CN111725802B - Method for judging transient stability of alternating current-direct current hybrid power grid based on deep neural network - Google Patents

Method for judging transient stability of alternating current-direct current hybrid power grid based on deep neural network Download PDF

Info

Publication number
CN111725802B
CN111725802B CN202010493766.3A CN202010493766A CN111725802B CN 111725802 B CN111725802 B CN 111725802B CN 202010493766 A CN202010493766 A CN 202010493766A CN 111725802 B CN111725802 B CN 111725802B
Authority
CN
China
Prior art keywords
power grid
neural network
direct current
deep neural
sample
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
Application number
CN202010493766.3A
Other languages
Chinese (zh)
Other versions
CN111725802A (en
Inventor
张业宇
郭剑波
曾平良
马士聪
赵兵
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Electric Power Research Institute Co Ltd CEPRI
Hangzhou Dianzi University
State Grid Hubei Electric Power Co Ltd
Original Assignee
China Electric Power Research Institute Co Ltd CEPRI
Hangzhou Dianzi University
State Grid Hubei Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by China Electric Power Research Institute Co Ltd CEPRI, Hangzhou Dianzi University, State Grid Hubei Electric Power Co Ltd filed Critical China Electric Power Research Institute Co Ltd CEPRI
Priority to CN202010493766.3A priority Critical patent/CN111725802B/en
Publication of CN111725802A publication Critical patent/CN111725802A/en
Application granted granted Critical
Publication of CN111725802B publication Critical patent/CN111725802B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • 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]

Landscapes

  • Engineering & Computer Science (AREA)
  • Power Engineering (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention relates to an alternating current-direct current hybrid power grid transient stability judgment method based on a deep neural network, and belongs to the field of power grid transient stability evaluation. The modern power grid has new characteristics of information physical fusion, complex interconnection of a large power grid and the like, and the transient stability evaluation method of the power system is greatly influenced. In order to adapt to new characteristics of a future power grid, the invention introduces a deep learning method into the transient stability judgment of the power system; a transient sample data set capable of reflecting characteristics of the alternating current and direct current network system is obtained through simulation, and a mapping relation between the characteristic data set and a stable result is trained by utilizing a deep learning framework. Compared with a common shallow learning method supporting a vector machine and a decision tree, the technical scheme adopted by the invention can quickly realize transient stability evaluation, and has higher evaluation accuracy and stronger generalization capability.

