CN112613666B - Power grid load prediction method based on graph convolution neural network and transfer learning - Google Patents
Power grid load prediction method based on graph convolution neural network and transfer learning Download PDFInfo
- Publication number
- CN112613666B CN112613666B CN202011569261.7A CN202011569261A CN112613666B CN 112613666 B CN112613666 B CN 112613666B CN 202011569261 A CN202011569261 A CN 202011569261A CN 112613666 B CN112613666 B CN 112613666B
- Authority
- CN
- China
- Prior art keywords
- power grid
- model
- nodes
- time
- graph convolution
- 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
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000013526 transfer learning Methods 0.000 title claims abstract description 21
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 17
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims abstract description 14
- 238000004364 calculation method Methods 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 claims 1
- 238000010276 construction Methods 0.000 abstract description 2
- 238000013508 migration Methods 0.000 abstract 1
- 230000005012 migration Effects 0.000 abstract 1
- 230000007547 defect Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
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
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/048—Activation functions
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a power grid load prediction method based on a graph convolution neural network and transfer learning. Aiming at the time-space characteristics of power grid load data, a power grid load prediction method suitable for a large-scale area is designed by adopting a graph convolution network and combining a migration learning method. The method comprises the steps of dividing the whole power grid into a plurality of sub-areas according to areas and independently constructing a model for each sub-area for prediction. In the model construction, a dynamic time warping algorithm is adopted to fuse time similarity, a graph convolution network is introduced by using a power grid topological structure to extract spatial features, time features are extracted through convolution operation in a time domain, and network models in different regions are trained by using a transfer learning method. The method can reduce the calculation pressure of the cloud server, improve the accuracy of power grid load prediction and reduce the time of model training.
Description
Technical Field
The invention relates to the technical field of power grid load prediction, in particular to a power grid load prediction method based on a graph convolution neural network and transfer learning.
Background
The power grid load prediction plays a very key role in tasks such as power utilization planning, operation scheduling and the like of a power system. Because the load of the power grid is influenced by various factors, the load is closely related to conditions such as economic level, industrial structure, meteorological conditions and the like of the region, and certain complexity is presented. In addition, the change of the load of the power grid has certain regularity and periodicity. Therefore, how to accurately predict the grid load is an important issue facing the power system. However, most of the previous research has mainly focused on the prediction of the load of a single node or the total load of the grid. The power grid is a wide area network with a large range, multiple regions and multiple nodes, so how to comprehensively consider the correlation among different regions of the power grid and the space-time characteristics among the nodes to realize the rapid and large-range accurate prediction of the load of each node of the power grid has great economic value and practical significance.
There are two main types of statistics for grid charge prediction, the first type is based on traditional statistical methods, such as autoregressive integrated moving average ARIMA, vector autoregressive VAR, etc. Such methods typically require data to satisfy some assumptions, but these grid data are too complex to satisfy these assumptions, and therefore they typically have poor prediction accuracy in practical applications. The second category of prediction methods is based on data-driven methods. In this class of methods, the performance of these methods is superior to statistical methods, based mainly on traditional machine learning, such as Support Vector Regression (SVR), k-nearest neighbor (KNN). However, these methods require many complex feature extractions and are not suitable for processing large amounts of complex grid data. With the rapid development of deep learning, breakthroughs are made in many fields (such as image processing and voice recognition). The method for obtaining the space-time correlation of the data by using the deep learning-based method, such as a Convolutional Neural Network (CNN) and a cyclic convolutional neural network (RNN), can be used for effectively predicting the load data of the power grid, but cannot sufficiently extract the space correlation between the power grid nodes. Also, generally only a single node can be predicted. In order to solve the problems, the invention provides a power grid load prediction method based on a graph convolution neural network and transfer learning.
Disclosure of Invention
The invention aims to provide a power grid load prediction method based on a graph convolution neural network and transfer learning.
In order to achieve the purpose, the technical scheme of the invention is as follows: a power grid load prediction method based on a graph convolution neural network and transfer learning comprises the following steps:
step S1: dividing a region to be predicted into a plurality of sub-regions according to the geographical position information, and independently constructing a model for each sub-region to predict the power grid load of the region;
step S2: extracting time-space information of the power grid nodes of each sub-region, including topological structure information of the power grid nodes and time sequence information of load data, and constructing a prediction model by adopting a graph convolution neural network and a full connection layer;
and step S3: training the constructed model by adopting the power grid load data of all nodes in the subareas;
and step S4: and sharing the trained network model to the adjacent subareas as an initial training model by adopting a transfer learning method, and training each subarea model.
