CN111914486A - Power system transient stability evaluation method based on graph attention network - Google Patents
Power system transient stability evaluation method based on graph attention network Download PDFInfo
- Publication number
- CN111914486A CN111914486A CN202010791074.7A CN202010791074A CN111914486A CN 111914486 A CN111914486 A CN 111914486A CN 202010791074 A CN202010791074 A CN 202010791074A CN 111914486 A CN111914486 A CN 111914486A
- Authority
- CN
- China
- Prior art keywords
- sample
- node
- layer
- power system
- transient stability
- 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.)
- Granted
Links
- 230000001052 transient effect Effects 0.000 title claims abstract description 68
- 238000011156 evaluation Methods 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 claims abstract description 37
- 239000011159 matrix material Substances 0.000 claims abstract description 33
- 238000013210 evaluation model Methods 0.000 claims abstract description 27
- 238000011176 pooling Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims description 35
- 101001094098 Homo sapiens Sodium- and chloride-dependent GABA transporter 3 Proteins 0.000 claims description 26
- 102100035254 Sodium- and chloride-dependent GABA transporter 3 Human genes 0.000 claims description 26
- 238000005070 sampling Methods 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 17
- 230000004913 activation Effects 0.000 claims description 13
- 101000639970 Homo sapiens Sodium- and chloride-dependent GABA transporter 1 Proteins 0.000 claims description 12
- 101001094079 Homo sapiens Sodium- and chloride-dependent GABA transporter 2 Proteins 0.000 claims description 12
- 102100033927 Sodium- and chloride-dependent GABA transporter 1 Human genes 0.000 claims description 12
- 102100035242 Sodium- and chloride-dependent GABA transporter 2 Human genes 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 11
- 238000013097 stability assessment Methods 0.000 claims description 9
- 238000000547 structure data Methods 0.000 claims description 8
- 238000002372 labelling Methods 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 6
- 238000004088 simulation Methods 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 4
- 230000014509 gene expression Effects 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims 1
- 230000006870 function Effects 0.000 description 13
- 230000008859 change Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 230000002776 aggregation Effects 0.000 description 4
- 238000004220 aggregation Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000368 destabilizing effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/08—Learning methods
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/04—Power grid distribution networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- 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
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Development Economics (AREA)
- Biophysics (AREA)
- Entrepreneurship & Innovation (AREA)
- Biomedical Technology (AREA)
- Marketing (AREA)
- Educational Administration (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Power Engineering (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a transient stability evaluation method of a power system based on a graph attention network, which comprises the following steps: acquiring operation data, wherein the operation data comprises an operation data matrix X constructed by real-time operation data of a power grid measured by a PMU and an adjacency matrix A constructed according to a power grid operation topology; and inputting the operation data into a transient stability evaluation model of the power system for evaluation to obtain a stability result. According to the method, the power grid topology is embedded into the evaluation model in an adjacent matrix mode, so that the difference of transient characteristics under different topologies is effectively considered, and the model mobility is improved; through the pooling layer before the classifier, the model can still be suitable under the condition that the system scale is changed, the problem that the model needs to be redesigned and trained under the condition that the system scale is enlarged or reduced is avoided while the evaluation performance is ensured, and the transient stability evaluation model cannot be learned deeply.
Description
Technical Field
The invention relates to deep learning transient stability evaluation of a power system, in particular to a power system transient stability evaluation method based on a graph attention network.
Background
The rapid and accurate transient stability assessment of the power system is of great significance. The transient stability evaluation technology of the power system based on the deep learning model is concerned, and a mapping relation between the observable operation characteristic quantity of the system and the stable result can be established through the learning of a large number of samples, so that the stability evaluation is rapidly carried out. At present, common deep learning models such as CNN, SAE, DBN and the like can not effectively consider the influence of a power grid topological structure when feature extraction is carried out, and the established transient state evaluation method has low adaptability to power grid topological change and has an unsatisfactory feature extraction effect on high-dimensional operation data. When the scale of the power grid changes, such as node increase and decrease, the model parameters trained before are difficult to be effectively utilized, and the model structure is often required to be redesigned and trained, so that the migration and application value of the model are influenced.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides a power system transient stability evaluation method for a graph attention network, so as to solve the problem that the power grid topology change is difficult to be effectively considered when the characteristics of a power system transient stability evaluation model based on deep learning are extracted, and the dynamic correlation between adjacent nodes can be better mined and embedded into a high-dimensional characteristic aggregation process, so that the application performance of the transient stability evaluation model can be obviously improved.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a power system transient stability evaluation method based on a graph attention network comprises the following steps:
acquiring operation data, wherein the operation data comprises an operation data matrix X constructed by real-time operation data of a power grid measured by a PMU and an adjacency matrix A constructed according to a power grid operation topology;
inputting the operation data into a transient stability evaluation model of the power system for evaluation to obtain a stability result;
the power system transient stability evaluation model is obtained by the following steps:
step 1: generating power system transient process samples under different operating topologies through power system simulation software, and taking the power system transient process samples as an original sample set; forming a graphic structure data G (A, X) by the adjacency matrix A of the power grid topology corresponding to the sample and the power grid operation state data X, and labeling the sample data according to whether the sample data is stable or not;
step 2: constructing a graph attention network for feature extraction of graph-like structure data G (A, X), the graph attention network being composed of a plurality of layers;
and step 3: performing pooling operation on output data of the last layer of graph attention network, and inputting the output data into a full-connection classifier layer to realize stable classification;
and 4, step 4: dividing the original sample set marked in the step 1 into a training set, a verification set and a test set according to a set proportion to obtain a transient stability evaluation model of the power system with optimal accuracy
Compared with the prior art, the invention has the beneficial effects that:
according to the method, the power grid topology is embedded into the evaluation model in an adjacent matrix mode, so that the difference of transient characteristics under different topologies is effectively considered, and the model mobility is improved; through the pooling layer before the classifier, the model can still be suitable under the condition that the system scale is changed, the problem that the model needs to be redesigned and trained under the condition that the system scale is enlarged or reduced is avoided while the evaluation performance is ensured, and the transient stability evaluation model cannot be learned deeply.