Description

Method for judging transient stability of alternating current-direct current hybrid power grid based on deep neural network
Technical Field
The invention belongs to the field of power system transient stability judgment, and particularly relates to an alternating current-direct current hybrid power grid transient stability judgment method based on a deep neural network.
Background
With the rapid advance of smart grid construction, a smart grid will present new characteristics of power electronization, information physical fusion, complex interconnection of a large power grid and the like in the future, which brings great challenges to Transient Stability Assessment (TSA) work of a future power system. With the rapid rise of artificial intelligence technology, the method for judging the transient stability of the power system based on machine learning and deep learning technology develops a new way from the aspect of pattern recognition and gradually enters the field of vision of people, so that the method is considered as one of new methods and key technologies for judging the transient stability of the power system in the future by people.
Due to the characteristics of complex and variable running modes, high dimensionality, nonlinearity and the like of the alternating-current and direct-current hybrid power grid, the processing capability of most shallow learning methods for the running characteristics of the power grid is slightly insufficient, and the generalization capability of the methods is greatly restricted when the complex transient state judgment classification problem is processed. In recent years, with the leap-type development of a deep learning technology (DL), a new idea is provided for power grid transient stability judgment research, that is, by building a learning model of a plurality of hidden layers, more useful characteristics are learned through a large amount of power grid data, so that the accuracy and generalization capability of power system transient stability judgment are improved.
Disclosure of Invention
In order to solve the problems, the invention provides a method for judging the transient stability of an alternating current-direct current hybrid power grid based on a Deep Neural Network (DNN).
The invention comprises the following steps:
the method comprises the following steps: the method comprises the following steps of constructing an alternating current-direct current hybrid power grid system and acquiring a transient original data set of the system, wherein the method comprises the following steps:
a39-node testing system of a new England 10 machine is adopted as a system prototype, a direct current model with two ends is added at a 16-node position of an original system, and the direct current power is arranged to be 1000MW, so that an alternating current-direct current series-parallel power grid system is formed.
The failure information setting is mainly considered from the following three aspects: load level of the system, duration of the fault, location of the fault occurrence. Respectively selecting and selecting 6 load levels of 80%, 90%, 100%, 110%, 120% and 130%, respectively, selecting 6 load levels of 4 cycles, 5 cycles, 6 cycles, 7 cycles, 8 cycles and 10 cycles, respectively, selecting 3 load levels of 0%, 50% and 80% of the positions of the fault, respectively, simulating the system by adopting PSD-BPA, and obtaining a transient original data set, wherein the positions of the fault are respectively 3 positions of 0%, 50% and 80% of the line, the simulation duration is 200 cycles, and the frequency is 50 Hz.
Step two: the sample data set is specifically described as follows:
the transient state sample data set of the alternating current-direct current hybrid power grid mainly comprises two parts: sample set X and label set Y. Taking a sample set X as the input of the deep neural network model, taking a label set Y as the target of model training, and simply describing X and Y as follows:
Figure BDA0002522023470000021
in the formula (1), the row vector is regarded as the number of transient samples, the column vector is regarded as the number of features of each sample, namely m samples in total, and each sample has n feature attributes and is used as the input of the deep neural network model.
The tag set Y considers two cases of stability and instability, namely Y has 2 categories which are specifically described as follows:
Figure BDA0002522023470000022
in the formula (2) [ y1,y2,y3……ym]Are all row vectors, and adopt One-Hot coding, i.e. y ═ 1,0]Represents a stabilizing label, y ═ 0,1]Representing a destabilization tag.
Step three: the selection of the transient sample characteristics of the system is specifically described as follows:
the key step of the deep neural network model is the selection of sample characteristic variables, and usually, the characteristics which have definite physical meanings and represent the running state of the system are selected. Dynamic features and static features; in terms of space, the generator parameter characteristics and the grid parameter characteristics can be considered; the characteristic parameters of the generator cover common factors influencing stability, such as a power angle, a rotor speed, an angular acceleration, a rotor kinetic energy and the like; the power grid mainly comprises total output, total load, bus voltage and the like; in addition, two characteristic variables of direct-current voltage and direct-current power are additionally added aiming at the transient stability of the alternating-current and direct-current series-parallel power grid system. Specific transient characteristics were selected as shown in table 1 below.
TABLE 1 sample feature selection
Figure BDA0002522023470000023
Figure BDA0002522023470000031
Step four: training sample data by using the built deep neural network model, wherein the specific description is as follows:
(1) inputting sample characteristic data into a deep neural network model, enabling the sample data to reach a first hidden layer, and obtaining an output result of the first hidden layer through function weighted transformation as follows:
g1＝f(w1x+b1) (3)
in the formula (3), f is an Activation Function (Activation Function) of the hidden layer, w1Is a weight matrix, b1Is an offset, g1Is the output result of the hidden layer.
Similarly, the output of the i-th hidden layer is:
