CN109886604A - A kind of transient stability evaluation in power system method based on one-dimensional convolutional neural networks - Google Patents
A kind of transient stability evaluation in power system method based on one-dimensional convolutional neural networks Download PDFInfo
- Publication number
- CN109886604A CN109886604A CN201910187314.XA CN201910187314A CN109886604A CN 109886604 A CN109886604 A CN 109886604A CN 201910187314 A CN201910187314 A CN 201910187314A CN 109886604 A CN109886604 A CN 109886604A
- Authority
- CN
- China
- Prior art keywords
- sample
- stability evaluation
- model
- transient stability
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The present invention relates to a kind of transient stability evaluation in power system methods based on one-dimensional convolutional neural networks technology, this method, as original input data, forms training set after then carrying out data prediction to it again first with the voltage magnitude and phase angle of the bus that WAMS system acquisition arrives in electric system and the active and reactive power of branch;Transient stability evaluation in power system model is established using one-dimensional convolutional neural networks.The model is made of one-dimensional convolutional layer, one-dimensional pond layer and full connection output layer.The present invention carries out feature extraction using convolutional layer and pond layer, and the stealth mode of mining data forms the high-order feature for being more advantageous to Transient Stability Evaluation, and then will assess in its full articulamentum.This method assessment accuracy rate with higher, is of great significance for electrical power system on-line safety and stability evaluation.
Description
Technical field
The present invention relates to transient stability evaluation in power system, more particularly to a kind of electricity based on one-dimensional convolutional neural networks
Force system Transient Stability Evaluation method.
Technical background
Increasingly sophisticated electric system makes system security and stability control face a severe test, it is therefore desirable to explore robustness
Height and fast and accurately Transient Stability Evaluation method, so that operations staff can take suitably when encountering dangerous system mode
Control Measure.The assessment of time-domain-simulation method is accurate high in traditional temporarily steady appraisal procedure but calculates that time-consuming;Direct method can be fast
Speed Transient Stability Evaluation is provided as a result, but approximation can only be provided, and be difficult to use in complicated electric power system.Therefore conventional method
The electric system not being suitable under current situation.In past several years, deep learning is in image, voice and natural language processing
Field obtains quantum jump.Deep learning designs corresponding deep layer network structure for different task, is calculated by backpropagation
Method finds the labyrinth in large data sets.Wherein, convolutional neural networks extract different local features using convolution algorithm, often
For image domains, it was proved to equally effective to sequence data later, and was employed successfully in natural language modeling and speech production.Area
Not in the two-dimensional convolution neural network of image domains, processing sequence data use one-dimensional convolutional neural networks.The present invention proposes one
Kind of the mentioned method of Transient Stability Evaluation method based on one-dimensional convolutional neural networks can effective digging system transient process the overall situation
Timing information has preferable assessment performance.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of electrical power system transient based on one-dimensional convolutional neural networks is steady
Determine appraisal procedure, so as to overcome the problems existing in the prior art.
In order to solve the above technical problems, the present invention adopts the following technical solutions:
Transient stability evaluation in power system method based on one-dimensional convolutional neural networks, which is characterized in that the method
Step includes:
Step 1: by time-domain-simulation method or the historical data that electric system is collected using WAMS system as
Learning sample collection is divided into 0.0083s between systematic sampling, is collected always 3 after failure removal since occurring preceding 2 weeks wave points for failure
Cycle, using the active and reactive power of the voltage magnitude of bus and phase angle and branch as the input of sample characteristics, that is, model;
Step 2: data initialization processing is carried out to sample set;
Step 3: sample data is labeled according to generator's power and angle curve, stablizes sample and is labeled as 1, unstable sample
It is labeled as 0;
Step 4: Transient Stability Evaluation model of the building based on one-dimensional convolutional neural networks, and step 3 acceptance of the bid is poured in
Data set is divided into training set and test set;
Step 5: to assess accuracy rate being standard to temporary in step 4 using the training set and test set that are constructed in step 4
State Stability Assessment model carries out optimizing, saves the model of optimum precision;
Step 6: the new electric power system data information that WAMS system acquisition obtains is handled using step 2 and step 4,
Building assessment data set;
Step 7: assessment data set being assessed using the optimum precision model in step 5, obtains stability result.
