the gradient penalty constant term.
Step S14: and if the trained multi-task learning model converges, predicting the carbon emission amount of the predicted data to obtain the predicted multi-energy carbon emission amount.
In this embodiment, the urban area energy history data sample library is input into a designed multi-task learning model for training and testing. In order to reduce the deviation of the recognition effect caused by unreasonable data division, the invention adopts a cross-validation training method, randomly and averagely divides an original data set into K (K is often 6-10) groups, takes each subset data as a validation set respectively, takes the rest K-1 groups of subset data as a training set, obtains K prediction models, and finally obtains the prediction moment values of different loads by comprehensively judging the K models. The predicted values of different loads are obtained through a prediction model and are respectively set as electric loads p e Thermal load p h Cold load p c Clean energy load p n And predicting the total carbon emission through energy planning. From the cost point of view, clean energy is first used to dissipate the electrical load, and if there is still a surplus, to dissipate the cold and heat loads, the total amount of carbon emissions in the urban area is calculated as follows:
Carbon＝c e ·ReLU(p e -p n )+c e ·ReLU(p c -ReLU(p n -p e ))+c h ·ReLU(p h -ReLU(ReLU(p n -p e )-p c ))
wherein c e ,c c ,c h The formula of the carbon emission factors corresponding to fossil energy consumed when electric, cold and heat loads cannot be completely consumed by clean energy, and the ReLU is an activation function:
ReLU(x)＝max(0,x)
the Chinese energy information network can know that the carbon emission factor of the Chinese power grid is c e ＝0.581tCO 2 The carbon emission factors of the cold load and the heat load are needed to be obtained according to the requirements of different areas, and the carbon emission factors of the cold load and the heat load of Zhejiang province are respectively c c ＝0.692tCO 2 /MW·h，c h ＝0.396tCO 2 /MW.h. Accordingly, the carbon emission prediction based on the multi-task learning is realized, and the effective prediction of the carbon emission can promote the follow-up energy planning scheme to meet the carbon emission requirement, thereby obtaining competitive advantages in the carbon trade market.
In the embodiment, historical load data of multiple energy sources are obtained, duration information to be predicted is determined, and feature extraction operation is performed on the historical load data according to the duration information to be predicted so as to obtain historical data to be predicted; transmitting the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain predicted data output after testing; constructing an overall loss function according to the prediction data, and judging whether the trained multi-task learning model converges or not based on the overall loss function; and if the trained multi-task learning model converges, predicting the carbon emission amount of the predicted data to obtain the predicted multi-energy carbon emission amount. The method and the device realize accurate prediction of different loads of the urban area with multiple energy flows by constructing the multi-task learning model, calculate the total carbon emission amount of the urban area through energy regulation and control, and finally realize short-term prediction of the carbon emission of the urban area. The multi-task learning can effectively reduce the prediction cost of multi-type energy sources, and can mine the high-dimensional coupling characteristic among different tasks so as to improve the prediction precision. The accurate prediction of the carbon emission can promote the regulation trend of the subsequent energy planning and improve the regulation potential of the energy-saving scheme.
Referring to fig. 2, the embodiment of the invention discloses a multi-energy flow carbon emission short-term prediction method based on a multi-task learning mechanism, which specifically comprises the following steps:
step S21: and acquiring multi-energy historical load data, determining duration information to be predicted, and performing feature extraction operation on the historical load data according to the duration information to be predicted to obtain the historical data to be predicted.
Step S22: and sending the historical data to be predicted to a preset multi-task learning model for training, so that the trained multi-task learning model tests the historical data to be predicted by utilizing a self shallow convolutional neural network to obtain the predicted data output after the test.
In this embodiment, a multi-task learning model is built to implement multi-energy flow load prediction. Because the selected characteristics comprise various load history requirements and meteorological information, the dimension explosion is easy to cause when the selected characteristics are directly input into a deep learning network for learning, and therefore, the shallow convolutional neural network (Convolutional Neural Network, CNN) is utilized to initially compress and couple the multi-functional load information. In order to shorten the training time of multi-task prediction and effectively utilize the high-dimensional correlation between different tasks to improve the prediction accuracy. The invention adopts a Swin-transducer model to perform multi-task learning on the high-dimensional characteristics obtained by the preliminary compression of the CNN. In this process, different tasks share the same weight data by adopting a weight sharing strategy. Thus, the prediction accuracy can be improved while the parameter number is reduced, and the training process of multi-task prediction can be accelerated. Finally, for the electric, thermal, cold and clean energy loads, a single connection network is adopted to realize data output, and the structure of the multi-task learning network designed by the invention is shown in figure 3.
Step S23: and acquiring multi-energy load actual data, calculating likelihood probability between the multi-energy load actual data and the predicted data, and constructing an overall loss function based on the likelihood probability.
Step S24: and judging whether the trained multi-task learning model converges or not based on the overall loss function.
