CN110414412A - The multiple disturbance precise recognition method of the Wide Area Power based on big data analysis and device - Google Patents
The multiple disturbance precise recognition method of the Wide Area Power based on big data analysis and device Download PDFInfo
- Publication number
- CN110414412A CN110414412A CN201910674207.XA CN201910674207A CN110414412A CN 110414412 A CN110414412 A CN 110414412A CN 201910674207 A CN201910674207 A CN 201910674207A CN 110414412 A CN110414412 A CN 110414412A
- Authority
- CN
- China
- Prior art keywords
- disturbance
- big data
- power grid
- layer
- features
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000007405 data analysis Methods 0.000 title claims abstract description 23
- 230000001052 transient effect Effects 0.000 claims abstract description 45
- 238000001914 filtration Methods 0.000 claims abstract description 44
- 230000008878 coupling Effects 0.000 claims abstract description 35
- 238000010168 coupling process Methods 0.000 claims abstract description 35
- 238000005859 coupling reaction Methods 0.000 claims abstract description 35
- 238000000926 separation method Methods 0.000 claims abstract description 20
- 238000013135 deep learning Methods 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 238000005562 fading Methods 0.000 claims description 27
- 238000005259 measurement Methods 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 7
- 238000012544 monitoring process Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 239000011475 Accrington brick Substances 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 239000000203 mixture Substances 0.000 abstract 1
- 238000011897 real-time detection Methods 0.000 abstract 1
- 230000035772 mutation Effects 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 230000010355 oscillation Effects 0.000 description 4
- 230000008447 perception Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 230000002087 whitening effect Effects 0.000 description 1
Classifications
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Remote Monitoring And Control Of Power-Distribution Networks (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
Abstract
The invention proposes a kind of multiple disturbance precise recognition method of the Wide Area Power based on big data analysis and devices, this method comprises: obtaining disturbance transient state big data；General characteristics are extracted, using filtering residuals as test statistics, real-time detection grid disturbance occurs；The depth network for extracting depth perturbation features is constructed, it is the stack self-encoding encoder with specific structure that the depth network, which is arranged,；By general characteristics and disturbance transient state big data together as the input of depth network, is successively trained, blend depth perturbation features and general characteristics, extract fused coupling feature, decoupling separation is carried out to it, is separated into different single disturbance types, completes the identification of disturbance type.The device includes data module, general characteristics module, depth network settings module, deep learning module and integrated separation module.This method and device can detect in time grid disturbance and recognize the type of multiple disturbance.
Description
Technical Field
The invention mainly relates to the technical field of power grid operation maintenance, in particular to a method and a device for accurately identifying multiple disturbances of a wide-area power grid based on big data analysis.
Background
The power failure caused by the power grid fault can cause serious loss to social life and production, and the timely monitoring of the power grid fault is undoubtedly the primary factor for solving or preventing the power grid fault. With the cross-regional interconnection of electric power systems in China, the power grid is larger and larger, and the structure is more and more complex. The interconnection of large power grids brings economic benefits and increases the possibility of large-scale power failure caused by disturbance. The reason for further expanding the power grid 7.1 accident in 2006 is that the fault-free tripping of the 500kV bulk-Zheng I line pair side switch cannot be accurately identified. Therefore, the power grid disturbance is sensed and identified quickly and effectively, so that the accurate and reliable action of the system is stabilized, the spread and the deterioration of the fault are inhibited, the random propagation of the cascading fault is blocked, and the safe and stable operation of the power grid is guaranteed.
However, the complex topology, power electronics and access of large-scale new energy of the interconnected power grid not only increase the diversity and complexity of power grid disturbance, but also enhance the coupling and association among various disturbances, causing multiple disturbances, resulting in the reduction of accuracy and reliability of disturbance identification. The main reasons for this are: the method has the advantages that multiple disturbances have an overlapping effect, and some useful characteristic information are mutually coupled and overlapped, so that the assumption of mutual independence of the original disturbance analysis method is destroyed, and disturbance characteristic intersection and failure are caused; secondly, attenuation and distortion of the disturbance transient traveling wave with a special shape generated by the power grid fault disturbance occur in the transmission process of the line, and various random noise influences are caused, so that the sudden change characteristic of the disturbance is not obvious. These all present unprecedented difficulties for accurate identification of grid disturbances.
In summary, how to sense and accurately identify multiple disturbances of the power grid in time becomes a technical problem to be solved urgently by those skilled in the art.
Disclosure of Invention
The invention aims to provide a method and a device for accurately identifying multiple disturbances of a wide-area power grid based on big data analysis, so as to solve the technical problem that the multiple disturbances of the power grid cannot be timely sensed and accurately identified in the prior art.
In order to solve the technical problems, the invention adopts the following technical scheme:
in a first aspect of the embodiments of the present invention, a wide area power grid multiple disturbance accurate identification method based on big data analysis is provided, including the steps of:
acquiring disturbance transient state big data of a wide area power grid in a monitoring range;
extracting conventional characteristics according to the disturbance transient big data, wherein the conventional characteristics comprise signal amplitude, phase, filtering residual error and time-varying fading factors; taking the filtering residual error as a test statistic, and detecting whether power grid disturbance occurs in real time;
constructing a depth network for extracting depth disturbance characteristics, wherein the depth network is a stacked self-encoder which is formed by stacking a plurality of automatic encoders and integrates a layer of BP neural network classifier on the top layer of the network;
the conventional features and the disturbance transient state big data are used as input layer data of a depth network together, the depth disturbance features and the conventional features are fused through layer-by-layer training of a stacked automatic encoder, and the fused coupling features are extracted;
and decoupling and separating the coupling characteristics into different single disturbance types.