gi＝f(wigi-1+bi) (4)
the activation functions of all hidden layers in the deep neural network model adopt Relu activation functions, and the specific mathematical description is as follows:
Figure BDA0002522023470000041
(2) after passing through a plurality of hidden layers, the output result of the deep neural network model is as follows:
y＝f'(wngn-1+bn) (6)
wherein wn、bnRespectively weight matrix and bias matrix, gn-1Is the output result of the previous layer, and f' is the activation function of the output layer, where the softmax function is used, and the specific mathematical description is as follows:
Figure BDA0002522023470000042
the outputs in equation (7) are all between 0 and 1, and their sum is 1, being the probability of each class respectively.
(3) The deep neural network model is essentially a classification model, and the loss function adopts a cross entropy (cross entropy) function, and the specific mathematical description is as follows:
Figure BDA0002522023470000043
in formula (8)
Figure BDA0002522023470000044
And for the actual output of the network, y is a label of a real result, J is used as an error measurement standard and represents the error between the output of the model and the real label, and an Adam algorithm is adopted for training to reduce the error to the minimum.
Step five: and (4) inputting the original data obtained in the step one into the deep neural network model obtained in the step four for training and testing through the processing and feature selection in the step two and the step three, and obtaining the result and the accuracy of the transient stability judgment of the alternating current-direct current series-parallel power grid.
The invention has the beneficial effects that: according to the method, the deep learning technology and the transient stability judgment of the AC/DC hybrid power grid are combined according to the advantages that the deep neural network has strong capability of processing high-dimensional data and can mine deeper data characteristics, the complex interconnection of the large power grid data and the high data dimension are better processed, the power grid data can be fully trained, the accuracy of the transient judgment of the AC/DC hybrid power grid can be improved, and the method has good robustness and generalization capability, so that the rapid transient stability judgment of the AC/DC hybrid power grid is realized.
Drawings
FIG. 1 is a flow chart of transient stability determination according to the present invention;
FIG. 2 is a diagram of the AC/DC grid topology of the present invention;
FIG. 3 is a diagram of the Deep Neural Network (DNN) architecture of the present invention;
FIG. 4 is a graphical illustration of the deep neural network training accuracy of the present invention;
FIG. 5 is a schematic diagram of the deep neural network training error of the present invention;
FIG. 6 is a schematic diagram of a confusion matrix for transient stability determination according to the present invention.
Detailed Description
The technical scheme of the invention is clearly and completely described in the following by combining the drawings and the specific embodiments of the specification.
Fig. 1 is a flow chart for transient determination of the present invention, and fig. 2 shows a topological structure of an ac/dc hybrid power grid adopted in the present invention, where the system has 10 generators, 39 buses and 46 lines in total, and the present invention arranges that the active power output of the generators is 6158MW and the active power load of the system is 6098.5 MW. A total of 4968 samples were generated from software simulations with PSD-BPA, with 4150 stable samples and 818 unstable samples. The data normalization processing is carried out on 4968 data samples, the z-score method is adopted by the invention, and the mathematical description is as follows:
Figure BDA0002522023470000051
in the formula (1), m and σ are the mean value and standard deviation of the data, x,
Figure BDA0002522023470000052
Respectively before and after processing, passing throughThe sample data after processing has the characteristics of mean 0 and variance 1.
4968 sample data are randomly divided into a training set and a test set, the proportion of the training set is 0.7, the proportion of the test set is 0.3, the proportion of stable and unstable samples in the data set is kept unchanged, the training set is used for training a DNN judgment model, and the test set is used for judging the performance of the model.
The method comprises the following steps: the method comprises the following steps of constructing an alternating current-direct current hybrid power grid system and acquiring a transient original data set of the system, wherein the method comprises the following steps:
a39-node testing system of a new England 10 machine is adopted as a system prototype, a direct current model with two ends is added at a 16-node position of an original system, and the direct current power is arranged to be 1000MW, so that an alternating current-direct current series-parallel power grid system is formed.
The failure information setting is mainly considered from the following three aspects: load level of the system, duration of the fault, location of the fault occurrence. Respectively selecting 6 load levels of 80%, 90%, 100%, 110%, 120% and 130%, respectively, selecting 6 duration times of 4 cycles, 5 cycles, 6 cycles, 7 cycles, 8 cycles and 10 cycles, respectively, and 3 positions of 0%, 50% and 80% of the line at the fault occurrence positions, simulating the system by adopting PSD-BPA, wherein the simulation time length is 200 cycles, the frequency is 50Hz, and acquiring a transient original data set.
Step two: the sample data set is specifically described as follows:
the transient state sample data set of the alternating current-direct current hybrid power grid mainly comprises two parts: sample set X and label set Y. The sample set X is used as the input of the deep neural network model, and the label set Y is used as the target of the deep neural network model training, and X and Y can be described simply as follows:
Figure BDA0002522023470000061
in the formula (1), the row vector is regarded as the number of transient samples, the column vector is regarded as the number of features of each sample, namely m samples in total, and each sample has n feature attributes and is used as the input of a judgment model.
