CN109799454A - Generator stator insulation residual breakdown strength prediction technique based on particle group optimizing - Google Patents
Generator stator insulation residual breakdown strength prediction technique based on particle group optimizing Download PDFInfo
- Publication number
- CN109799454A CN109799454A CN201811637846.0A CN201811637846A CN109799454A CN 109799454 A CN109799454 A CN 109799454A CN 201811637846 A CN201811637846 A CN 201811637846A CN 109799454 A CN109799454 A CN 109799454A
- Authority
- CN
- China
- Prior art keywords
- breakdown strength
- generator stator
- stator insulation
- data
- residual breakdown
- 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
- Tests Of Circuit Breakers, Generators, And Electric Motors (AREA)
Abstract
A kind of generator stator insulation residual breakdown strength prediction technique based on particle group optimizing, comprising the following steps: generator degradation the data obtained is subjected to correlation analysis first, existing dimension difference between different data is then eliminated by normalization;Data after normalization are classified, sample set data and inspection set data are divided into;Sample set data are obtained into optimal penalty parameter c and kernel functional parameter g using particle swarm optimization algorithm;Optimal penalty parameter c and kernel functional parameter g are substituted into the modeling of support vector machines generator stator insulation residual breakdown strength prediction model, support vector machines generator stator insulation residual breakdown strength prediction model is obtained;Double verification is carried out to obtained prediction model by sample set data and inspection set data, generator stator insulation residual breakdown strength is predicted using the prediction model after verifying.The present invention can accurately predict generator residual breakdown strength, provide reference for generator maintenance.
Description
Technical field
The invention belongs to operation state of generator detection fields, and in particular to a kind of power generation owner based on particle group optimizing
Insulate residual breakdown strength prediction technique, establishes prediction model by support vector machines, can provide effectively for the maintenance of generator
With reference to.
Background technique
Due to the effect by the multiple factors such as heat, electricity, environmental stress, it is old to will lead to insulation in the process of running for generator
Change, influences the normal operation of generator.Especially because developing on a large scale period at the beginning of China's 21 century in electric load, therefore at present
There is the operation time limit of large quantities of generators longer, there is more serious insulation ageing problem.It mostly uses and periodically stops in the past
The method of machine overhauling inspects periodically generator, and maintenance judges operation state of generator, certainly will will cause a large amount of power generation in this way
Machine load loss and the unnecessary wasting of resources.If can judge in advance operation state of generator and take phase to generator accordingly
Measure is answered, the operational efficiency of generator can be greatly improved, economic loss caused by reducing because of generator periodic maintenance.
Operation state of generator generally judged by generator stator insulation residual breakdown strength, the residue of generator
Disruptive field intensity is again and non-destructive electric parameter is in close relations.By the Accurate Prediction to generator stator insulation residual breakdown strength,
Effectively whether shutdown maintenance can be needed to provide reference for generator, there is biggish economic significance and practical value.
Summary of the invention
It is an object of the invention to be directed to above-mentioned the problems of the prior art, a kind of power generation based on particle group optimizing is provided
Owner's insulation residual breakdown strength prediction technique, can calculate to a nicety out generator stator insulation residual breakdown strength.
To achieve the goals above, the technical solution adopted by the present invention be the following steps are included:
Step 1: generator degradation the data obtained is carried out correlation analysis first, low non-broken of correlation is rejected
Bad property electric parameter, the high electric parameter of retention relationship;Then existing dimension difference between different data is eliminated by normalization;
Data after normalization are classified, sample set data and inspection set data are divided into;
Step 2: sample set data are obtained optimal penalty parameter c and kernel functional parameter g using particle swarm optimization algorithm;
Step 3: optimal penalty parameter c and kernel functional parameter g are substituted into the breakdown of support vector machines generator stator insulation residue
In the modeling of field-strength prediction model, support vector machines generator stator insulation residual breakdown strength prediction model is obtained;
Step 4: being hit by sample set data and inspection set data to obtained support vector machines generator stator insulation residue
It wears field-strength prediction model and carries out double verification, generator stator insulation residual breakdown strength is carried out using the prediction model after verifying
Prediction.