In an embodiment of the present invention, in step S1, the grid area division is performed as follows:
step S11: according to the geographical position information, the power grid node topological structure is divided into a plurality of sub-graphs, nodes with more connection points in each sub-graph are extracted as central nodes, and the sub-graphs with similar sizes are divided according to the central nodes. And overlapping partial nodes in the nearby area on the subgraph with few connection points to ensure that the subgraph has similar size.
In the embodiment of the present invention, in step S2, the time-space information of each sub-region is integrated into the model construction through the time warping algorithm, and the load of each node of the power grid is predicted by using the graph convolution neural network. The method comprises the following specific steps:
s21, calculating the similarity between different nodes in the subgraph by using a dynamic time warping algorithm, wherein the calculation method comprises the following steps:
max(n，m)≤K≤n+m, where (n, m) is the length of two different time series of information, dist (·) represents a distance computation function;
step S22: constructing a topological graph G containing spatial information according to the connectivity of the power grid topological structure S According to the sequence similarity calculated by the dynamic time warping algorithm, the first 5 percent of most similar nodes are taken to be connected to form a new topological graph G ST And time information and space information of the power grid nodes are fused.
In the embodiment of the present invention, in the step S3, the specific operations are as follows:
step S31: respectively carrying out graph convolution operation on power grid load data of one week, one day and one hour before all nodes of a sub-region, extracting spatial information in the data, then carrying out convolution operation on a time domain, extracting time correlation, carrying out nonlinear transformation in a network layer by adopting a ReLU as an activation function, fusing features extracted from data at different moments, and finally carrying out prediction through a full connection layer.
Step S32: comparing the prediction output of the constructed model with the true value, calculating a loss value, optimizing the model, adopting a mean square error as a loss function, then continuously optimizing parameters of the frame by using a back propagation algorithm, continuously calculating a parameter gradient in the back propagation algorithm, continuously adapting the learning rate by using an RMSprop algorithm, and optimizing the parameters to enable the model to reach an optimal solution.
In the embodiment of the present invention, in step S4, a transfer learning method is adopted, in a source domain, a trained model is used as an initial model of an adjacent region, then sub-regions similar to or close to the source domain are selected as target domains, and power grid load data of each sub-region is adopted to train the model.
Compared with the prior art, the invention has the following beneficial effects: the grid load prediction based on the graph convolution neural network and the transfer learning, provided by the invention, can overcome the defect of too large calculated amount caused by cloud computing in the existing method, and meanwhile, the space-time correlation of the grid load data is integrated into a model, so that the prediction precision is improved. The defect of long training time is overcome by adopting transfer learning, and the model training efficiency is improved.
Drawings
Fig. 1 is an overall model frame in an embodiment of the present invention.
Detailed Description
The technical scheme of the invention is specifically explained below with reference to the accompanying drawings.
The invention discloses a power grid load prediction method based on a graph convolution neural network and transfer learning, which is specifically realized according to the following steps as shown in figure 1:
step S1: according to the geographical position information, the power grid node topological structure is divided into a plurality of sub-graphs, nodes with more connection points in each sub-graph are extracted as central nodes, and the sub-graphs with similar sizes are divided according to the central nodes. When the subgraph is divided specifically, the subgraph is divided into areas with similar node numbers as much as possible, but the same node number is difficult to guarantee in the specific implementation process, and a zero padding method can be adopted, or nodes in different areas are used in an overlapping mode, so that the subgraph has similar size.
Step S2: extracting time-space information of the power grid nodes of each sub-region, including topological structure information of the power grid nodes and time sequence information of load data, and constructing a prediction model by adopting a graph convolution neural network and a full connection layer;
firstly, calculating the similarity between different nodes in the subgraph by using a dynamic time warping algorithm, wherein the calculation method comprises the following steps:
max (n, m) is more than or equal to K and less than or equal to n + m, wherein (n, m) is the length of two different time sequence information, and Dist (DEG) represents a distance calculation function;
further, constructing a topological graph G containing spatial information according to the connectivity of the power grid topological structure S According to the sequence similarity calculated by the dynamic time warping algorithm, the first 5 percent of most similar nodes are taken to be connected to form a new topological graph G ST And time information and space information of the power grid nodes are fused.