Step S25: and if the trained multi-task learning model converges, predicting the carbon emission amount of the predicted data to obtain the predicted multi-energy carbon emission amount.
The specific step flow of the method is shown in fig. 4, (1) all historical electric, cold and heat loads of the urban area to be predicted are obtained, and clean energy output and historical load data of multiple energy sources of the corresponding area are obtained; (2) Performing characteristic extraction operation on the historical load data to obtain historical data to be predicted, and constructing a multi-energy historical data sample library according to the historical data to be predicted; (3) The historical data to be predicted is sent to a multi-task learning model for training, and the trained multi-task learning model is tested to obtain predicted data; (4) constructing an overall loss function according to the prediction data; (5) Judging whether the trained multi-task learning model converges or not based on the overall loss function; (6) If the trained multi-task learning model converges, predicting the carbon emission amount of the predicted data to obtain the predicted multi-energy carbon emission amount; if the trained multi-task learning model is not converged, the method jumps to the step of transmitting the historical data to be predicted to a preset multi-task learning model for training so as to obtain a new overall loss function.
In the embodiment, historical load data of multiple energy sources are obtained, duration information to be predicted is determined, and feature extraction operation is performed on the historical load data according to the duration information to be predicted so as to obtain historical data to be predicted; transmitting the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain predicted data output after testing; constructing an overall loss function according to the prediction data, and judging whether the trained multi-task learning model converges or not based on the overall loss function; and if the trained multi-task learning model converges, predicting the carbon emission amount of the predicted data to obtain the predicted multi-energy carbon emission amount. The method and the device realize accurate prediction of different loads of the urban area with multiple energy flows by constructing the multi-task learning model, calculate the total carbon emission amount of the urban area through energy regulation and control, and finally realize short-term prediction of the carbon emission of the urban area. The multi-task learning can effectively reduce the prediction cost of multi-type energy sources, and can mine the high-dimensional coupling characteristic among different tasks so as to improve the prediction precision. The accurate prediction of the carbon emission can promote the regulation trend of the subsequent energy planning and improve the regulation potential of the energy-saving scheme.
Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The invention provides a multi-energy flow carbon emission short-term prediction method based on a multi-task learning mechanism, which is described in detail above, and specific examples are applied to illustrate the principle and the implementation of the invention, and the description of the examples is only used for helping to understand the method and the core idea of the invention; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.
Claims (7)
1. A multi-energy flow carbon emission short-term prediction method based on a multi-task learning mechanism, comprising:
acquiring multi-energy historical load data, determining duration information to be predicted, and performing feature extraction operation on the historical load data according to the duration information to be predicted to obtain historical data to be predicted;
transmitting the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain predicted data output after testing;
constructing an overall loss function according to the prediction data, and judging whether the trained multi-task learning model converges or not based on the overall loss function;
and if the trained multi-task learning model converges, predicting the carbon emission amount of the predicted data to obtain the predicted multi-energy carbon emission amount.
2. The short-term prediction method of carbon emission in a multi-energy stream based on a multi-task learning mechanism according to claim 1, wherein the steps of transmitting the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain the predicted data output after testing include:
and sending the historical data to be predicted to a preset multi-task learning model for training, so that the trained multi-task learning model tests the historical data to be predicted by utilizing a self shallow convolutional neural network to obtain the predicted data output after the test.
3. The short-term prediction method of carbon emission of multi-energy stream based on the multi-task learning mechanism according to claim 1, wherein the determining the duration information to be predicted, performing feature extraction operation on the historical load data according to the duration information to be predicted, so as to obtain the historical data to be predicted, includes:
determining duration information to be predicted and a data feature extraction range according to service requirements, screening historical load data to be extracted from the historical load data according to the duration information to be predicted and the data feature extraction range, and performing feature extraction operation on the historical load data to be extracted to obtain historical data to be predicted;
and constructing a multi-energy historical data sample base according to the historical data to be predicted.
4. The method for short-term prediction of carbon emission in a multi-energy stream based on a multi-task learning mechanism according to claim 3, wherein the steps of transmitting the historical data to be predicted to a preset multi-task learning model for training, and testing the trained multi-task learning model to obtain the predicted data output after testing comprise the steps of:
the historical data to be predicted is sent to a preset multi-task learning model to be trained, so that the trained multi-task learning model adopts a cross-validation training method to determine data sets from the multi-energy historical data sample library, the number of the data sets is determined, then a predicted data set is screened out from the data sets, other data sets except the predicted data set are used as training data sets, the training data sets are used for training the multi-task learning model to obtain the trained multi-task learning model, the trained multi-task learning model is tested by the predicted data sets to obtain predicted sub-data, then the step of determining the predicted data sets from the data sets is skipped until the number of the obtained predicted sub-data sets is the same as the number of the data sets, the predicted sub-data is obtained, and the predicted data is determined based on the predicted sub-data.