The step one carries out correlation point to generator degradation the data obtained using Wilson's relevance formula
Analysis:
In above formula, Cov (X, Y) indicates the covariance of X and Y, and Var [X] is the variance of X, and Var [Y] is the variance of Y, X generation
Table non-destructive electric parameter, Y indicate generator stator insulation residual breakdown strength.
Normalized function expression described in step 1 is as follows:
In formula, xiIndicate former data group, yiData after indicating normalization, xmaxFor a number maximum in former data group
According to xminFor a data the smallest in former data group, n is the data amount check in former data group.
Particle swarm optimization algorithm described in step 2, particle more new-standard cement are as follows:
Wherein, i=1,2...n, n are particle number, and 1≤d≤D, D are space dimensionality, and k is the number of iterations, wkIt is inertia
Coefficient indicates the weight to speed control;Accelerator coefficient is referred to as recognized, indicates the study to itself history optimum point;Claim
For social accelerator coefficient, the study to all particle history optimum points is indicated;WithIt is the random value on [0,1],
Every single-step iteration all generates at random;Iteration particle passes through the more new particle position individual extreme value pbest and global extremum gbest each time
It sets;
The update rule such as following formula of each coefficient:
In above formula, wini、wfin、c1ini、c1fin、c2ini、c2finRespectively inertia coeffeicent, cognition accelerator coefficient and society
The initial value and final value of meeting accelerator coefficient, kmaxFor the iterative steps upper limit.
W is taken in the coefficient update rule of particle swarm optimization algorithmini=0.9, wfin=0.4, c1ini=c2fin=2.5, c1fin
=c2ini=0.5.
Radial basis kernel function is used when predicting generator stator insulation residual breakdown strength, function expression is as follows:
K(x,xi)=exp (- γ | | x-xi0||2), γ > 0;
In formula, x is former data group, xi0For kernel function center, γ is kernel function width parameter, control function radial effect model
It encloses.
The step four first hits the generator stator insulation residue that sample set data substitution prediction model is predicted
Field strength is worn, compares, judges whether in the error range of permission with actual generator stator insulation residual breakdown strength, if not
Then return step two in the error range of permission change population initial velocity or penalty parameter c and kernel functional parameter g
Search Range to penalty parameter c and kernel functional parameter g optimizing, carries out the second step again if to the extent permitted by the error
Inspection set data are substituted into support vector machines generator stator insulation residual breakdown strength prediction model, repeat identical operation by card.
Compared with prior art, the present invention is with following the utility model has the advantages that first by generator degradation the data obtained
Correlation analysis is carried out, the determining and higher non-destructive electric parameter of the major insulation residual breakdown strength degree of correlation participates in generator
The prediction modeling of major insulation residual breakdown strength carries out data using normalization algorithm for the difference for eliminating dimension between data
Normalization, then obtains optimal support vector machines penalty parameter c and kernel functional parameter g by particle swarm optimization algorithm.Will
To optimal penalty parameter c and kernel functional parameter g substitution establish support vector machines generator stator insulation residual breakdown strength prediction
Model, and dual test is carried out to prediction model by sample set data and inspection set data, judge that established model is accurate
Whether degree reaches standard, if it does not meet the standards, then continues through particle swarm optimization algorithm and finds penalty parameter c and kernel functional parameter g
Optimal value.The present invention can accurately predict generator residual breakdown strength, provide reference for generator maintenance, have significant
Economy and practical value.
Detailed description of the invention
Particle swarm optimization algorithm flow chart Fig. 1 of the invention;
The flow chart of Fig. 2 forecast analysis generator stator insulation residual breakdown strength of the present invention;
Specific embodiment
Present invention will be described in further detail below with reference to the accompanying drawings.