The tag set Y considers two cases of stability and instability, namely Y has 2 categories which are specifically described as follows:
Figure BDA0002522023470000062
in the formula (2) [ y1,y2,y3……ym]All are line vectors, and adopt One-Hot coding, m is the corresponding number of samples, i.e. y is [1,0 ]]Represents a stabilizing label, y ═ 0,1]Representing a destabilization tag.
Step three: the selection of the transient sample characteristics of the system is specifically described as follows:
the key step of the deep neural network model is the selection of sample characteristic variables, and usually, the characteristics which have definite physical meanings and represent the running state of the system are selected. Dynamic features and static features; in terms of space, the generator parameter characteristics and the grid parameter characteristics can be considered; the characteristic parameters of the generator cover common factors influencing stability such as a power angle, a rotor speed, an angular acceleration, rotor kinetic energy and the like, and the power grid mainly comprises total output, total load, bus voltage and the like; in addition, the invention relates to the research of transient stability of an alternating current-direct current hybrid power grid system, and two characteristic variables of direct current voltage and direct current power are additionally added. Specific transient characteristics were selected as shown in table 1 below.
T0 in table 1 represents the time of stable operation before the occurrence of the fault, t1 is the time of the occurrence of the fault, t2 is the time before the removal of the fault, t3 is the time after the removal of the fault, in order to prevent sudden change of the system operation state, it is considered that the 15 th, 25 th, 35 th, 45 th, 55 th, 85 th, 125 th and 200 th cycles after the clearing of the fault are 8 times, the characteristic attribute is consistent with the time of t3, and in order to prevent the occurrence of the later sudden change, the selection span of the characteristic time point is large. Part of the constructed sample feature set is directly obtained by simulation software, and part of the features is obtained by calculation of simulation data.
Table 2 sample feature selection
Figure BDA0002522023470000063
Figure BDA0002522023470000071
Step four: training sample data by using a built deep neural network model (DNN), as shown in fig. 3, specifically described as follows:
(1) inputting sample characteristic data into a deep neural network model, enabling the sample data to reach a first hidden layer, and obtaining an output result of the first hidden layer through function weighted transformation as follows:
g1＝f(w1x+b1) (4)
in the formula (3), f is an Activation Function (Activation Function) of the hidden layer, w1Is a weight matrix, b1Is an offset, g1Is the output result of the hidden layer.
Similarly, the output of the i-th hidden layer is:
gi＝f(wigi-1+bi) (5)
the activation functions of all hidden layers in the deep neural network model adopt Relu activation functions, and the specific mathematical description is as follows:
Figure BDA0002522023470000072
(2) after passing through multiple hidden layers, the output results of the obtained model are as follows:
y＝f'(wngn-1+bn) (7)
wherein wn、bnRespectively weight matrix and bias matrix, gn-1Is the output result of the previous layer, and f' is the activation function of the output layer, where the softmax function is used, and the specific mathematical description is as follows:
Figure BDA0002522023470000081
the outputs in equation (7) are all between 0 and 1, and their sum is 1, being the probability of each class respectively.
(3) The deep neural network model is essentially a classification model, and the loss function adopts a cross entropy (cross entropy) function, and the specific mathematical description is as follows:
Figure BDA0002522023470000082
in formula (8)
Figure BDA0002522023470000083
And for the actual output of the network, y is a label of a real result, J is used as an error measurement standard and represents the error between the output of the model and the real label, and an Adam algorithm is adopted for training to reduce the error to the minimum.
Step five: and (3) inputting the original data obtained in the first step into the deep neural network model obtained in the fourth step for training and performance testing through the processing and feature selection of the second step and the third step, and comparing the result with other shallow learning methods, wherein the result is shown in table 1.
TABLE 1 transient stability determination results
Figure BDA0002522023470000084
According to the table 1, it can be seen that DNN has an advantage in determining accuracy compared with other methods, and the DNN can well mine deep features of a sample set both in a training set and a test set, and accurately and quickly complete training and determining of a model. It can be seen that the judgment accuracy of the test set of DNN is 4.43%, 1.35% and 1.63% higher than that of PCA + SVM, PCA + SVM + optimization and Decision Tree (DT). The training process of the model is shown in fig. 4 below. As can be seen from fig. 4, the DNN training process converges quickly, the curve of the verification set is relatively stable, and the determination accuracy is kept high. Fig. 5 is a graph of the training error of the DNN model, from which it can be seen that the error converges faster and eventually stays at a lower error level.
To better see the process of DNN model decision, fig. 6 below shows a confusion matrix for a certain DNN decision. The abscissa of the confusion matrix is Target Class, namely the actual label of the sample, and the ordinate is Output Class, namely the actual judgment result of the DNN model. As can be seen from fig. 6, the total of 8 samples have a judgment error, and compared with the test set sample population, the samples having the judgment error are fewer, and the judgment performance of the DNN model population is superior.
Finally, the above embodiments are only intended to illustrate the technical solutions of the present invention and not to limit the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, and all of them should be covered by the claims of the present invention.