Drawings
Fig. 1 is a flowchart of a power system transient stability evaluation method based on a graph attention network.
Fig. 2 is a topological block diagram of an example new england 10 machine 39 node system.
FIG. 3 is a schematic diagram of attention coefficient calculation and feature aggregation of the attention network layer of the graph according to the present invention.
FIG. 4 is a schematic diagram of a multi-head attention mechanism of the multi-head attention layer of the present invention.
FIG. 5 is a diagram of a transient stability assessment model according to the present invention.
Detailed Description
Example (b):
the technical scheme of the invention is further explained by combining the drawings and the embodiment as follows:
the invention introduces a graph attention network to obtain a power system transient stability evaluation method based on the graph attention network. Transient characteristics of the power system under different operation topologies are captured from the adjacency matrix by using the graph attention network, so that the model has better evaluation capability on transient stability under the condition of topology change of the power system. On the other hand, the pooling layer technology is added in front of the classifier layer, so that the problem that the model needs to be redesigned and trained when the system is enlarged or reduced in scale is avoided while the evaluation performance is ensured.
With reference to fig. 1, the method for evaluating transient stability of a power system based on a graph attention network provided in this embodiment includes:
acquiring operation data, wherein the operation data comprises an operation data matrix X constructed by real-time operation data of a power grid measured by a PMU and an adjacency matrix A constructed according to a power grid operation topology;
inputting the operation data into a transient stability evaluation model of the power system for evaluation to obtain a stability result;
the power system transient stability evaluation model is obtained by the following steps:
step 1: generating power system transient process samples under different operating topologies through power system simulation software, and taking the power system transient process samples as an original sample set; forming a graphic structure data G (A, X) by the adjacency matrix A of the power grid topology corresponding to the sample and the power grid operation state data X, and labeling the sample data according to whether the sample data is stable or not;
step 2: constructing a graph attention network for feature extraction of graph-like structure data G (A, X), the graph attention network being composed of a plurality of layers;
and step 3: performing pooling operation on output data of the last layer of graph attention network, and inputting the output data into a full-connection classifier layer to realize stable classification;
and 4, step 4: and (3) dividing the original sample set marked in the step (1) into a training set, a verification set and a test set according to a set proportion so as to obtain the transient stability evaluation model of the power system with the optimal accuracy.
Therefore, the power grid topology is embedded into the evaluation model in an adjacent matrix mode, so that the difference of transient characteristics under different topologies is effectively considered, and the model mobility is improved; through the pooling layer before the classifier, the model can still be suitable under the condition that the system scale is changed, the problem that the model needs to be redesigned and trained under the condition that the system scale is enlarged or reduced is avoided while the evaluation performance is ensured, and the transient stability evaluation model cannot be learned deeply.
Optionally, in step 1, X is organized by taking nodes as a unit, and includes node voltage amplitude, phase angle, injected active power and reactive power of each sampling point during a period from a time before a fault occurs to a time after the fault is removed, and a sampling interval is 0.01s, so that only time series data from the time before the fault occurs to the time after the fault is removed is collected as an original input feature, a response time of the model is short, and a control measure is taken for a scheduling center as much as possible.
A and X in the derived sample feature G (a, X) are respectively represented as:
in the formula, ai,jA value of 0 or 1, ai,j1 denotes that bus i is connected to bus j, ai,j0 represents that the bus i is not connected with the bus j, and diag (A) is set to 1; u shapek,yRepresents the kth bus voltage amplitude at the y time of the system; thetak,yRepresents the kth bus voltage phase angle at the y time of the system; pk,yThe active power injected by the kth bus node at the y moment of the system is represented; qk,yRepresenting the reactive power injected by the kth bus node at the y-th moment of the system. If the bus node is neither a generator node nor a load node, Pk,y=Qk,y0. i, j, k ═ 1,2, …, n; y is 1,2, …, t; n is the total number of the system buses, and t is the sampling time. The node numbering sequence has no special rule, and the node numbering sequence in the matrixes A and X only needs to be ensured to be the same.
In order to reduce the influence of the dimension on the model, carrying out the following data standardization processing on a matrix X representing the power grid operation state in the sample set according to columns, wherein the standardization formula is as follows:
in the formula, XyFor the feature sample data collected at the y-th sampling instant,the normalized value of the characteristic sample data collected at the y-th sampling moment,characteristic sample collected for the y sampling timeThe mean value of the present data is,normalizing the features of t moments to obtain a sample feature data set X, wherein y is 1,2, …, t, the standard deviation of the feature sample data acquired at the y-th sampling momentstd。
Carrying out stability labeling on the sample according to a transient stability simulation result, wherein the relative power angle difference of any two generators is larger than 180 degrees, the transient instability of the system is caused, and the sample stability label is [0,1 ]; otherwise, the transient stability of the system is stable, and the sample stability label is [1,0 ].