5. The method for short-term prediction of carbon emissions in a multi-energy stream based on a multi-task learning mechanism of claim 4, wherein said determining said predicted data based on each of said predicted sub-data comprises:
acquiring all the predicted sub-data, calculating the average value of all the predicted sub-data, and taking the average value as the predicted data;
and outputting the predicted data by using the trained neural network in the multi-task learning model to obtain the predicted data.
6. The method for short-term prediction of carbon emission in a multi-energy stream based on a multi-task learning mechanism according to claim 1, wherein after the determining whether the trained multi-task learning model converges based on the overall loss function, further comprising:
if the trained multi-task learning model is not converged, the step of transmitting the historical data to be predicted to a preset multi-task learning model for training is skipped to obtain a new overall loss function.
7. The method for short-term prediction of carbon emissions in a multi-energy stream based on a multi-task learning mechanism according to any one of claims 1 to 6, wherein said constructing an overall loss function from said prediction data comprises:
acquiring multi-energy load actual data, and calculating likelihood probability between the multi-energy load actual data and the predicted data;
an overall loss function is constructed based on the likelihood probabilities.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310570606.8A CN116485036A (en) | 2023-05-16 | 2023-05-16 | Multi-energy-flow carbon emission short-term prediction method based on multi-task learning mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310570606.8A CN116485036A (en) | 2023-05-16 | 2023-05-16 | Multi-energy-flow carbon emission short-term prediction method based on multi-task learning mechanism |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116485036A true CN116485036A (en) | 2023-07-25 |
Family
ID=87213938
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310570606.8A Pending CN116485036A (en) | 2023-05-16 | 2023-05-16 | Multi-energy-flow carbon emission short-term prediction method based on multi-task learning mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116485036A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117369282A (en) * | 2023-11-17 | 2024-01-09 | 上海四方无锡锅炉工程有限公司 | Control method for adaptive hierarchical air supply and solid waste CFB boiler thereof |
-
2023
- 2023-05-16 CN CN202310570606.8A patent/CN116485036A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117369282A (en) * | 2023-11-17 | 2024-01-09 | 上海四方无锡锅炉工程有限公司 | Control method for adaptive hierarchical air supply and solid waste CFB boiler thereof |
| CN117369282B (en) * | 2023-11-17 | 2024-04-19 | 上海四方无锡锅炉工程有限公司 | Control method for adaptive hierarchical air supply and solid waste CFB boiler thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107591844B (en) | Active power distribution network robust reconstruction method considering node injection power uncertainty | |
| CN108898251B (en) | Offshore wind farm power prediction method considering meteorological similarity and power fluctuation | |
| CN110912205B (en) | Wind power system operation standby dynamic scheduling optimization method based on scene set | |
| CN110380444B (en) | Capacity planning method for distributed wind power orderly access to power grid under multiple scenes based on variable structure Copula | |
| CN112464566A (en) | Transformer oil temperature prediction method based on genetic algorithm and BP neural network | |
| Chen et al. | Probabilistic operational reliability of composite power systems considering multiple meteorological factors | |
| CN116485036A (en) | Multi-energy-flow carbon emission short-term prediction method based on multi-task learning mechanism | |
| CN103971175A (en) | Short-term load prediction method of multistage substations | |
| CN112994115A (en) | New energy capacity configuration method based on WGAN scene simulation and time sequence production simulation | |
| CN110837915A (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
| CN102880907A (en) | Method and device for correcting wind speed | |
| CN111612244A (en) | QRA-LSTM-based method for predicting nonparametric probability of photovoltaic power before day | |
| CN110969306A (en) | Power distribution low-voltage distribution area load prediction method and device based on deep learning | |
| CN116826710A (en) | Peak clipping strategy recommendation method and device based on load prediction and storage medium | |
| CN109978236B (en) | Small hydropower station short-term power refined prediction method based on feature combination | |
| CN111680712A (en) | Transformer oil temperature prediction method, device and system based on similar moments in the day | |
| CN114611757A (en) | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network | |
| CN113449847A (en) | Offshore wind power rolling prediction method considering second-level time series wind speed change | |
| CN108694475B (en) | Short-time-scale photovoltaic cell power generation capacity prediction method based on hybrid model | |
| CN115313422B (en) | Operation control method of electricity-gas comprehensive energy system for actively supporting main network frequency | |
| CN116167465A (en) | Solar irradiance prediction method based on multivariate time series ensemble learning | |
| CN115764855A (en) | Real-time adjustable capacity and available electric quantity prediction method for electric vehicle quick charging station | |
| CN114707422A (en) | Intelligent power check method based on load prediction | |
| CN114759579A (en) | Power grid active power optimization control system, method and medium based on data driving | |
| Zhang et al. | A hybrid load forecasting method based on neural network in smart grid |
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 |