Optionally, extracting conventional features according to the disturbance transient big data includes:
acquiring an electrical measurement signal from the disturbance transient state big data, and establishing a state space model according to the electrical measurement signal;
setting a strong tracking Kalman filtering algorithm with a time-varying fading factor;
and estimating a state variable in the state space model by a strong tracking Kalman filtering algorithm to obtain a signal amplitude, a phase, a filtering residual error and a time-varying fading factor.
Optionally, the establishing of the state space model according to the electrical measurement signal includes:
selecting a state variable:
Xk＝[A1cosφ1,A1sinφ1,…,Ancosφn,Ansinφn,…,ANcosφN,ANsinφN,Ad,Adα]T
establishing a state space model as follows:
wherein,
Hk＝[sin(ω1kTs),cos(ω1kTs),…,sin(ωnkTs),cos(ωnkTs),…,sin(ωNkTs),cos(ωNkTs),1-kTs]
estimating state variables in the state space model through a strong tracking Kalman filtering algorithm to obtain conventional characteristics including signal amplitude, phase, filtering residual error and time-varying fading factors, wherein the conventional characteristics include: the signal amplitude and phase are calculated as follows:
wherein,is a state variable x2n-1,k，x2n,kAn estimated value of (d);
estimated filtered residualIs the difference between the actual measured value and the estimated value of the output signal, i.e.:
wherein,is a state variable XkAn estimate of (d).
Optionally, the method for setting a strong tracking kalman filter algorithm with a time-varying fading factor includes the steps of:
and introducing a time-varying fading factor, adjusting a state prediction error covariance matrix and a gain matrix in real time on line, and keeping the sequence of the filtering residual errors orthogonal.
Optionally, decoupling and separating the coupling features includes:
a mapping function from containing multiple types of disturbance features to being separated into different single disturbance features is constructed through deep learning, and decoupling and separation of the coupling features are achieved.
Optionally, the conventional feature and the disturbance transient big data are used together as input layer data of the deep network, and the method further includes the following steps:
expressing the disturbance transient big data into a form of a plurality of groups of space-time sequences, and carrying out normalization processing on the plurality of groups of space-time sequences and the conventional characteristics, wherein a normalization formula is as follows:
where x denotes the original raw signal value, xnorIs a normalized value, xmaxAnd xminRespectively represent the maximum and minimum values of each set of space-time sequences or conventional feature vectors.
Optionally, constructing a depth network for extracting depth perturbation features includes the steps of:
performing adaptive hidden layer-by-layer number optimization according to the reconstructed residual error by combining the minimized reconstructed residual error and the white noise degree of the reconstructed residual error, and determining the layer number of the depth network;
after the layer number of the depth network is determined, the number of nodes of the hidden layer is determined by researching the relation between the node number of the hidden layer and the identification precision, and the node number of each layer of the depth network is gradually decreased from bottom to top layer by layer.
According to a second aspect of the embodiment of the invention, a wide-area power grid multiple disturbance accurate identification device based on big data analysis is provided, and the device comprises a data module, a conventional characteristic module, a deep network setting module, a deep learning module and a coupling separation module.
The data module is used for acquiring disturbance transient big data of a wide area power grid in a monitoring range; the conventional characteristic module is used for extracting conventional characteristics according to the disturbance transient big data, taking the filtering residual error as test statistic and detecting whether disturbance occurs in real time; the depth network setting module is used for constructing a depth network for extracting depth disturbance characteristics, and the depth network is set to be a stacked self-encoder which is formed by stacking a plurality of automatic encoders and integrates a layer of BP neural network classifier on the top layer of the network; the deep learning module is used for taking the conventional characteristics and the disturbance transient state big data as the input layer data of the deep network, fusing the deep disturbance characteristics and the conventional characteristics through the layer-by-layer training of the stacked automatic encoder, and extracting the fused coupling characteristics; and the coupling separation module is used for decoupling and separating the coupling characteristics into different single disturbance types.
Optionally, a conventional feature module to:
acquiring an electrical measurement signal from the disturbance transient state big data, and establishing a state space model according to the electrical measurement signal; setting a strong tracking Kalman filtering algorithm with a time-varying fading factor; and estimating the state variable in the state space model by a strong tracking Kalman filtering algorithm to obtain conventional characteristics including signal amplitude, phase, filtering residual error and time-varying fading factors.
Optionally, the coupling separation module is configured to:
a mapping function from containing multiple types of disturbance features to being separated into different single disturbance features is constructed through deep learning, and decoupling and separation of the coupling features are achieved.
Compared with the prior art, the invention can achieve the following technical effects:
according to the wide area power grid multiple disturbance accurate identification method based on big data analysis, the filtering residual error is used as the test statistic, and the occurrence of disturbance can be quickly detected; in addition, the invention can automatically learn the space-time structure information of the disturbance transient big data space-time sequence through a depth network with a specific structure, extract deeper disturbance characteristics, map the characteristics of input data from an original mode space to a new characteristic space which is more beneficial to perception and identification, and obtain the depth disturbance characteristics; in addition, the extracted conventional features are used as original input signals and added into an input layer of the depth network, and an improved depth learning model with the depth features extracted layer by layer and the conventional features optimized layer by layer is obtained through the layer-by-layer training of the stacked automatic encoder, so that the depth learning features and the conventional features are effectively fused under the condition that the overall framework of the depth network is not changed, the training effect of the depth learning model is improved, the output coupling features are more comprehensive and accurate and are easy to identify, and the accurate identification of the multiple disturbances of the power grid can be realized after the coupling separation.