In the step 1, using transient emulation or WAMS historical data as learning sample collection, it is divided between systematic sampling
0.0083s collects always 3 cycles after failure removal, by the voltage magnitude and phase of bus since occurring preceding 2 weeks wave points for failure
The active power of angle and branch is as the input of sample characteristics, that is, model, wherein since fault clearing time is not a fixation
Constant, therefore will on the basis of maximum fault clearing time to insufficient sequence carry out zero padding operation.
Initial method in the step 2 will carry out initialization process to sample set, initialize formula are as follows:
WhereinIndicate the primitive character value of the t moment of initial characteristics jth column time series,After indicating initialization
Characteristic value,WithRespectively indicate mean value and standard deviation that all sample jth in sample set arrange all characteristic values.
In the step 3, for each sample, transient stability mark is carried out to it according to transient stability index TSI,
The formula of TSI are as follows:
Wherein Δ δ max is any two generator's power and angles difference maximum value in the end 4s after disturbing.For each sample, such as
Fruit TSI is positive, then system is stable, and sample label is labeled as 1;If TSI is negative, system is unstable, sample label
It is labeled as 0.
In the step 4, the transient stability evaluation in power system model based on one-dimensional convolutional neural networks of building is one
It is a by two one-dimensional convolutional layer 1D-CONV1 and 1D-CONV2, an one-dimensional pond layer 1D-POOL, an one-dimensional maximum pond layer
1D-MaxPOOL and full articulamentum FC composition.
Appraisal procedure according to claim 1, it is characterised in that: pass through the data under various typical operation modes
Collection carries out parameter optimization to the Transient Stability Evaluation model in step 4, keeps best model.To the on-line prediction in step 6
Data set obtains Stability Assessment result using the assessment models kept in step 4.
Detailed description of the invention
Specific embodiments of the present invention will be described in further detail with reference to the accompanying drawing:
Fig. 1 shows a kind of transient stability evaluation in power system method based on one-dimensional convolutional neural networks of the present invention
Flow chart;
Fig. 2 shows model structures of the invention;
Fig. 3 shows the topology diagram of 10 machine of New England, 39 node system in this example.
Specific embodiment
Step 1: by time-domain-simulation method or the historical data that electric system is collected using WAMS system as
Learning sample collection is divided into 0.0083s between systematic sampling, is collected always 3 after failure removal since occurring preceding 2 weeks wave points for failure
Cycle, using the active and reactive power of the voltage magnitude of bus and phase angle and branch as the input of sample characteristics, that is, model;
Step 2: data initialization processing is carried out to sample set;
Step 3: sample data is labeled according to generator's power and angle curve, stablizes sample and is labeled as 1, unstable sample
It is labeled as 0;
Step 4: Transient Stability Evaluation model of the building based on one-dimensional convolutional neural networks, and step 3 acceptance of the bid is poured in
Data set is divided into training set/test set;
Step 5: to assess accuracy rate being standard to temporary in step 4 using the training set and test set that are constructed in step 4
State Stability Assessment model carries out optimizing, saves the model of optimum precision;
Step 6: the new electric power system data information that WAMS system acquisition obtains is handled using step 2 and step 4,
Building assessment data set;
Step 7: assessment data set being assessed using the optimum precision model in step 5, obtains stability result.
In the step 1, using transient emulation or WAMS historical data as learning sample collection, it is divided between systematic sampling
0.0083s collects always 3 cycles after failure removal, by the voltage magnitude and phase of bus since occurring preceding 2 weeks wave points for failure
The active and reactive power of angle and branch is as the input of sample characteristics, that is, model, wherein since fault clearing time is not one
The constant of a fixation, therefore zero padding operation will be carried out to insufficient sequence on the basis of maximum fault clearing time.
Initial method in the step 2 will carry out initialization process to sample set, initialize formula are as follows:
WhereinIndicate the primitive character value of the t moment of initial characteristics jth column time series,After indicating initialization
Characteristic value,WithRespectively indicate mean value and standard deviation that all sample jth in sample set arrange all characteristic values.
In the step 3, for each sample, transient stability mark is carried out to it according to generator's power and angle curve,
Calculation formula are as follows:
Wherein Δ δ max is any two generator's power and angles difference maximum value in the end 4s after disturbing.For each sample, such as
Fruit TSI is positive, then system is stable, and sample label is labeled as 1;If TSI is negative, system is unstable, sample label
It is labeled as 0.