Optionally, in step 2, a multi-head graph attention network is constructed for feature extraction of graph-like structure data, and evaluation results of different preference models are integrated to enrich feature extraction capability of the models.
The feature extraction part is composed of three multi-head graph attention network layers GAT1, GAT2 and GAT3, Elu is adopted as an activation function, and Elu function expressions are as follows:
the number of multiple GAT1 layers is marked as P1The output characteristic dimension of each node in a single attention head is 64, and the number of multi-head of the GAT2 layer is recorded as P2The output feature dimension of each node in a single attention head is 32. The multi-head outputs of the GAT1 and the GAT2 are directly spliced into adjacent outputs with higher dimension, so that the characteristic dimension of the final output of each node of the GAT1 layer is 64 XP1And the final output characteristic dimension of each node of the GAT2 layer is 32 xP2Expressed as:
the number of multiple heads of the GAT3 layer is P3The output feature dimension of each node in a single attention head is 16. The multiple outputs of GAT3 are aggregated by taking an average value, as shown in equation (6). Therefore, each node of the GAT3 layer finally outputs the feature dimensionIs 16.
In the formulae (5) and (6), xiIs a bus i original characteristic, x'iFor the new characteristic of the bus i, sigma is an activation function; p is the number of multiple heads; wpIs a feature variation matrix; alpha is alphaij,pIs the attention coefficient generated by the bus node j to the bus node i in the attention head p; j ∈ N (i) represents that node j is a neighbor node of node i, which can be obtained by the adjacency matrix A.
Optionally, in step 3, performing pooling operation on output data of the last layer of graph attention network, and inputting the output data into the fully-connected classifier layer to implement stable classification. Namely, in order to avoid the need of repeatedly training the classifier when the number of bus nodes is increased or reduced, a pooling layer is added between the GAT3 layer and the fully-connected classifier layer, and the expression is as follows:
X′=(max{xi},min{xi},mean{xi}) (7)
where X' is the input characteristic of the fully connected layer, max { XiIs the maximum value of each column of features in the GAT3 layer output result, min { x }iIs the minimum value of each column of characteristics in the output result of the GAT3 layer, mean { x }iThe mean value of each column of features in the output result of the GAT3 layer. Because the output characteristic dimension of each node is 16 in the GAT3 layer design, the characteristic dimension is fixed to be 16 multiplied by 3 to 48 after passing through the pooling layer, and cannot be changed according to the change of the number of the bus nodes of the system, and the need of redesigning a classifier and training when the system scale is changed is avoided.
The full-connection classifier layer is composed of two full-connection layers FC1 and FC2, wherein Elu is adopted as an activation function for FC1, Softmax is adopted as an activation function for FC2, the number of neurons in the FC1 layer is 16, and the number of neurons in the FC2 layer is 2.
Optionally, in step 4, the original sample set marked in step 1 is divided into a training set, a verification set, and a test set according to a ratio of 6:2: 2. Model parameters are optimized on a training set through a back propagation algorithm, and in consideration of the fact that the sample category in the transient evaluation of the power system has a significant imbalance problem, the following weighted loss function is adopted as a training target:
wherein m is the number of samples in a single training, yk,0And yk,1Is that sample k belongs to the label of instability and stability, Pk,0And Pk,1The probability value of the instability and the stability of the sample k given by the discriminator is in a value range of [0,1],w0Weight, w, for a sample that is unstable1The weight is the weight in case the sample is stable. In practice, w may be set according to the ratio of the unstable samples to the stable samples in the set of sampled training samples0And w1The proportional value of (c).
When the evaluation accuracy of the verification set reaches the historical highest value in the training process and the value is not updated in the subsequent 200 times of training processes, the training is finished, and meanwhile, the model with the optimal accuracy is stored. And (5) testing the application effect of the stability evaluation model by adopting a test set sample.
Optionally, training is performed on original sample sets of different topologies, so that the graph attention network can perform feature extraction in a self-adaptive manner according to the change of the power grid topology, and effective features are formed so as to perform stable evaluation. And (3) forming a power grid operation state data matrix X by the online acquired power grid real-time data according to a formula (2), constructing an adjacent matrix A according to the power grid real-time operation topology, and obtaining a transient stability evaluation result by using the best evaluation model which is stored in the step (4) and is represented on the verification set.
The invention is further illustrated by the following set of examples.
In the example, the new england 10 machine 39 node system is taken as an example to perform model performance evaluation. The new england 10 machine 39 node contains 10 generators, 39 buses and 46 transmission lines under the full connection condition, as shown in fig. 2.