Furthermore, the invention provides that the conventional characteristics are extracted through a strong tracking Kalman filtering algorithm with a time-varying fading factor, when the power grid disturbance occurs, the operation signal corresponding to the power grid has mutation, the traditional Kalman filtering algorithm cannot respond to the mutation signal in time, and the ability of timely tracking the mutation signal is lost.
Furthermore, the invention constructs a mapping function from containing multiple disturbance types to being separated into different single disturbance types through deep learning, thereby realizing the decoupling and separation of the coupling characteristics.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a schematic flow chart of an embodiment of a wide area power grid multiple disturbance accurate identification method based on big data analysis according to the present invention;
FIG. 2 is a schematic main flow diagram of a technical scheme of a preferred embodiment of the wide area power grid multiple disturbance accurate identification method based on big data analysis;
fig. 3 is a schematic diagram of a deep network structure according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The present invention will be described in more detail with reference to the following embodiments in order to make the technical aspects of the present invention more apparent and understandable.
Example 1
The embodiment 1 of the invention provides a wide area power grid multiple disturbance accurate identification method based on big data analysis, and referring to fig. 1, the method comprises the following steps:
s101, obtaining disturbance transient state big data of the wide area power grid in the monitoring range.
Preferably, in the embodiment of the present invention, the perturbation transient big data is space-time multi-scale big data. The space-time multi-scale big data comprises disturbance transient waveform data with multiple time scales and multiple space scales. The multi-time scale data information covers data information including real-time data, historical data and predicted future data, and the multi-space scale data information covers data information including different power elements, intervals, voltage levels and the like. The data in the spatio-temporal database has a stereo, panoramic information stream and can be sufficiently faithful to express spatio-temporal phenomena.
The space-time multi-scale big data types are various, including but not limited to real-time data, historical data, text data, multimedia data, time sequence data and other various structured and semi-structured data, unstructured data and the like, and space-time multi-scale stereo information and panoramic information transmitted by nanosecond-level power grid transient energy can be displayed.
The embodiment of the invention analyzes laboratory simulation data and actual measurement data obtained on site by an actual system, constructs a knowledge base of a power grid space-time big data disturbance event, researches the characteristic quantity difference of various disturbances (such as short-circuit trip, fault-free trip, generator tripping, load shedding, line fault-free trip and system oscillation disturbance) under the influence of different power grid structures, different power grid scales and different external environments, researches the evolution mechanism of multiple disturbances, and provides basis and foundation for accurate sensing and identification of the multiple disturbances.
Preferably, in order to obtain qualitative characteristics of power grid disturbance, the method researches and analyzes mechanism characteristics and electrical quantity change characteristics of multiple common single disturbances (such as short circuit, generator tripping and load shedding) and multiple disturbances causing low-frequency oscillation, explores an evolution mechanism of the multiple disturbances, performs qualitative theoretical analysis on a transient process after the disturbance occurs, and lays a foundation for understanding various disturbance properties.
In order to obtain quantitative characteristics of power grid disturbance, various typical power grid models are set up on an EMTP (empirical mode transfer protocol) and MATLAB (matrix laboratory) simulation platform, multiple disturbances such as short-circuit fault, line fault-free tripping, generator breakdown, load shedding, system oscillation, various double disturbances (such as two-phase grounding short circuit + generator shedding, generator shedding + load shedding), triple disturbances (two-phase grounding short circuit + generator shedding + load shedding, generator shedding + load shedding) and the like of an actual power grid at different positions, different structures and different strength conditions inside and outside a region are simulated, various disturbances are analyzed from a time domain and a frequency domain respectively, time-frequency domain characteristics of single disturbance are researched, characteristic differences of the single disturbance are contrastively analyzed, and characteristic quantity parameters of different disturbance types are highlighted. The method comprises the steps of simulating and generating multiple disturbances generated by superposition of various single disturbances, comparing and analyzing the difference between the system impact caused by the multiple disturbances and the system impact caused by each single disturbance, researching the characteristic changes of time domain and frequency domain of the single disturbances after superposition, determining the characteristic value capable of representing the multiple disturbances, qualitatively analyzing the relation and difference between the characteristics of the single disturbances and the multiple disturbances, and providing basis and foundation for perception and identification of the subsequent multiple disturbances.
In order to fully reflect disturbance characteristics through wide-area measurement information, various typical power grid models are set up in a laboratory, various disturbance scenes (such as short circuit, generator tripping, load shedding, low-frequency oscillation and the like) are set through a PW (pseudo wire) port online relay protection test system, an energy flow signal is detected by using a signal acquisition module, a plurality of groups of actually-measured disturbance transient waveforms are recorded by using a 50MHz digital oscilloscope, a disturbance waveform difference rule under different power grid scales and power grid structure conditions is searched, and a theoretical research result is further improved and perfected.
In the step S101, the collection of disturbance transient big data is completed, and a complete data basis and a preliminary theoretical basis are provided for accurate sensing and identification of power grid disturbance.
And S102, extracting conventional features according to the disturbance transient big data.
As an implementable manner, the conventional features include signal amplitude, phase, filtered residual, and time-varying fading factor.
And S103, detecting whether power grid disturbance occurs in real time by taking the filtering residual error as a test statistic.
And S104, constructing a depth network for extracting the depth disturbance characteristics.
The deep network is set to be a stacked self-encoder which is formed by stacking a plurality of automatic encoders and integrates a layer of BP neural network classifier at the top layer of the network.
And S105, fusing the depth disturbance feature and the conventional feature through depth learning, and outputting the coupling feature.
Specifically, the conventional features and the disturbance transient state big data are used as the input layer data of the depth network together, and the depth disturbance features and the conventional features are fused through layer-by-layer training of a stacked automatic encoder.