In the step 4, the transient stability evaluation in power system model based on deep neural network of building be one by
Two one-dimensional convolutional layer 1D-CONV1 and 1D-CONV2, an one-dimensional pond layer 1D-POOL, an one-dimensional maximum pond layer 1D-
MaxPOOL and full articulamentum FC composition.
Appraisal procedure according to claim 1, it is characterised in that: pass through the data under various typical operation modes
Collection carries out parameter optimization to the Transient Stability Evaluation model in step 4, keeps best model.To the on-line prediction in step 6
Data set obtains Stability Assessment result using the assessment models kept in step 4.Meaning of the present invention is as follows: this hair
It is bright based on one-dimensional convolutional neural networks technology, be easy to acquire data for input, taken using one-dimensional convolutional neural networks
Transient stability evaluation in power system model is built, the generalization and convergence speed of model are improved using Dropout technology and Adam algorithm
Degree.This method can be handled the disturbed timing response data of electric system, have certain engineering use value.
The present invention is further detailed below by one group of example.
This example is by taking 10 machine of New England, 39 node system as an example, and emulation is obtained 10200, sample, wherein stable sample
7714,2486, unstability sample, stabilization/unstability sample proportion is 3:1.Training set/test is divided by 8:2 to data set
Collection, each subset are concentrated from total sample by uniformly random sampling and are taken out, and guarantee stabilization/unstability sample proportion and population sample one
It causes.Mode input will be used as after the measurement data initialization process of system according to step 1 to step 3, whether by the stabilization of mark
Label updates model parameter as output, by training, keeps optimal models.Model is tested, assessment accuracy rate is obtained
Respectively 97.55% and 94.38%, it is compared with other methods, the mentioned method superiority with higher of the present invention.
The assessment result of 1 model of table
Tab.1Assessment results of models
From it can be seen that inventive algorithm has highest assessment performance under clean data set, accuracy rate and being recalled in table
Rate is respectively 97.55% and 94.38%, embodies the superior performance of model.Although the accuracy rate of service machine learning model is all
It can be seen that three kinds of machine learning models are more inclined in the case where sample proportion unevenness in 90% or more, but from recall rate index
To in the higher sample of the ratio that is classified as;ANN is higher compared to performance for other two kinds of models, but because its structure cannot be preferably
Using time series data, therefore its accuracy rate and recall rate are still not so good as model of the present invention.Specific embodiments described herein
Only illustrate as the applicating example of the method for the present invention, is not considered as limiting the invention, the technical field of the invention
Technical staff can within the scope of the invention be replaced specific embodiment, changes, supplements or modify.
Claims (6)
1. a kind of transient stability evaluation in power system method based on one-dimensional convolutional neural networks, which is characterized in that including following
Step:
Step 1: by time-domain-simulation method or the historical data collected using WAMS system to electric system as study
Sample set is divided into 0.0083s between systematic sampling, collects 3 cycles after failure removal always since occurring preceding 2 weeks wave points for failure
Data, using the active and reactive power of the voltage magnitude of bus and phase angle and branch as the input of sample characteristics, that is, model;
Step 2: data initialization processing is carried out to sample set;
Step 3: sample data is labeled according to generator's power and angle curve, stablizes sample and is labeled as 1, unstable sample mark
It is 0;
Step 4: Transient Stability Evaluation model of the building based on one-dimensional convolutional neural networks, and the data that step 3 acceptance of the bid is poured in
Collection is divided into training set and test set;
Step 5: steady to the transient state in step 4 as standard to assess accuracy rate using the training set and test set that are constructed in step 4
Determine assessment models and carry out optimizing, saves the model of optimum precision;
Step 6: the new electric power system data information that WAMS system acquisition obtains being handled using step 2 and step 4, is constructed
Assessment data set;
Step 7: assessment data set being assessed using the optimum precision model in step 5, obtains stability result.