Corresponding to the step 1, a large number of power system transient process samples under different operating topologies are generated through power system simulation software to serve as an original sample set. The simulation is set as follows: consider a fully wired system and "N-1" and "N-2" modes of operation, the latter two being generated by randomly disconnecting the 1-2 loops of the line. For each operation topology, the load level is changed between 75% and 120% in a step length of 5%, the adjustment amount of the generated power is randomly distributed to different generator sets, and an operation mode set is formed in a mode of keeping the trend convergence. In the aspect of fault setting, three-phase grounding short-circuit faults are set at the beginning and the tail end of each circuit of the power transmission line, and the fault line is cut off by 0.1s of reliable protection after disturbance occurs. The sampling interval of the dynamic data of the system is 0.01s, the simulation time length is 4s, and finally 85593 samples are generated.
The embodiment sample set comprises operation modes under different topologies, and the power grid topology corresponding to the sample is expressed in the form of an adjacency matrix; since the system of the embodiment has 39 buses in total, the adjacency matrix is a symmetric square matrix with the size of 39 × 39, and is taken as one of the characteristics of the sample; the acquisition range is from the moment before the fault occurs to the moment after the fault is removed, and because the sample fault duration is 0.1s, the number of sampling points is 11; the bus voltage amplitude, phase angle, injected active power and reactive power are used as the input of the sample characteristic, namely the model, so that the size of the node characteristic matrix X is 39X 44. A and X in the sample feature G (a, X) are respectively represented as:
in the formula, ai,jA value of 0 or 1, ai,j1 denotes that bus i is connected to bus j, aij0 represents that the bus i is not connected with the bus j, and diag (A) is set to 1; u shapek,yRepresents the kth bus voltage amplitude at the y time of the system; thetak,yRepresents the kth bus voltage phase angle at the y time of the system; pk,yThe active power injected by the kth bus node at the y moment of the system is represented; qk,yPresentation systemReactive power injected by the kth bus node at the y-th moment. If the bus node is neither a generator node nor a load node, Pk,y=Qk,y=0。i,j,k=1,2,...,n;y1,2, t; n is the total number of system bus bars, 39 in the example, and t is the sampling time, 11 in the example. The node numbering sequence has no special rule, and the node numbering sequence in the matrixes A and X only needs to be ensured to be the same.
In order to reduce the influence of the dimension on the model, carrying out the following data standardization processing on a matrix X representing the power grid operation state in the sample set according to columns, wherein the standardization formula is as follows:
in the formula, XyFor the feature sample data collected at the y-th sampling instant,the normalized value of the characteristic sample data collected at the y-th sampling moment,is the mean value of the characteristic sample data acquired at the y-th sampling moment,for the standard deviation of the feature sample data acquired at the y-th sampling instant,yt, where t is 11 in the example, and the features at t times are normalized to obtain a sample feature dataset Xstd。
Carrying out stability labeling on the sample according to a transient stability simulation result, wherein the relative power angle difference of any two generators is larger than 180 degrees, the transient instability of the system is caused, and the sample stability label is [0,1 ]; otherwise, the system transient is stable and the sample label is [1,0 ]. The number of final full-wiring system and "N-1" operation mode samples is 19287, wherein 3224 unstable samples and 16063 stable samples are obtained; 14041 destabilizing samples and 52265 stable samples of the 66306 "N-2" run mode samples.
Corresponding to the step 2, a graph attention network is constructed for feature extraction of graph-like structure data, and single-head GAT layer forward propagation comprises two steps of attention coefficient calculation and feature aggregation, and a schematic diagram is shown in FIG. 3. Attention coefficient alpha generated by bus node j to bus node iijThe calculation formula is as follows:
wherein LeakyReLU is a nonlinear activation function,and W is a parameter of the drawing attention layer, xiFor the input features of the bus node i, k ∈ N (i) indicates that the bus node k is a neighbor node of the node i.
The multi-view attention network can integrate the evaluation results of different preference models to enrich the feature extraction capability of the models, and the schematic view is shown in fig. 4. The feature extraction part is composed of three multi-head graph attention network layers GAT1, GAT2 and GAT3, Elu is adopted as an activation function, and Elu function expressions are as follows:
in the embodiment, the multi-head number P of GAT1 layers14, each node in a single attention head outputs a multi-head number P with the characteristic dimension of 64 and GAT2 layers2Each node in a single attention head outputs a feature dimension of 32, 4. The multi-head outputs of the GAT1 and the GAT2 are directly spliced into adjacent outputs with higher dimension, so that the characteristic dimension of the final output of each node of the GAT1 layer is 64 XP1256, the final output characteristic dimension of each node of GAT2 layer is 32 XP2128, expressed as:
the number of multiple heads of the GAT3 layer is P3Each node in a single attention head outputs a feature dimension of 16, 4. The multiple outputs of GAT3 are aggregated by taking an average value, as shown in equation (6). So each node at GAT3 level eventually outputs a feature dimension of 16.
In the formulae (6) and (7), xiIs a bus i original characteristic, x'iFor the new characteristic of the bus i, sigma is an activation function; p is the number of multiple heads; wpIs a feature variation matrix; alpha is alphaij,pIs the attention coefficient generated by the bus node j to the bus node i in the attention head p; j ∈ N (i) represents that node j is a neighbor node of node i, which can be obtained by the adjacency matrix A.
And corresponding to the step 3, performing pooling operation on the output data of the last layer of graph attention network, and inputting the output data into a full-connection layer classifier to realize stable classification.