And S106, decoupling and separating the coupling characteristics into different single disturbance types.
The embodiment of the invention takes the filtering residual error as the statistic for detecting the power grid disturbance, can quickly detect the occurrence of the power grid disturbance and determine the initial moment of the power grid disturbance; and conventional characteristics and deep disturbance characteristics are fused through deep learning, so that the space-time characteristics of disturbance transient original data are more accurately reflected, and accurate identification of multiple disturbances of the power grid is further realized.
Example 2
Embodiment 2 of the present invention provides a preferred embodiment of a wide area power grid multiple disturbance accurate identification method based on big data analysis. The main flow diagram of the technical solution of this embodiment is shown in fig. 2.
In this embodiment, the large data is collected in the same manner as in the embodiment, which is not described again, and after the disturbance transient large data is collected, the following steps are performed:
s201, time domain denoising and disturbance moment detection of the disturbance transient state big data.
The electrical quantity (e.g., voltage or current) obtained by the wide-area measurement system is generally a mixed signal containing a power frequency fundamental component, harmonic components, a transient component, and some random white noise, which can be described as:
yk＝zk+vk (1)
wherein, ykIs a measured electrical measurement signal (e.g. voltage or current) { ykE R, K1, 2, …, K (K is the sample length), vkIs white Gaussian random noise, and vk～N(0,Rk) I.e. vkSubject to a mathematical expectation of 0 and variance ofNormal distribution of (2); z is a radical ofkThe electrical quantity signal after removing noise can represent the sum of fundamental wave, each harmonic and direct current component, namely:
wherein N is the highest order of the harmonic wave, An、φnAmplitude and phase of the fundamental (n ═ 1) or nth harmonic, respectively, ωn＝2πnf0(f0At fundamental frequency, typically 50Hz), TsFor a sampling period, Ad exp(-αkTs) Is a direct current component.
The direct current component A in the above formulad exp(-αkTs) Expansion with taylor approximation yields:
the formula (3) is a general expression for obtaining the electric quantity signal, and is a basis for establishing state space models of the types (4) - (5). Wherein α is a time constant, AdTo attenuate the magnitude of the dc component.
In order to extract conventional characteristic parameters (such as amplitude, phase angle and the like) of a disturbance transient by using a kalman filter algorithm, state space modeling needs to be performed on an electrical measurement signal, and therefore, a 2N + 1-dimensional state variable is selected:
Xk＝[x1,k,x2,k,…,x2n-1,k,x2n,k,…,x2N-1,k,x2N,k,x2N+1,k]
＝[A1cosφ1,A1sinφ1,…,Ancosφn,Ansinφn…,ANcosφN,ANsinφN,Ad,Adα]Tthen, the following state space model can be established:
wherein etakRepresentative of observation noise, HkRepresents an observation matrix of the system, an
Hk＝[sin(ω1kTs),cos(ω1kTs),…,sin(ωnkTs),cos(ωnkTs),…,sin(ωNkTs),cos(ωNkTs),1-kTs] (5)
When disturbance occurs, the operation signal corresponding to the power grid has mutability, the traditional Kalman filtering algorithm cannot respond to the mutation signal in time, and the ability of timely tracking the mutation signal is lost. Therefore, a strong tracking Kalman filtering algorithm with a time-varying fading factor is researched, and the real-time accurate tracking of the sudden change component of the disturbance signal is realized. As an implementable manner, the kalman filter equation with time-varying fading factor is as follows:
and (3) state prediction value:
state estimation value:
filter gain array Kk：Kk＝Pk|k-1Hk T[HkPk|k-1Hk T+Rk]-1；
Prediction error covariance matrix Pk|k-1：Pk|k-1＝λkPk-1|k-1+Qk；
Estimation error covariance matrix Pk|k：Pk|k＝[I-KkHk]Pk|k-1；
Residual sequence ek：
Fading factor lambdak：λk＝max{1,trace(Nk)/trace(Mk)}
Wherein trace () represents trace-seeking, and
Mk＝HkPk|k-1Hk T
the algorithm adjusts a state prediction error covariance matrix and a filter gain matrix on line in real time by introducing a time-varying fading factor, so that effective information is extracted from residual errors to the maximum extent, and the tracking capability and robustness of Kalman filtering on mutation signals are improved.
And (3) estimating state variables in the model (4) by applying strong tracking Kalman filtering, so as to obtain characteristic quantities such as signal amplitude, phase and the like. Namely:
wherein,is a state variable x2n-1,k，x2n,kEstimated value of (A), XkIs a 2N +1 dimensional state variable, x2n-1,kAnd x2n,kAre each XkThe 2n-1, 2 n-dimensional component of (1), wherein:
Xk＝[x1,k,x2,k,…,x2n-1,k,x2n,k,…,x2N-1,k,x2N,k,x2N+1,k]
＝[A1cosφ1,A1sinφ1,…,Ancosφn,Ansinφn…,ANcosφN,ANsinφN,Ad,Adα]T
x2n-1,kand x2n,kIs the state estimate of kalman filtering.
Estimated residual sequenceIs the difference between the actual measured value and the estimated value of the output signal, i.e.:
wherein,is a state variable XkAn estimate of (d).
In a steady state (normal operation of a system), Kalman filtering can well track signal waveforms, and absolute values of errors obtained at the momentAre small. When the system is disturbed, the absolute value of the error is obtained because the filter can not immediately follow the sudden change of the signalLarger, the moment of occurrence of the disturbance can be detected accordingly.