2. Transient Stability Evaluation method according to claim 1, it is characterised in that: in the step 1, utilize transient emulation
Or WAMS historical data is divided into 0.0083s between systematic sampling as learning sample collection, one since occurring preceding 2 weeks wave points for failure
3 cycles after failure removal directly are collected, using the active and reactive power of the voltage magnitude of bus and phase angle and branch as sample
The input of feature, that is, model, wherein since fault clearing time is not a fixed constant, it will be with maximum failure removal
Zero padding operation is carried out to insufficient sequence on the basis of time.
3. appraisal procedure according to claim 1, it is characterised in that: the initial method in the step 2 will be to sample
Collection carries out initialization process, initializes formula are as follows:
WhereinIndicate the primitive character value of the t moment of initial characteristics jth column time series,Feature after indicating initialization
Value,WithRespectively indicate mean value and standard deviation that all sample jth in sample set arrange all characteristic values.
4. appraisal procedure according to claim 1, it is characterised in that: in the step 3, for each sample, foundation
Generator's power and angle curve to it carries out transient stability mark, its calculation formula is:
Wherein Δ δ max is any two generator's power and angles difference maximum value in the end 4s after disturbing, for each sample, if TSI
It is positive, then system is stable, and sample label is labeled as 1;If TSI is negative, system be it is unstable, sample label is labeled as
0。
5. Transient Stability Evaluation method according to claim 1, it is characterised in that: in the step 4, building based on complete
The transient stability evaluation in power system model of facial nerve network be one by two one-dimensional convolutional layer 1D-CONV1 and 1D-CONV2,
One one-dimensional pond layer 1D-POOL, an one-dimensional maximum pond layer 1D-MaxPOOL and full articulamentum FC composition.
6. appraisal procedure according to claim 1, it is characterised in that: by the data set under various typical operation modes,
Parameter optimization is carried out to the Transient Stability Evaluation model in step 4, best model is kept, to the on-line prediction data in step 6
Collection, using the assessment models kept in step 4, obtains Stability Assessment result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910187314.XA CN109886604A (en) | 2019-03-13 | 2019-03-13 | A kind of transient stability evaluation in power system method based on one-dimensional convolutional neural networks |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910187314.XA CN109886604A (en) | 2019-03-13 | 2019-03-13 | A kind of transient stability evaluation in power system method based on one-dimensional convolutional neural networks |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109886604A true CN109886604A (en) | 2019-06-14 |
Family
ID=66931947
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910187314.XA Pending CN109886604A (en) | 2019-03-13 | 2019-03-13 | A kind of transient stability evaluation in power system method based on one-dimensional convolutional neural networks |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109886604A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110163540A (en) * | 2019-06-28 | 2019-08-23 | 清华大学 | Electric power system transient stability prevention and control method and system |
CN110336277A (en) * | 2019-07-11 | 2019-10-15 | 福州大学 | A kind of transient stability evaluation in power system method based on depth confidence network |
CN110417005A (en) * | 2019-07-23 | 2019-11-05 | 清华大学 | In conjunction with the transient stability catastrophe failure screening technique of deep learning and simulation calculation |
CN110705831A (en) * | 2019-09-06 | 2020-01-17 | 华中科技大学 | Power angle instability mode pre-judgment model construction method after power system fault and application thereof |
CN110879917A (en) * | 2019-11-08 | 2020-03-13 | 北京交通大学 | Electric power system transient stability self-adaptive evaluation method based on transfer learning |
CN111626416A (en) * | 2020-04-24 | 2020-09-04 | 黑龙江瑞兴科技股份有限公司 | Automatic rail circuit fault diagnosis method based on deep convolutional neural network |
CN111797919A (en) * | 2020-06-30 | 2020-10-20 | 三峡大学 | Dynamic security assessment method based on principal component analysis and convolutional neural network |
CN112017070A (en) * | 2020-07-17 | 2020-12-01 | 中国电力科学研究院有限公司 | Method and system for evaluating transient stability of power system based on data enhancement |
CN112729834A (en) * | 2021-01-20 | 2021-04-30 | 北京理工大学 | Bearing fault diagnosis method, device and system |
CN116992255A (en) * | 2023-07-13 | 2023-11-03 | 华北电力大学 | Screening method and system for transient voltage stability characteristic quantity and electronic equipment |