In order to avoid the need to train the classifier repeatedly when the system scale changes, i.e. the number of bus nodes increases or decreases, a pooling layer is added between the GAT3 layer and the fully-connected classifier layer, which is expressed as:
X′=(max{xi},min{xi},mean{xi}) (7)
where X' is the input characteristic of the fully connected layer, max { XiIs the maximum value of each column of features in the GAT3 layer output result, min { x }iIs the minimum value of each column of characteristics in the output result of the GAT3 layer, mean { x }iThe mean value of each column of features in the output result of the GAT3 layer. Because the output characteristic dimension of each node is 16 in the GAT3 layer design, the characteristic dimension is fixed to be 16 multiplied by 3 to 48 after passing through the pooling layer, and cannot be changed according to the change of the number of the bus nodes of the system, and the need of redesigning a classifier and training when the system scale is changed is avoided.
The full-connection classifier layer is composed of two full-connection layers FC1 and FC2, wherein Elu is adopted as an activation function for FC1, Softmax is adopted as an activation function for FC2, the number of neurons in the FC1 layer is 16, and the number of neurons in the FC2 layer is 2. The complete evaluation model is finally formed as shown in FIG. 5
Corresponding to the step 4, dividing the original sample set marked in the step 1 into a training set, a verification set and a test set according to the ratio of 6:2: 2. Model parameters are optimized on a training set through a back propagation algorithm, and in consideration of the fact that the sample category in the transient evaluation of the power system has a significant imbalance problem, the following weighted loss function is adopted as a training target:
wherein m is the number of samples in a single training, yk,0And yk,1Is a label that the sample k belongs to instability and stability, and is a Boolean value, Pk,0And Pk,1The probability value of the instability and the stability of the sample k given by the discriminator is in a value range of [0,1],w0Weight, w, for a sample that is unstable1The weight is the weight in case the sample is stable. In an embodiment, the weight w of a destabilized sample0To 7, stabilize w of the sample 11 is taken. When the evaluation accuracy of the verification set reaches the historical highest value in the training process and the value is not updated in the subsequent 200 times of training processes, the training is finished, and meanwhile, the model with the optimal accuracy is stored.
And finally, replacing the real-time operation data of the power grid measured by the PMU with the data of the test set, and inputting the data into the evaluation model to obtain a transient stability evaluation result, thereby measuring the performance of the evaluation model. In the research of solving transient stability evaluation by deep learning, the problem of unbalanced samples can be encountered, that is, in the samples obtained by an actual system or simulation, the number of unstable samples is less than that of stable samples, and meanwhile, in an actual scene, the influence of 'missed judgment' and 'false judgment' on the power system is different. Therefore, for the performance measurement of the model, besides the conventional accuracy ACC, other indexes including a missing judgment rate MA, a misjudgment rate FA and a third-level index G-mean are introduced. As shown in table 1, the following evaluation indexes are defined according to the transient stability confusion matrix:
TABLE 1
The ACC is the primary index for judging the performance of the model, but other indexes need to be considered because the cost of the system is different between the problem of unbalanced samples and the cost of judging unstable samples as stable and the cost of judging stable samples as unstable.
The invention was compared with other deep learning model based evaluation methods, with model input X for the comparison method being the same as the method of the invention, and the results are shown in table 2.
TABLE 2
As can be seen from the table, due to different operation topological modes of the sample centralized system, the method enables the model to extract the transient characteristic difference under different topologies by embedding the adjacency matrix into the model, so that all evaluation indexes of the method are optimal, the accuracy rate is 99.02%, and the rate of missing judgment and the rate of erroneous judgment are near 1%. DNN, SAE and SSAE adopt the same model structure, and the difference is that SAE and SSAE add a pre-training process, and SSAE also adds a sparse penalty term in the pre-training process, so that the optimal effect is obtained in three algorithms, but the accuracy is still 2.95% lower than that of the method of the invention. SVM is comparable to DBN in current samples, but still has a gap compared to the method of the present invention. Although the CNN is also an aggregation model of local features, the CNN cannot learn the topological information of the samples because the topological structure of each sample is different, and therefore the accuracy of the test set does not exceed 95% after the final training.
In summary, in addition to evaluating performance, the present invention has the following three advantages from the practical point of view: according to the method, only time sequence data from the moment before the fault occurs to the moment after the fault is removed are collected and used as original input features, the data sampling time is short, and the method has good feature mining capability; the power grid topology is embedded into the evaluation model in an adjacent matrix mode, so that the difference of transient characteristics under different topologies is effectively considered, and the model mobility is improved; through the pooling layer before the classifier, on one hand, the model can still be suitable under the condition of system scale change, so that the model still keeps the applicability under different system scales without redesigning and training the model when the system scale changes, and on the other hand, the evaluation model is ensured to have more accurate evaluation capability compared with the traditional evaluation model.
The above embodiments are only for illustrating the technical concept and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and implement the present invention accordingly, and not to limit the protection scope of the present invention accordingly. All equivalent changes or modifications made in accordance with the spirit of the present disclosure are intended to be covered by the scope of the present disclosure.