And step S3: training the constructed graph convolution network model by adopting the power grid load data of all nodes in the sub-region;
firstly, respectively carrying out graph convolution operation on power grid load data of one week, one day and one hour before all nodes of a sub-region, extracting spatial information in the data, then carrying out convolution operation on a time domain, extracting time correlation, carrying out nonlinear transformation in a network layer by adopting ReLU as an activation function, fusing features extracted from data at different moments, and finally carrying out prediction through a full connection layer.
Graph convolution is as follows;
g θ * G x＝g θ (UΛU T )x＝Ug O (L)xU T x
in the above formula G Represents the graph convolution operation, g θ The method comprises the following steps that (1) a convolution kernel is adopted, x is power grid load data, and x adopts data of one day, one week and one hour before a predicted point; l is a Laplace matrix
A is an adjacent matrix of the constructed subgraph, D is the degree of the adjacent matrix, and a matrix U is a feature vector of the normalized Laplace matrix; spatial features of different nodes can be extracted through graph convolution;
further, CNN by convolution of the time domain
Wherein
Representing the output of the r-th layer, the subscript T indicates the current usage data period, after which non-linear similarities can be extracted using ReLU as the activation function. And finally, fusing the characteristics through a full connection layer.
Furthermore, the prediction results corresponding to the data of one week, one day and one hour are fused through full connection respectively:
Y＝W h ⊙Y h +W d ⊙Y d +W w ⊙Y w
therein asRepresenting a matrix dot product; y is the prediction result after fusion, Y h 、Y d 、Y w Respectively representing the prediction of corresponding data one hour before, one day before and one week before;
and further, comparing the prediction output of the constructed model with a true value, calculating a loss value, optimizing the model, adopting a mean square error as a loss function, then continuously optimizing parameters of the frame by using a back propagation algorithm, continuously calculating a parameter gradient in the back propagation algorithm, continuously self-adapting the learning rate by using an RMSprop algorithm, and optimizing the parameters to enable the model to reach an optimal solution.
The RMSprop algorithm uses a variable MeanSquare (w, t) to store the average value of the square of the gradient of a period of time before each weight value when the learning rate is updated for the t time, and the RMSprop algorithm adapts to the learning rate according to the variable and continuously optimizes parameters to enable the structure to reach the optimal solution.
Mean square error function (MSE):
RMSprop formula:
And step S4: and sharing the trained network model to the adjacent subareas as an initial model by adopting a transfer learning method, and training each subarea model to accelerate the model training speed.
Compared with the prior art, the invention has the following beneficial effects: the invention provides a power grid load prediction method based on a graph convolution neural network and transfer learning, which can overcome the defect of large calculation amount caused by cloud computing in the existing method, and simultaneously, the space-time correlation of power grid load data is integrated into a graph model, so that the prediction precision is improved. The transfer learning is adopted, so that the defect of long training time is overcome, and the model training efficiency is improved.
The above are preferred embodiments of the present invention, and all changes made according to the technical solutions of the present invention that produce functional effects do not exceed the scope of the technical solutions of the present invention belong to the protection scope of the present invention.