Claims (4)

1. The method for judging the transient stability of the alternating current-direct current hybrid power grid based on the deep neural network is characterized by comprising the following steps of:
the method comprises the following steps: building AC-DC hybrid power grid system and acquiring system transient original data set
The method comprises the following steps of (1) adopting a new England 10 machine 39-node testing system as a system prototype, adding a direct current model at two ends at a 16-node position of the system prototype, arranging the direct current power to be 1000MW, and forming an alternating current-direct current series-parallel power grid system;
the following three aspects are considered when setting the fault information: load level of the system, duration of the fault, location of the fault occurrence; simulating the system by adopting PSD-BPA, wherein the simulation time length is 200 cycles, the frequency is 50Hz, and a transient original data set is obtained;
step two: describing a sample data set
The transient state sample data set of the alternating current-direct current hybrid power grid mainly comprises two parts: a sample set X and a label set Y; taking a sample set X as the input of the deep neural network model, taking a label set Y as the target of deep neural network model training, and describing X and Y as follows:
Figure FDA0003200610930000011
in the formula (1), the row vector is the number of transient samples, the column vector is the number of features of each sample, namely m samples in total, and each sample has n feature attributes;
the tag set Y considers two cases of stability and instability, namely Y has 2 categories which are specifically described as follows:
Figure FDA0003200610930000012
in the formula (2) [ y1,y2,y3……ym]Are all row vectors, and adopt One-Hot coding, i.e. y ═ 1,0]Represents a stabilizing label, y ═ 0,1]Represents a destabilization label;
step three: selecting system transient sample features
Selecting sample characteristics including the power angle, the rotor speed, the angular acceleration and the rotor kinetic energy of the generator; the total output, total load, bus voltage, direct current voltage and direct current power of the power grid;
step four: training sample data by adopting deep neural network model
(1) Inputting sample characteristic data into a deep neural network model, enabling the sample characteristic data to reach a first-layer hidden layer, and obtaining an output result of the first-layer hidden layer through function weighted transformation as follows:
g1＝f(w1x+b1) (3)
in formula (3), f is the activation function of the first hidden layer, w1Is the weight matrix of the first layer hidden layer, b1Is the bias matrix of the first layer hidden layer, g1Is the output result of the first layer hidden layer, x is the sample feature data;
similarly, the output of the i-th hidden layer is:
gi＝f(wigi-1+bi) (4)
wherein wiIs the weight matrix of the i-th hidden layer, biIs the bias matrix of the i-th hidden layer, gi-1Is the output result of the hidden layer of the (i-1) th layer;
the activation function expressions for all hidden layers are:
Figure FDA0003200610930000021
(2) after passing through a plurality of hidden layers, obtaining an output result of the deep neural network model:
y＝f'(wngn-1+bn) (5)
wherein wn、bnWeight matrix and bias matrix of the n-th hidden layer, gn-1The output result of the previous layer is obtained, f' is an activation function of the output layer, and a softmax function is adopted, the output of the softmax function is between 0 and 1, the sum of the softmax function is 1, and the softmax function is the probability of each category;
step five: and (4) inputting the original data obtained in the step one into the deep neural network model obtained in the step four through the processing of the step two and the step three to train and test, and obtaining the result of judging the transient stability of the alternating current-direct current hybrid power grid.
2. The alternating current-direct current hybrid power grid transient stability judgment method based on the deep neural network according to claim 1, wherein the method comprises the following steps: the load levels of the system in the step 1 are respectively selected from 6 load levels of 80%, 90%, 100%, 110%, 120% and 130%.
3. The alternating current-direct current hybrid power grid transient stability judgment method based on the deep neural network according to claim 1, wherein the method comprises the following steps: the failure durations in step 1 are 6 durations of 4 cycles, 5 cycles, 6 cycles, 7 cycles, 8 cycles and 10 cycles.
4. The alternating current-direct current hybrid power grid transient stability judgment method based on the deep neural network according to claim 1, wherein the method comprises the following steps: the positions of the fault occurrence in step 1 are respectively 0%, 50% and 80% of the line, and are 3 positions.
CN202010493766.3A 2020-06-03 2020-06-03 Method for judging transient stability of alternating current-direct current hybrid power grid based on deep neural network Active CN111725802B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010493766.3A CN111725802B (en) 2020-06-03 2020-06-03 Method for judging transient stability of alternating current-direct current hybrid power grid based on deep neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010493766.3A CN111725802B (en) 2020-06-03 2020-06-03 Method for judging transient stability of alternating current-direct current hybrid power grid based on deep neural network

Publications (2)

Publication Number Publication Date
CN111725802A CN111725802A (en) 2020-09-29
CN111725802B true CN111725802B (en) 2021-10-08