Claims (10)
1. A power system transient stability assessment method based on a graph attention network is characterized by comprising the following steps:
acquiring operation data, wherein the operation data comprises an operation data matrix X constructed by real-time operation data of a power grid measured by a PMU and an adjacency matrix A constructed according to a power grid operation topology;
inputting the operation data into a transient stability evaluation model of the power system for evaluation to obtain a stability result;
the power system transient stability evaluation model is obtained by the following steps:
step 1: generating power system transient process samples under different operating topologies through power system simulation software, and taking the power system transient process samples as an original sample set; forming a graphic structure data G (A, X) by the adjacency matrix A of the power grid topology corresponding to the sample and the power grid operation state data X, and labeling the sample data according to whether the sample data is stable or not;
step 2: constructing a graph attention network for feature extraction of graph-like structure data G (A, X), the graph attention network being composed of a plurality of layers;
and step 3: performing pooling operation on output data of the last layer of graph attention network, and inputting the output data into a full-connection classifier layer to realize stable classification;
and 4, step 4: and (3) dividing the original sample set marked in the step (1) into a training set, a verification set and a test set according to a set proportion so as to obtain the transient stability evaluation model of the power system with the optimal accuracy.
2. The method for evaluating the transient stability of the power system based on the graph attention network as claimed in claim 1, wherein in the step 1, X is organized in units of nodes and comprises node voltage amplitude, phase angle, injected active power and reactive power of each sampling point from a time before the fault occurs to a time after the fault is removed.
3. The graph attention network-based power system transient stability assessment method according to claim 2, wherein a and X in the derived sample features G (a, X) are respectively represented as:
in the formula, ai,jA value of 0 or 1, ai,j1 denotes that bus i is connected to bus j, ai,j0 represents that the bus i is not connected with the bus j, and diag (A) is set to 1; u shapek,yRepresents the kth bus voltage amplitude at the y time of the system; thetak,yRepresents the kth bus voltage phase angle at the y time of the system; pk,yThe active power injected by the kth bus node at the y moment of the system is represented; qk,yThe reactive power injected by the kth bus node at the y moment of the system is represented; if the bus node is neither a generator node nor a load node, Pk,y=Qk,y0; i, j, k ═ 1,2, …, n; y is 1,2, …, t; n is the total number of the system buses, and t is the sampling time.
4. The power system transient stability assessment method based on graph attention network as claimed in claim 2 or 3, characterized in that the matrix X representing the power grid operation state in the sample set is subjected to the following data standardization process according to the following column formula:
in the formula, XyFor the feature sample data collected at the y-th sampling instant,the normalized value of the characteristic sample data collected at the y-th sampling moment,is the mean value of the characteristic sample data acquired at the y-th sampling moment,the standard deviation of the characteristic sample data collected at the y-th sampling moment is 1,2, …, t, and the characteristics at the t momentsObtaining a sample feature data set X after standardizationstd;
Carrying out stability labeling on the sample according to a transient stability simulation result, wherein the relative power angle difference of any two generators is larger than 180 degrees, the transient instability of the system is caused, and the sample stability label is [0,1 ]; otherwise, the transient stability of the system is stable, and the sample stability label is [1,0 ].
5. The method for evaluating transient stability of power system based on graph attention network as claimed in claim 1, wherein in the step 2, the graph attention network is composed of three multi-head graph attention network layers GAT1, GAT2 and GAT3, Elu is adopted as activation function, and Elu function expression is as follows:
6. the graph attention network-based power system transient stability assessment method of claim 5, wherein the multi-headed outputs of GAT1 and GAT2 are directly spliced into adjacent outputs; the number of multiple GAT1 layers is marked as P1The output characteristic dimension of each node in a single attention head is 64, and the number of multi-head of the GAT2 layer is recorded as P2The output feature dimension of each node in a single attention head is 32; the final output characteristic dimension of each node of the GAT1 layer is 64 multiplied by P1And the final output characteristic dimension of each node of the GAT2 layer is 32 xP2Expressed as:
the number of multiple heads of the GAT3 layer is P3Each node in a single attention head outputs a feature dimension of 16; the multi-head output of the GAT3 is aggregated and then output in an averaging mode, and as shown in formula (6), the final output characteristic dimension of each node of the GAT3 layer is 16;
in the formulae (5) and (6), xiIs a bus i original characteristic, x'iFor the new characteristic of the bus i, sigma is an activation function; p is the number of multiple heads; wpIs a feature variation matrix; alpha is alphaij,pIs the attention coefficient generated by the bus node j to the bus node i in the attention head p; j ∈ N (i) represents that node j is a neighbor node of node i, which can be obtained by the adjacency matrix A.
7. The graph attention network-based power system transient stability assessment method according to claim 5 or 6, wherein in said step 3, a pooling layer is added between the GAT3 layer and the fully-connected classifier layer, which is expressed as:
X′=(max{xi},min{xi},mean{xi}) (7)
where X' is the input feature of the fully connected classifier layer, max { XiIs the maximum value of each column of features in the GAT3 layer output result, min { x }iIs the minimum value of each column of characteristics in the output result of the GAT3 layer, mean { x }iThe mean value of each column of features in the output result of the GAT3 layer.
8. The power system transient stability assessment method based on graph attention network as claimed in claim 1, wherein in the step 4, the labeled original sample set is divided into a training set, a verification set and a test set according to a ratio of 6:2: 2; optimizing the model parameters on the training set through a back propagation algorithm, and adopting a weighting loss function as a training target; controlling the training times according to the evaluation accuracy of the verification set, and storing a model with the optimal accuracy; and (5) testing the application effect of the stability evaluation model by using the test set sample.