Claims (3)
1. A power grid load prediction method based on a graph convolution neural network and transfer learning is characterized in that: comprises the following steps of (a) carrying out,
step S1: according to the geographical position information, dividing a region to be predicted into a plurality of sub-regions, and independently constructing a model for each sub-region to predict the power grid load of the region;
step S2: extracting time-space information of the power grid nodes of each sub-area, including topological structure information of the power grid nodes and time sequence information of load data, and constructing a prediction model by adopting a graph convolution neural network and a full connection layer;
and step S3: training the constructed model by adopting the power grid load data of all nodes in the subareas;
and step S4: sharing the trained network model to the adjacent subareas as an initial training model by adopting a transfer learning method, and training each subarea model;
in step S2, the specific space-time diagram configuration is performed as follows:
step S21: the similarity between different nodes in the subgraph is calculated by using a dynamic time warping algorithm, and the calculation method is as follows:
where (n, m) is the length of two different time series information, dist (·) represents a distance computation function;
step S22: constructing a topological graph G containing spatial information according to the connectivity of the power grid topological structure s Calculating sequence similarity according to dynamic time warping algorithm, and connecting the first 5% most similar nodesForm a new topological graph G ST Time information and space information of the power grid nodes are fused;
in the step S3, the step of,
step S31: respectively adopting power grid load data of one week, one day and one hour before all nodes of a sub-region to respectively carry out graph convolution operation, extracting spatial information in the data, then carrying out convolution operation on a time domain, extracting time correlation, adopting ReLU as an activation function in a network layer to carry out nonlinear transformation, fusing the extracted characteristics of the data at different moments, and finally carrying out prediction through a full connection layer;
graph convolution is as follows;
g θ * G x＝g θ (UΛU T )x＝Ug θ (L)xU T x
in the above formula G Represents the graph convolution operation, g θ The method comprises the following steps that (1) a convolution kernel is adopted, x is power grid load data, and x adopts data of one day, one week and one hour before a predicted point; l is a Laplace matrix
A is an adjacent matrix of the constructed subgraph, D is the degree of the adjacent matrix, and a matrix U is a feature vector of the normalized Laplace matrix; spatial features of different nodes can be extracted through graph convolution;
further, by convolution of the time domain CNN:
wherein
The output of the r-th layer is shown, subscript T shows the current data cycle of use, and then nonlinear similarity can be extracted by using ReLU as an activation function; finally, fusing the characteristics through a full connection layer;
furthermore, the prediction results corresponding to the data of one week, one day and one hour are fused through full connection respectively:
Y＝W h ⊙Y h +W d ⊙Y d +W w ⊙Y w
this wherein |, represents a matrix dot product; y is the prediction result after fusion, Y h 、Y d 、Y w Respectively representing the prediction of corresponding data one hour before, one day before and one week before;
step S32: comparing the prediction output of the constructed model with the true value, calculating a loss value, optimizing the model, adopting a mean square error as a loss function, then continuously optimizing parameters of the frame by using a back propagation algorithm, continuously calculating a parameter gradient in the back propagation algorithm, continuously adapting the learning rate by using an RMSprop algorithm, and optimizing the parameters to enable the model to reach an optimal solution.
2. The grid load prediction method based on the graph convolution neural network and the transfer learning as claimed in claim 1, wherein: in step S1, the area division is performed as follows:
step S11: dividing a power grid node topological structure into a plurality of subgraphs according to geographical position information, extracting nodes with more connection points in each subgraph as central nodes, and dividing the subgraphs with similar sizes according to the central nodes; and overlapping partial nodes in the nearby area on the subgraph with few connection points to ensure that the subgraph has similar size.
3. The grid load prediction method based on the graph convolution neural network and the transfer learning as claimed in claim 1, wherein: in the step S4, a transfer learning method is adopted, the trained model is used as an initial model of an adjacent area on a source area, then sub-areas similar to or close to the source area are selected as target areas, and the model is trained by adopting power grid load data of the sub-areas.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011569261.7A CN112613666B (en) | 2020-12-26 | 2020-12-26 | Power grid load prediction method based on graph convolution neural network and transfer learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011569261.7A CN112613666B (en) | 2020-12-26 | 2020-12-26 | Power grid load prediction method based on graph convolution neural network and transfer learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112613666A CN112613666A (en) | 2021-04-06 |
| CN112613666B true CN112613666B (en) | 2023-02-14 |
Family
ID=75248151