Meanwhile, when the power system is in normal operation, the filtering residual error of the power system should be a set of zero-mean, uncorrelated random variables. When the system is disturbed, the mathematical expectation of the residual error is not zero, and the distribution is no longer the standard normal distribution. Therefore, based on the theory of normal distribution hypothesis test, the filter residual is used as the test statistic, so that the occurrence of disturbance can be quickly detected. Further, the time-varying fading factor of the adaptive adjustment suddenly increases at the time of occurrence of the disturbance start-stop, and therefore the fading factor can be extracted as the feature amount recognition disturbance.
S202, mining the spatio-temporal structure information of the spatio-temporal multi-scale big data and extracting high-order deep features.
The grid space-time multi-scale big data has obvious space-time structure and nonlinear relation, and the shallow structure has very limited capability of mining the nonlinear structure information, so that the deep network modeling of the space-time sequence is researched, and the space-time structure information and deep abstract characteristics in the data are fully mined. The noise-removed space-time sequence of the conventional features can be obtained in step S201, and a deep network suitable for extracting the disturbance features, which is a stacked self-encoder formed by stacking a plurality of Automatic Encoders (AE) and integrating a layer of BP neural network classifier on the top layer of the network, is constructed by combining the particularity of space-time multi-scale big data.
Aiming at the design problem of the network structure depth (including the number of network layers and the number of nodes of a hidden layer) in the deep learning, the invention explores a multidimensional model evaluation system considering minimization of the reconstructed residual and the white whitening degree of the reconstructed residual according to the reconstructed residual, and provides an optimization mechanism of self-adaption hidden layer number, thereby determining the number of layers of the deep network. After the number of the network layers is determined, the number of the hidden layer nodes is determined by researching the relation between the number of the hidden layer nodes and the identification precision. In addition, the node number of each layer of the deep network is gradually decreased from bottom to top, so that redundant information is removed.
The deep learning network generally has over-fitting training, namely the accuracy of the model is good and the generalization capability of the model is poor. In this regard, preferably, in the embodiment of the present invention, a regularization parameter optimization method for a deep learning network is proposed, that is, in a conventional unsupervised objective function, a regularization term is added:
where M is the number of training data pairs, Y is the desired output data vector, a is the output vector of the training data through the deep network, λ is the regularization coefficient (governing the relative importance of the two terms in equation (8)), and W is the deep network ownership value parameter. In the formula (8), the first term is a mean square error term; the second term is regularization of the network weight parameter W, i.e. a network weight attenuation term, which aims to reduce the variation amplitude of the weight value and prevent the learned network transition fitting.
The time-space structure information of the time-space sequence can be automatically learned through the deep network modeling and learning, deeper disturbance characteristics are extracted, and the characteristics of the input data are mapped to a new characteristic space which is more beneficial to perception and identification from the original mode space.
S203, accurately sensing and identifying multiple disturbances.
In order to realize the separation of the coupling characteristics, a mapping function (namely neuron functions of each layer of a depth network after training) from the disturbance characteristics containing multiple types (such as two types of characteristics containing an excisional generator and an excisional load) to the disturbance types separated into different single types (such as the excisional generator and the excisional load) is supervised and learned by a depth learning algorithm by utilizing data of a large number of known disturbance types (such as double disturbance of a cutter and a excisional load), so that the decoupling separation of the coupling characteristics is realized.
In the identification of the actual power grid disturbance type, the importance degrees of the extracted multiple characteristic quantities are not equal, namely, the different types of characteristic quantities account for different proportions in classification identification, for example, the current amplitude characteristic can distinguish short circuit and load shedding (or load shedding) better than the frequency characteristic. Therefore, under the condition that the influence of each characteristic quantity on the final identification effect is different, the embodiment of the invention provides a power grid disturbance identification mechanism fusing the depth characteristic and the conventional characteristic by taking the conventional characteristic as guidance and taking the depth network as a pipeline.
Specifically, in order to fuse the conventional feature quantities such as the estimated signal amplitude, the estimated residual error, the fading factor and the like into the depth abstract feature, as shown in fig. 3, the extracted conventional feature quantity is used as an original input signal, added to an input layer of a depth network, and trained layer by layer through a stacked automatic encoder, so as to obtain an improved depth learning model with the depth feature extracted layer by layer and the conventional feature optimized layer by layer. In this way, perfect fusion of deep learning features and conventional features is achieved without changing the overall framework of the deep network. Meanwhile, in order to eliminate the difference between different data orders, the input signals in fig. 3, i.e., the denoised space-time sequence and the conventional eigenvector, are normalized. The normalization formula is as follows:
where x denotes the original raw signal value, xnorIs a normalized value, xmaxAnd xminRespectively represent each groupThe maximum and minimum values of the spatio-temporal sequence or the conventional feature vector. After the input data is subjected to normalization processing, all the data are mapped to [0,1 ]]Within the range, each data is in the same order of magnitude, achieving comparability between different data.
In consideration of different influences of various different types of characteristic quantities on a final classification result, namely different proportions of the different types of characteristic quantities in the classification action, the invention also provides a weight classifier for fusing the depth characteristic with the conventional characteristic. Namely: adding a layer of BP network (see figure 3) classifier at the top layer of the deep network, and adjusting the weight parameters of the BP classifier by using supervised learning through samples with known disturbance types so as to adjust the weight of various features according to classification results; and meanwhile, fine-tuning parameters of all layers through a back propagation algorithm of the multilayer neural network.
The feature fusion weight classifier researched by the invention is based on a stacked automatic encoder framework, takes deep learning as a pipeline, and is added with the guidance of conventional features, so that the training effect can be effectively improved, and the accurate identification of the power grid disturbance is realized.