9. The graph attention network-based power system transient stability assessment method according to claim 8, characterized in that the following weighted loss function is adopted as a training target:
wherein m is the number of samples in a single training, yk,0And yk,1Is that sample k belongs to the label of instability and stability, Pk,0And Pk,1The probability value of the instability and the stability of the sample k given by the discriminator is in a value range of [0,1],w0Weight, w, for a sample that is unstable1The weight is the weight under the condition that the sample is stable;
when the evaluation accuracy of the verification set reaches the historical highest value in the training process and the value is not updated in the subsequent 200 times of training processes, the training is finished, and meanwhile, the model with the optimal accuracy is stored.
10. The graph attention network-based power system transient stability evaluation method according to claim 7, wherein the full-connection classifier layer is composed of two full-connection layers FC1 and FC2, FC1 adopts Elu as an activation function, FC2 adopts Softmax as an activation function, FC1 layer has 16 neurons, and FC2 layer has 2 neurons.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010791074.7A CN111914486B (en) | 2020-08-07 | 2020-08-07 | Power system transient stability evaluation method based on graph attention network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010791074.7A CN111914486B (en) | 2020-08-07 | 2020-08-07 | Power system transient stability evaluation method based on graph attention network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111914486A true CN111914486A (en) | 2020-11-10 |
CN111914486B CN111914486B (en) | 2023-09-15 |
Family
ID=73283725
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010791074.7A Active CN111914486B (en) | 2020-08-07 | 2020-08-07 | Power system transient stability evaluation method based on graph attention network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111914486B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112365091A (en) * | 2020-11-30 | 2021-02-12 | 中国人民解放军国防科技大学 | Radar quantitative precipitation estimation method based on classification node map attention network |
CN112561303A (en) * | 2020-12-10 | 2021-03-26 | 清华大学 | Power system dynamic analysis method based on integrated learning and power grid topological change |
CN112834921A (en) * | 2020-12-23 | 2021-05-25 | 浙江工业大学 | Motor fault diagnosis method based on residual error map attention network |
CN113435575A (en) * | 2021-06-22 | 2021-09-24 | 国网黑龙江省电力有限公司 | Gate graph neural network transient stability evaluation method based on unbalanced data |
CN114006370A (en) * | 2021-10-29 | 2022-02-01 | 中国电力科学研究院有限公司 | Power system transient stability analysis and evaluation method and system |
CN114004155A (en) * | 2021-11-01 | 2022-02-01 | 清华大学 | Transient stability assessment method and device considering topological structure characteristics of power system |
CN114209323A (en) * | 2022-01-21 | 2022-03-22 | 中国科学院计算技术研究所 | Method for recognizing emotion and emotion recognition model based on electroencephalogram data |
CN116167289A (en) * | 2023-04-26 | 2023-05-26 | 南方电网数字电网研究院有限公司 | Power grid operation scene generation method and device, computer equipment and storage medium |
CN117370934A (en) * | 2023-12-04 | 2024-01-09 | 环球数科集团有限公司 | Multi-mode data enhancement method of sensitive information discovery model |
CN117609824A (en) * | 2023-11-09 | 2024-02-27 | 武汉华源电力设计院有限公司 | Active power distribution network topology identification and fault diagnosis analysis method, device and equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103995734A (en) * | 2014-06-18 | 2014-08-20 | 中国南方电网有限责任公司电网技术研究中心 | Electric power system hybrid real-time simulation system and simulation method based on RTDS |
CN105098763A (en) * | 2014-05-16 | 2015-11-25 | 国家电网公司 | Transient voltage stable on-line assessment method based on WAMS and EMS |
CN107132772A (en) * | 2017-04-21 | 2017-09-05 | 南方电网科学研究院有限责任公司 | The real-time emulation system and method for alternating current-direct current power network |
CN110676841A (en) * | 2019-09-12 | 2020-01-10 | 天津大学 | Transient stability analysis method for power electronic power system based on direct method |
-
2020
- 2020-08-07 CN CN202010791074.7A patent/CN111914486B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105098763A (en) * | 2014-05-16 | 2015-11-25 | 国家电网公司 | Transient voltage stable on-line assessment method based on WAMS and EMS |
CN103995734A (en) * | 2014-06-18 | 2014-08-20 | 中国南方电网有限责任公司电网技术研究中心 | Electric power system hybrid real-time simulation system and simulation method based on RTDS |
CN107132772A (en) * | 2017-04-21 | 2017-09-05 | 南方电网科学研究院有限责任公司 | The real-time emulation system and method for alternating current-direct current power network |
CN110676841A (en) * | 2019-09-12 | 2020-01-10 | 天津大学 | Transient stability analysis method for power electronic power system based on direct method |
Non-Patent Citations (1)
Title |
---|
钟智等: "《基于图注意力深度网络的电力系统暂态稳定评估》", 《电网技术》, vol. 45, no. 6, pages 2122 - 2130 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112365091B (en) * | 2020-11-30 | 2022-06-14 | 中国人民解放军国防科技大学 | Radar quantitative precipitation estimation method based on classification node map attention network |
CN112365091A (en) * | 2020-11-30 | 2021-02-12 | 中国人民解放军国防科技大学 | Radar quantitative precipitation estimation method based on classification node map attention network |
CN112561303A (en) * | 2020-12-10 | 2021-03-26 | 清华大学 | Power system dynamic analysis method based on integrated learning and power grid topological change |