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011569261.7A Active CN112613666B (en) | 2020-12-26 | 2020-12-26 | Power grid load prediction method based on graph convolution neural network and transfer learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112613666B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114742258B (en) * | 2022-02-23 | 2022-11-29 | 吉林化工学院 | Migratable residential household power load scene prediction method |
| CN114780870B (en) * | 2022-04-15 | 2023-03-24 | 北京骑胜科技有限公司 | Order quantity prediction method, system, device, server, terminal and storage medium |
| CN115564180B (en) * | 2022-09-01 | 2023-10-10 | 北京京能清洁能源电力股份有限公司北京分公司 | Power network reliability assessment method based on big data analysis |
| CN115983448B (en) * | 2022-12-14 | 2023-09-12 | 南京信息工程大学 | Multi-energy load prediction method based on space-time diagram neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109754113A (en) * | 2018-11-29 | 2019-05-14 | 南京邮电大学 | Load forecasting method based on dynamic time warping Yu length time memory |
| CN110717526A (en) * | 2019-09-23 | 2020-01-21 | 华南理工大学 | Unsupervised transfer learning method based on graph convolution network |
| CN111582610A (en) * | 2020-07-13 | 2020-08-25 | 清华四川能源互联网研究院 | Prediction method for family energy decomposition based on convolutional neural network |
| CN111860951A (en) * | 2020-06-12 | 2020-10-30 | 北京工业大学 | Rail transit passenger flow prediction method based on dynamic hypergraph convolutional network |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10795935B2 (en) * | 2016-02-05 | 2020-10-06 | Sas Institute Inc. | Automated generation of job flow definitions |
| CN111525587B (en) * | 2020-04-01 | 2022-10-25 | 中国电力科学研究院有限公司 | Reactive load situation-based power grid reactive voltage control method and system |
-
2020
- 2020-12-26 CN CN202011569261.7A patent/CN112613666B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109754113A (en) * | 2018-11-29 | 2019-05-14 | 南京邮电大学 | Load forecasting method based on dynamic time warping Yu length time memory |
| CN110717526A (en) * | 2019-09-23 | 2020-01-21 | 华南理工大学 | Unsupervised transfer learning method based on graph convolution network |
| CN111860951A (en) * | 2020-06-12 | 2020-10-30 | 北京工业大学 | Rail transit passenger flow prediction method based on dynamic hypergraph convolutional network |
| CN111582610A (en) * | 2020-07-13 | 2020-08-25 | 清华四川能源互联网研究院 | Prediction method for family energy decomposition based on convolutional neural network |
Non-Patent Citations (2)
| Title |
|---|
| Domain-Adversarial Transfer Learning for Robust Intrusion Detection in the Smart Grid;Yongxuan Zhang;《2019 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm)》;20191125;全文 * |
| 基于可变输入LSTM和迁移学习的非侵入式电力负荷分解;梁家阁;《国优秀博硕士学位论文全文数据库(硕士) 工程科技II辑》;20200615;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112613666A (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112613666B (en) | Power grid load prediction method based on graph convolution neural network and transfer learning | |
| Zhang et al. | Gaussian mutational chaotic fruit fly-built optimization and feature selection | |
| He et al. | Graph attention spatial-temporal network with collaborative global-local learning for citywide mobile traffic prediction | |
| Cai et al. | Discrete particle swarm optimization for identifying community structures in signed social networks | |
| CN111696345A (en) | Intelligent coupled large-scale data flow width learning rapid prediction algorithm based on network community detection and GCN | |
| CN110851662B (en) | Heterogeneous information network link prediction method based on meta-path | |
| Pu et al. | UFNGBM (1, 1): A novel unbiased fractional grey Bernoulli model with Whale Optimization Algorithm and its application to electricity consumption forecasting in China | |
| CN113326884B (en) | Efficient learning method and device for large-scale heterograph node representation | |
| Xue et al. | A graph regularized point process model for event propagation sequence | |
| Wang et al. | Adaptive multi-receptive field spatial-temporal graph convolutional network for traffic forecasting | |
| Ramirez-Gonzalez et al. | Convolutional neural nets with hyperparameter optimization and feature importance for power system static security assessment | |
| Meirom et al. | Optimizing tensor network contraction using reinforcement learning | |
| Liu et al. | DCENet: A dynamic correlation evolve network for short-term traffic prediction | |
| Bai et al. | Evaluation for development effect of enterprise innovation with neural network from low-carbon economy | |
| Lee et al. | Wind prediction based on general regression neural network | |
| Li et al. | Path planning of mobile robot based on dynamic chaotic ant colony optimization algorithm | |
| Li et al. | A sequence and network embedding method for bus arrival time prediction using GPS trajectory data only | |
| Xu et al. | Traffic flow forecasting with spatial-temporal graph convolutional networks in edge-computing systems | |
| Zhang et al. | A novel approach of tensor‐based data missing estimation for Internet of Vehicles | |
| CN115001978A (en) | Cloud tenant virtual network intelligent mapping method based on reinforcement learning model | |
| Yin et al. | Nodetrans: A graph transfer learning approach for traffic prediction | |
| Deng et al. | Learning structured components: Towards modular and interpretable multivariate time series forecasting | |
| CN115080795A (en) | Multi-charging-station cooperative load prediction method and device | |
| Fei et al. | Orthogonal spatial-temporal graph convolutional networks for traffic flow forecasting | |
| CN103440540B (en) | A kind of parallel method of land utilization space layout artificial immunity Optimized model |
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 |