Example 3
The embodiment 3 of the invention provides a wide area power grid multiple disturbance accurate identification device based on big data analysis. The device comprises a data module, a conventional characteristic module, a deep network setting module, a deep learning module and a coupling separation module.
The data module is used for acquiring disturbance transient big data of a wide area power grid in a monitoring range; the conventional characteristic module is used for extracting conventional characteristics according to the disturbance transient big data, and detecting whether disturbance occurs in real time by taking the filtering residual error as test statistic; the depth network setting module is used for constructing a depth network for extracting depth disturbance characteristics, and the depth network is set to be a stacked self-encoder which is formed by stacking a plurality of automatic encoders and integrates a layer of BP neural network classifier on the top layer of the network; the deep learning module is used for taking the conventional characteristics and the disturbance transient state big data as the input layer data of the deep network, fusing the deep disturbance characteristics and the conventional characteristics through the layer-by-layer training of the stacked automatic encoder, and extracting the fused coupling characteristics; and the coupling separation module is used for decoupling and separating the coupling characteristics into different single disturbance types.
Preferably, the routine features module is configured to:
acquiring an electrical measurement signal from the disturbance transient state big data, and establishing a state space model according to the electrical measurement signal; setting a strong tracking Kalman filtering algorithm with a time-varying fading factor; and estimating the state variable in the state space model by a strong tracking Kalman filtering algorithm to obtain conventional characteristics including signal amplitude, phase, filtering residual error and time-varying fading factors.
Preferably, the coupling separation module is configured to:
a mapping function from containing multiple types of disturbance features to being separated into different single disturbance features is constructed through deep learning, and decoupling and separation of the coupling features are achieved.
In summary, under the new situation of interconnected power grid development, the invention provides a new method for sensing and identifying power grid disturbance based on information driving by taking a big data processing technology as a means around deep utilization of wide-area space-time multi-scale big data, and compared with the prior art, the method has the following beneficial effects: the power grid disturbance can be detected in time, and the disturbance type identification accuracy is improved. Based on the technical scheme of the invention, the panoramic situation perception capability of the power grid is expected to be improved, so that the whole transient process of complex multiple fault disturbance and cascading faults is controlled, the functions of automatic early warning, fault self-healing, preventive control and the like of the smart power grid are boosted, and the method has important scientific research significance and engineering application prospect.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may be made by those skilled in the art without departing from the principle of the invention.
Claims (10)
1. The wide area power grid multiple disturbance accurate identification method based on big data analysis is characterized by comprising the following steps:
acquiring disturbance transient state big data of a wide area power grid in a monitoring range;
extracting conventional features according to the disturbance transient big data; the conventional characteristics comprise signal amplitude, phase, filtering residual error and time-varying fading factor; taking the filtering residual error as a test statistic to detect whether power grid disturbance occurs in real time;
constructing a depth network for extracting depth disturbance characteristics, wherein the depth network is a stacked self-encoder which is formed by stacking a plurality of automatic encoders and integrates a layer of BP neural network classifier on the top layer of the network;
the conventional features and the disturbance transient state big data are used as input layer data of the depth network together, the depth disturbance features and the conventional features are fused through layer-by-layer training of the stacked automatic encoder, and the fused coupling features are extracted;
and decoupling and separating the coupling characteristics into different single disturbance types.
2. The wide-area power grid multiple disturbance accurate identification method based on big data analysis according to claim 1, wherein the step of extracting conventional features according to the disturbance transient big data comprises:
acquiring an electrical measurement signal from the disturbance transient state big data, and establishing a state space model according to the electrical measurement signal;
setting a strong tracking Kalman filtering algorithm with a time-varying fading factor;
and estimating the state variable in the state space model through the strong tracking Kalman filtering algorithm to obtain a signal amplitude, a phase, a filtering residual error and a time-varying fading factor.
3. The wide-area power grid multiple disturbance accurate identification method based on big data analysis as claimed in claim 2, wherein:
the step of establishing a state space model according to the electrical measurement signal comprises the following steps:
selecting a state variable:
Xk＝[A1 cosφ1,A1 sinφ1,…,An cosφn,An sinφn,…,AN cosφN,AN sinφN,Ad,Adα]T
wherein A is1、φ1Representing the amplitude and phase of the fundamental wave, An、φnRespectively representing the amplitude and phase of the nth harmonic, alpha being a time constant, AdTo attenuate the magnitude of the dc component;
establishing a state space model as follows:
wherein eta iskRepresenting observation noise, ykRepresenting electrical measurement signals, v, obtained by actual measurementkIs Gaussian random white noise, HkAn observation matrix representing the system;
Hk＝[sin(ω1kTs),cos(ω1kTs),…,sin(ωnkTs),cos(ωnkTs),…,sin(ωNkTs),cos(ωNkTs),1-kTs]
wherein, ω isn＝2πnf0，f0Is the fundamental frequency, N denotes the nth harmonic, N is the highest order of the harmonic, TsK represents a sampling sequence number for a sampling period;
estimating a state variable in the state space model through the strong tracking Kalman filtering algorithm to obtain a signal amplitude, a phase, a filtering residual error and a time-varying fading factor, wherein the step comprises the following steps of calculating the signal amplitude and the phase according to the following formula:
wherein,is a state variable x2n-1,k，x2n,kAn estimated value of (d); x is the number of2n-1,kAnd x2n,kAre each XkThe 2n-1, 2 n-dimensional component of (1);
estimated filtered residualIs the difference between the actual measured value and the estimated value of the output signal, i.e.:
wherein,is a state variable XkAn estimate of (d).