Family

ID=72565650

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010493766.3A Active CN111725802B (en) 2020-06-03 2020-06-03 Method for judging transient stability of alternating current-direct current hybrid power grid based on deep neural network

Country Status (1)

Country Link
CN (1) CN111725802B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112583039B (en) * 2020-12-21 2022-11-15 中国科学院电工研究所 Data-model combined driven AC/DC power distribution and utilization system coordination control method
CN113852100A (en) * 2021-09-24 2021-12-28 广东电网有限责任公司 TCPST emergency control starting criterion generation method based on data driving
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

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108832619A (en) * 2018-05-29 2018-11-16 北京交通大学 Transient stability evaluation in power system method based on convolutional neural networks
CN109783997A (en) * 2019-03-12 2019-05-21 华北电力大学 A kind of transient stability evaluation in power system method based on deep neural network
CN110994604A (en) * 2019-12-12 2020-04-10 南京理工大学 Electric power system transient stability evaluation method based on LSTM-DNN model
CN111193260A (en) * 2020-01-16 2020-05-22 清华大学 Power system transient stability automatic evaluation method of self-adaptive expansion data

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9842585B2 (en) * 2013-03-11 2017-12-12 Microsoft Technology Licensing, Llc Multilingual deep neural network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108832619A (en) * 2018-05-29 2018-11-16 北京交通大学 Transient stability evaluation in power system method based on convolutional neural networks
CN109783997A (en) * 2019-03-12 2019-05-21 华北电力大学 A kind of transient stability evaluation in power system method based on deep neural network
CN110994604A (en) * 2019-12-12 2020-04-10 南京理工大学 Electric power system transient stability evaluation method based on LSTM-DNN model
CN111193260A (en) * 2020-01-16 2020-05-22 清华大学 Power system transient stability automatic evaluation method of self-adaptive expansion data

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Deep Learning Based Transient Stability Assessment for Grid-Connected Inverter;Xianghua Yu等;《 2018 IEEE International Power Electronics and Application Conference and Exposition(PEAC)》;20181227;全文 *
基于深度置信网络的电力系统暂态稳定评估方法;朱乔木等;《中国电机工程学报》;20180205;第38卷(第3期);全文 *

Also Published As

Publication number Publication date
CN111725802A (en) 2020-09-29

Similar Documents

Publication Publication Date Title
CN111725802B (en) Method for judging transient stability of alternating current-direct current hybrid power grid based on deep neural network
CN108551167B (en) XGboost algorithm-based power system transient stability discrimination method
CN111709555B (en) Method and system for optimal configuration of distributed power supply
CN111914486A (en) Power system transient stability evaluation method based on graph attention network
CN113659565B (en) Online prediction method for frequency situation of new energy power system
Lin et al. Parametric fault diagnosis based on fuzzy cerebellar model neural networks
CN109255726A (en) A kind of ultra-short term wind power prediction method of Hybrid Intelligent Technology
CN112200694A (en) Dominant instability mode identification model construction and application method based on graph neural network
CN115017787A (en) Wind power plant voltage ride through characteristic equivalent modeling method and system based on intelligent algorithm
CN111478314A (en) Transient stability assessment method for power system
CN111652478B (en) Umbrella algorithm-based power system voltage stability evaluation misclassification constraint method
CN110634082A (en) Low-frequency load shedding system operation stage prediction method based on deep learning
CN114006370A (en) Power system transient stability analysis and evaluation method and system
CN111401792A (en) Dynamic safety assessment method based on extreme gradient lifting decision tree
Jia et al. Voltage stability constrained operation optimization: An ensemble sparse oblique regression tree method
CN114266396A (en) Transient stability discrimination method based on intelligent screening of power grid characteristics
Wang et al. Dynamic multiobjective squirrel search algorithm based on decomposition with evolutionary direction prediction and bidirectional memory populations
CN112465136A (en) Hydroelectric generating set degradation prediction method and system
CN116305683A (en) Power system transient stability evaluation method and system based on sample equalization
CN114157552B (en) Distribution network fault detection method based on twin time sequence diagram network
CN116204771A (en) Power system transient stability key feature selection method, device and product
Karapidakis Machine learning for frequency estimation of power systems
CN111814394B (en) Power system safety assessment method based on correlation and redundancy detection
CN113435575B (en) Gate graph neural network transient stability evaluation method based on unbalanced data
Li et al. Visualization of CNN transient voltage classification based on feature recognition and enhancement

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