CN112834921A (en) * | 2020-12-23 | 2021-05-25 | 浙江工业大学 | Motor fault diagnosis method based on residual error map attention network |
CN112834921B (en) * | 2020-12-23 | 2022-07-15 | 浙江工业大学 | Motor fault diagnosis method based on residual error map attention network |
CN113435575A (en) * | 2021-06-22 | 2021-09-24 | 国网黑龙江省电力有限公司 | Gate graph neural network transient stability evaluation method based on unbalanced data |
CN113435575B (en) * | 2021-06-22 | 2022-07-29 | 国网黑龙江省电力有限公司 | Gate graph neural network transient stability evaluation method based on unbalanced data |
CN114006370A (en) * | 2021-10-29 | 2022-02-01 | 中国电力科学研究院有限公司 | Power system transient stability analysis and evaluation method and system |
CN114006370B (en) * | 2021-10-29 | 2024-06-18 | 中国电力科学研究院有限公司 | Power system transient stability analysis and evaluation method and system |
CN114004155A (en) * | 2021-11-01 | 2022-02-01 | 清华大学 | Transient stability assessment method and device considering topological structure characteristics of power system |
CN114004155B (en) * | 2021-11-01 | 2024-04-12 | 清华大学 | Transient stability evaluation method and device considering topological structure characteristics of power system |
CN114209323A (en) * | 2022-01-21 | 2022-03-22 | 中国科学院计算技术研究所 | Method for recognizing emotion and emotion recognition model based on electroencephalogram data |
CN114209323B (en) * | 2022-01-21 | 2024-05-10 | 中国科学院计算技术研究所 | Method for identifying emotion and emotion identification model based on electroencephalogram data |
CN116167289A (en) * | 2023-04-26 | 2023-05-26 | 南方电网数字电网研究院有限公司 | Power grid operation scene generation method and device, computer equipment and storage medium |
CN116167289B (en) * | 2023-04-26 | 2023-09-15 | 南方电网数字电网研究院有限公司 | Power grid operation scene generation method and device, computer equipment and storage medium |
CN117609824A (en) * | 2023-11-09 | 2024-02-27 | 武汉华源电力设计院有限公司 | Active power distribution network topology identification and fault diagnosis analysis method, device and equipment |
CN117609824B (en) * | 2023-11-09 | 2024-05-07 | 武汉华源电力设计院有限公司 | Active power distribution network topology identification and fault diagnosis analysis method, device and equipment |
CN117370934A (en) * | 2023-12-04 | 2024-01-09 | 环球数科集团有限公司 | Multi-mode data enhancement method of sensitive information discovery model |
CN117370934B (en) * | 2023-12-04 | 2024-03-22 | 环球数科集团有限公司 | Multi-mode data enhancement method of sensitive information discovery model |
Also Published As
Publication number | Publication date |
---|---|
CN111914486B (en) | 2023-09-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111914486A (en) | Power system transient stability evaluation method based on graph attention network | |
CN111443259A (en) | Active power distribution network fault diagnosis method and system based on local abnormal factor detection | |
CN110417011B (en) | Online dynamic security assessment method based on mutual information and iterative random forest | |
CN111478314B (en) | Transient stability evaluation method for power system | |
CN110994604B (en) | Power system transient stability assessment method based on LSTM-DNN model | |
CN103245881A (en) | Power distribution network fault analyzing method and device based on tidal current distribution characteristics | |
CN111415059B (en) | Practical model machine construction and online application method | |
CN112241836B (en) | Virtual load leading parameter identification method based on incremental learning | |
CN110705831A (en) | Power angle instability mode pre-judgment model construction method after power system fault and application thereof | |
CN111725802B (en) | Method for judging transient stability of alternating current-direct current hybrid power grid based on deep neural network | |
CN114004155A (en) | Transient stability assessment method and device considering topological structure characteristics of power system | |
CN112818525A (en) | Hydroelectric generating set state evaluation method and system | |
CN112821424A (en) | Power system frequency response analysis method based on data-model fusion drive | |
CN112926686B (en) | BRB and LSTM model-based power consumption anomaly detection method and device for big power data | |
CN108805419B (en) | Power grid node importance calculation method based on network embedding and support vector regression | |
CN114385403A (en) | Distributed cooperative fault diagnosis method based on double-layer knowledge graph framework | |
CN116029618B (en) | Dynamic safety partition assessment method and system for power system | |
CN111965442A (en) | Energy internet fault diagnosis method and device under digital twin environment | |
CN115659844B (en) | Simulation method and device based on wind power plant dynamic model and computer equipment | |
CN113949079B (en) | Power distribution station user three-phase unbalance prediction optimization method based on deep learning | |
CN110956010B (en) | Large-scale new energy access power grid stability identification method based on gradient lifting tree | |
CN115409317A (en) | Transformer area line loss detection method and device based on feature selection and machine learning | |
CN116068303A (en) | Private capacity-increasing on-line monitoring method for special transformer based on data driving | |
CN116204771A (en) | Power system transient stability key feature selection method, device and product | |
CN113111588A (en) | NO of gas turbineXEmission concentration prediction method and device |
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 |