4. The wide-area power grid multiple disturbance accurate identification method based on big data analysis according to claim 2, wherein the step of setting a strong tracking Kalman filtering algorithm with a time-varying fading factor comprises the steps of:
and introducing a time-varying fading factor into a Kalman filtering algorithm, adjusting a state prediction error covariance matrix and a gain matrix in real time on line, and keeping the sequence of the filtering residual errors orthogonal.
5. The wide-area power grid multiple disturbance accurate identification method based on big data analysis according to claim 1, wherein the step of decoupling and separating the coupling characteristics comprises the steps of:
a mapping function from containing multiple types of disturbance features to being separated into different single disturbance features is constructed through deep learning, and decoupling and separation of the coupling features are achieved.
6. The wide-area power grid multiple disturbance accurate identification method based on big data analysis according to claim 1, wherein the step uses the conventional characteristics and the disturbance transient big data together as the input layer data of the deep network, and further comprises the steps of:
expressing the disturbance transient state big data into a form of a plurality of groups of space-time sequences, and carrying out normalization processing on the plurality of groups of space-time sequences and the conventional characteristics, wherein a normalization formula is as follows:
where x denotes the original raw signal value, xnorIs a normalized value, xmaxAnd xminRespectively represent the maximum and minimum values of each set of space-time sequences or conventional feature vectors.
7. The wide-area power grid multiple disturbance accurate identification method based on big data analysis according to claim 1, wherein the step of constructing the deep network for extracting deep disturbance features comprises the steps of:
performing adaptive hidden layer-by-layer number optimization according to the reconstructed residual error and by combining the minimized reconstructed residual error and the white noise degree of the reconstructed residual error, and determining the layer number of the depth network;
after the number of layers of the depth network is determined, the number of nodes of the hidden layer is determined by researching the relation between the number of nodes of the hidden layer and the identification precision, and the number of nodes of each layer of the depth network is gradually decreased from bottom to top layer by layer.
8. The wide-area power grid multiple disturbance accurate identification device based on big data analysis is characterized by comprising a data module, a conventional characteristic module, a deep network setting module, a deep learning module and a coupling separation module;
the data module is used for acquiring disturbance transient state big data of a wide area power grid in a monitoring range;
the conventional feature module is configured to extract a conventional feature according to the disturbance transient big data, where the conventional feature includes a filtering residual error; taking the filtering residual error as a test statistic to detect whether disturbance occurs in real time;
the depth network setting module is used for constructing a depth network for extracting depth disturbance characteristics, and the depth network is set to be a stacked self-encoder which is formed by stacking a plurality of automatic encoders and integrates a layer of BP neural network classifier on the top layer of the network;
the deep learning module is used for taking the conventional features and the disturbance transient state big data as input layer data of the deep network, fusing the deep disturbance features and the conventional features through layer-by-layer training of the stacked automatic encoder, and extracting fused coupling features;
and the coupling separation module is used for decoupling and separating the coupling characteristics into different single disturbance types.
9. The wide-area power grid multiple disturbance accurate identification device based on big data analysis as claimed in claim 8, wherein the conventional feature module is configured to:
acquiring an electrical measurement signal from the disturbance transient state big data, and establishing a state space model according to the electrical measurement signal;
setting a strong tracking Kalman filtering algorithm with a time-varying fading factor;
and estimating the state variables in the state space model through the strong tracking Kalman filtering algorithm to obtain conventional characteristics including signal amplitude, phase, filtering residual error and time-varying fading factors.
10. The wide-area power grid multiple disturbance accurate identification device based on big data analysis according to claim 8, wherein the coupling separation module is configured to:
a mapping function from containing multiple types of disturbance features to being separated into different single disturbance features is constructed through deep learning, and decoupling and separation of the coupling features are achieved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910674207.XA CN110414412B (en) | 2019-07-25 | 2019-07-25 | Wide-area power grid multiple disturbance accurate identification method and device based on big data analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910674207.XA CN110414412B (en) | 2019-07-25 | 2019-07-25 | Wide-area power grid multiple disturbance accurate identification method and device based on big data analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110414412A true CN110414412A (en) | 2019-11-05 |
| CN110414412B CN110414412B (en) | 2023-03-31 |
Family
ID=68362979
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910674207.XA Active CN110414412B (en) | 2019-07-25 | 2019-07-25 | Wide-area power grid multiple disturbance accurate identification method and device based on big data analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110414412B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111046327A (en) * | 2019-12-18 | 2020-04-21 | 河海大学 | Prony analysis method suitable for low-frequency oscillation and subsynchronous oscillation identification |
| CN111355247A (en) * | 2020-02-18 | 2020-06-30 | 清华大学 | Power grid low-frequency oscillation prediction method and device |
| CN111551819A (en) * | 2020-04-16 | 2020-08-18 | 国网湖南省电力有限公司 | Micro-grid fault detection method and device and storage medium |