-
2019
- 2019-03-13 CN CN201910187314.XA patent/CN109886604A/en active Pending
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110163540A (en) * | 2019-06-28 | 2019-08-23 | 清华大学 | Electric power system transient stability prevention and control method and system |
CN110336277A (en) * | 2019-07-11 | 2019-10-15 | 福州大学 | A kind of transient stability evaluation in power system method based on depth confidence network |
CN110336277B (en) * | 2019-07-11 | 2021-03-30 | 福州大学 | Power system transient stability evaluation method based on deep belief network |
CN110417005A (en) * | 2019-07-23 | 2019-11-05 | 清华大学 | In conjunction with the transient stability catastrophe failure screening technique of deep learning and simulation calculation |
CN110705831A (en) * | 2019-09-06 | 2020-01-17 | 华中科技大学 | Power angle instability mode pre-judgment model construction method after power system fault and application thereof |
CN110879917A (en) * | 2019-11-08 | 2020-03-13 | 北京交通大学 | Electric power system transient stability self-adaptive evaluation method based on transfer learning |
CN111626416A (en) * | 2020-04-24 | 2020-09-04 | 黑龙江瑞兴科技股份有限公司 | Automatic rail circuit fault diagnosis method based on deep convolutional neural network |
CN111797919A (en) * | 2020-06-30 | 2020-10-20 | 三峡大学 | Dynamic security assessment method based on principal component analysis and convolutional neural network |
CN112017070A (en) * | 2020-07-17 | 2020-12-01 | 中国电力科学研究院有限公司 | Method and system for evaluating transient stability of power system based on data enhancement |
CN112729834A (en) * | 2021-01-20 | 2021-04-30 | 北京理工大学 | Bearing fault diagnosis method, device and system |
CN112729834B (en) * | 2021-01-20 | 2022-05-10 | 北京理工大学 | Bearing fault diagnosis method, device and system |
CN116992255A (en) * | 2023-07-13 | 2023-11-03 | 华北电力大学 | Screening method and system for transient voltage stability characteristic quantity and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109886604A (en) | A kind of transient stability evaluation in power system method based on one-dimensional convolutional neural networks | |
CN108832619A (en) | Transient stability evaluation in power system method based on convolutional neural networks | |
CN109921414A (en) | A kind of transient stability evaluation in power system method based on deep learning Fusion Model | |
CN111914486B (en) | Power system transient stability evaluation method based on graph attention network | |
CN109960860A (en) | Transient stability evaluation in power system method based on differential evolution extreme learning machine | |
CN109063276B (en) | Wind power plant dynamic equivalent modeling method suitable for long-time domain wind speed fluctuation | |
CN109783997A (en) | A kind of transient stability evaluation in power system method based on deep neural network | |
CN102157938B (en) | Method for online identifying lower-voltage stability nodes of power system | |
CN105938578A (en) | Large-scale photovoltaic power station equivalent modeling method based on clustering analysis | |
CN110601174B (en) | Load modeling and online correction method based on deep learning | |
CN108667069A (en) | A kind of short-term wind power forecast method returned based on Partial Least Squares | |
CN106451418A (en) | Online clustering equivalent modeling method of photovoltaic power plant | |
CN109066651B (en) | Method for calculating limit transmission power of wind power-load scene | |
CN112564107A (en) | Transient stability assessment method for power system | |
CN109711435A (en) | A kind of support vector machines on-Line Voltage stability monitoring method based on genetic algorithm | |
CN113937764A (en) | Low-voltage distribution network high-frequency measurement data processing and topology identification method | |
CN113612237A (en) | Method for positioning resonance-induced subsynchronous oscillation source in offshore wind farm | |
CN111654392A (en) | Low-voltage distribution network topology identification method and system based on mutual information | |
CN112350318B (en) | AC power distribution network topology identification method based on breadth-first search algorithm | |
CN105914752A (en) | Pilot node selection method based on clustering by fast search and density peaks | |
CN105701265A (en) | Double-fed wind generator modeling method and apparatus | |
CN102946099B (en) | Method for real-time discrimination of transient power angle stability of multi-machine interconnected electric power system | |
CN111293693A (en) | Doubly-fed wind turbine converter control parameter identification method based on extended Kalman filtering | |
CN108092272A (en) | A kind of voltage stabilization on-line monitoring method based on the Kalman filtering that fades | |
CN106684854B (en) | A kind of active power distribution network voltage limit risk analysis method based on node equivalent |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190614 |