| CN112183368A (en) * | 2020-09-29 | 2021-01-05 | 国网四川省电力公司经济技术研究院 | LSTM-based quick identification method for low-frequency oscillation modal characteristics of power system |
| CN112448403A (en) * | 2020-11-16 | 2021-03-05 | 国网四川省电力公司经济技术研究院 | Decoupling configuration method for energy storage of power distribution network |
| CN113240057A (en) * | 2021-07-12 | 2021-08-10 | 武汉中原电子信息有限公司 | High-precision error detection method and system based on electric power data acquisition |
| CN113486868A (en) * | 2021-09-07 | 2021-10-08 | 中南大学 | Motor fault diagnosis method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426839A (en) * | 2015-11-18 | 2016-03-23 | 清华大学 | Power system overvoltage classification method based on sparse autocoder |
| CN108846410A (en) * | 2018-05-02 | 2018-11-20 | 湘潭大学 | Power Quality Disturbance Classification Method based on sparse autocoding deep neural network |
-
2019
- 2019-07-25 CN CN201910674207.XA patent/CN110414412B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426839A (en) * | 2015-11-18 | 2016-03-23 | 清华大学 | Power system overvoltage classification method based on sparse autocoder |
| CN108846410A (en) * | 2018-05-02 | 2018-11-20 | 湘潭大学 | Power Quality Disturbance Classification Method based on sparse autocoding deep neural network |
Non-Patent Citations (3)
| Title |
|---|
| 吴智利等: "低频采样下基于卡尔曼滤波的同步相量测量算法的研究", 《电力系统保护与控制》 * |
| 贺姗等: "带渐消因子的容积卡尔曼滤波算法", 《科技与创新》 * |
| 陈晓静等: "一种实时电能质量扰动分类方法", 《电工技术学报》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111046327B (en) * | 2019-12-18 | 2021-10-19 | 河海大学 | Prony analysis method suitable for low-frequency oscillation and subsynchronous oscillation identification |
| CN111046327A (en) * | 2019-12-18 | 2020-04-21 | 河海大学 | Prony analysis method suitable for low-frequency oscillation and subsynchronous oscillation identification |
| CN111355247A (en) * | 2020-02-18 | 2020-06-30 | 清华大学 | Power grid low-frequency oscillation prediction method and device |
| CN111355247B (en) * | 2020-02-18 | 2022-04-01 | 清华大学 | Power grid low-frequency oscillation prediction method and device |
| CN111551819A (en) * | 2020-04-16 | 2020-08-18 | 国网湖南省电力有限公司 | Micro-grid fault detection method and device and storage medium |
| CN111551819B (en) * | 2020-04-16 | 2022-04-29 | 国网湖南省电力有限公司 | Micro-grid fault detection method and device and storage medium |
| CN112183368A (en) * | 2020-09-29 | 2021-01-05 | 国网四川省电力公司经济技术研究院 | LSTM-based quick identification method for low-frequency oscillation modal characteristics of power system |
| CN112183368B (en) * | 2020-09-29 | 2024-02-02 | 国网四川省电力公司经济技术研究院 | LSTM-based rapid identification method for low-frequency oscillation modal characteristics of power system |
| CN112448403A (en) * | 2020-11-16 | 2021-03-05 | 国网四川省电力公司经济技术研究院 | Decoupling configuration method for energy storage of power distribution network |
| CN112448403B (en) * | 2020-11-16 | 2022-05-17 | 国网四川省电力公司经济技术研究院 | Decoupling configuration method for energy storage of power distribution network |
| CN113240057A (en) * | 2021-07-12 | 2021-08-10 | 武汉中原电子信息有限公司 | High-precision error detection method and system based on electric power data acquisition |
| CN113486868A (en) * | 2021-09-07 | 2021-10-08 | 中南大学 | Motor fault diagnosis method and system |
| CN113486868B (en) * | 2021-09-07 | 2022-02-11 | 中南大学 | Motor fault diagnosis method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110414412B (en) | 2023-03-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110414412B (en) | Wide-area power grid multiple disturbance accurate identification method and device based on big data analysis | |
| CN113837477B (en) | Method, device and equipment for predicting power grid faults under typhoon disasters driven by data | |
| CN110082640B (en) | Distribution network single-phase earth fault identification method based on long-time memory network | |
| CN102221651B (en) | Fault on-line diagnosis and early warning method of flameproof dry-type transformer for mine | |
| CN111756034A (en) | Transient voltage stability evaluation method for power system based on graph space-time network | |
| CN108535572B (en) | Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics | |
| CN111046327B (en) | Prony analysis method suitable for low-frequency oscillation and subsynchronous oscillation identification | |
| Souhe et al. | Fault detection, classification and location in power distribution smart grid using smart meters data | |
| CN113569928B (en) | Train running state detection data missing processing model and reconstruction method | |
| CN110619182A (en) | Power transmission line parameter identification and power transmission network modeling method based on WAMS big data | |
| Shi et al. | Expected output calculation based on inverse distance weighting and its application in anomaly detection of distributed photovoltaic power stations | |
| CN102087311B (en) | Method for improving measurement accuracy of power mutual inductor | |
| CN106529816B (en) | The method of adjustment and system of power line channel | |
| Ren et al. | Online anomaly detection using machine learning and HPC for power system synchrophasor measurements | |
| CN117349614A (en) | Frequency stability prediction method based on self-attention mechanism and space-time diagram convolution network | |
| CN103778479A (en) | Adaptive information fault-tolerant protection method | |
| Kanwal et al. | Artificial intelligence based faults identification, classification, and localization techniques in transmission lines-a review | |
| Horvath et al. | Sensor fault diagnosis of inland navigation system using physical model and pattern recognition approach | |
| Kozionov et al. | Wavelet-based sensor validation: Differentiating abrupt sensor faults from system dynamics | |
| CN106646106B (en) | Electric network fault detection method based on outlier's detection technology | |
| CN111521905A (en) | Power grid fault diagnosis method based on intelligent fusion of multi-dimensional data fault degrees | |
| CN108092272A (en) | A kind of voltage stabilization on-line monitoring method based on the Kalman filtering that fades | |
| CN109684749B (en) | Photovoltaic power station equivalent modeling method considering operating characteristics | |
| CN112964948B (en) | Series reactor running state online monitoring method and system based on wave recording data inversion | |
| George et al. | Machine Learning Based Setting-free Reach Element For Zone-1 Distance Protection |
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 |