Referring to fig. 2, the present invention is based on the generator stator insulation residual breakdown strength prediction technique of particle group optimizing include with
Lower step:
Step 1: generator degradation the data obtained is carried out correlation point using Wilson's relevance formula first
Analysis, the low non-destructive electric parameter of rejecting correlation, the high electric parameter of retention relationship, Wilson's relevance formula are as follows:
In above formula, Cov (X, Y) indicates the covariance of X and Y, and Var [X] is the variance of X, and Var [Y] is the variance of Y, X generation
Table non-destructive electric parameter, Y indicate generator stator insulation residual breakdown strength.
Then existing dimension difference between different data is eliminated by normalization, function expression is as follows:
In formula, xiIndicate former data group, yiData after indicating normalization, xmaxFor a number maximum in former data group
According to xminFor a data the smallest in former data group, n is the data amount check in former data group.
Data after normalization are classified, sample set data and inspection set data are divided into.
Step 2: sample set data are obtained optimal penalty parameter c and kernel functional parameter g using particle swarm optimization algorithm;
Referring to Fig. 1, first confirm that svmtrain function is the fitness function of particle swarm optimization algorithm, that is,
Svmtrain is support vector machines training function.Then to one initial velocity of population assignment, and initialization is calculated in advance with this
Survey accuracy rate and initialization optimal penalty parameter c and with function parameter g.Then particle group velocity is updated, global prediction is calculated
Accuracy rate is simultaneously and the optimal accuracy rate of history relatively obtains current optimal accuracy rate and current optimal penalty parameter c and kernel functional parameter
G next determines whether to reach termination condition i.e. the number of iterations.Current optimal penalty parameter c is exported if reaching termination condition
With kernel functional parameter g, parameter c, g at this time is exactly the optimal penalty parameter c obtained by population optimizing and kernel functional parameter
g.If not reaching termination condition, particle group velocity is updated, continually looks for optimal penalty parameter c and kernel functional parameter g.
Specifically, particle swarm optimization algorithm is initialized as a group random particles first, then obtained by continuous iteration
Optimal solution, iteration particle needs to use individual extreme value pbest and global extremum gbest and updates particle position each time.
Particle more new-standard cement is as follows:
Wherein, i=1,2...n, n are particle number, and 1≤d≤D, D are space dimensionality, and k is the number of iterations, wkIt is inertia
Coefficient indicates the weight to speed control;Accelerator coefficient is referred to as recognized, indicates the study to itself history optimum point;Claim
For social accelerator coefficient, the study to all particle history optimum points is indicated;WithIt is the random value on [0,1],
Every single-step iteration all generates at random;The update rule such as following formula of each coefficient:
In above formula, wini、wfin、c1ini、c1fin、c2ini、c2finRespectively inertia coeffeicent, cognition accelerator coefficient and society
The initial value and final value of meeting accelerator coefficient, kmaxFor the iterative steps upper limit.Generally desirable wini=0.9, wfin=0.4, c1ini=c2fin
=2.5, c1fin=c2ini=0.5.wkDescending change with iterations going on, this is conducive to population and jumps out in the early stage
Local optimum, later period carry out local optimal searching, and accelerating algorithm convergence improves search precision;WithRule change can make particle
Group is biased to itself optimum point in the early stage, and the later period is biased to all optimum points, is conducive to all populations in this way and converges on together in the later period
One value.
Step 3: optimal penalty parameter c and kernel functional parameter g are substituted into the breakdown of support vector machines generator stator insulation residue
In the modeling of field-strength prediction model, support vector machines generator stator insulation residual breakdown strength prediction model is obtained;
Step 4: being hit by sample set data and inspection set data to obtained support vector machines generator stator insulation residue
It wears field-strength prediction model and carries out double verification, the generator stator insulation for first being predicted sample set data substitution prediction model
Residual breakdown strength compares with actual generator stator insulation residual breakdown strength, judges whether the error range in permission
Interior, the return step two if not in the error range of permission changes population initial velocity or penalty parameter c and kernel function
The Search Range of parameter g carries out again to penalty parameter c and kernel functional parameter g optimizing if to the extent permitted by the error
Inspection set data are substituted into support vector machines generator stator insulation residual breakdown strength prediction model, repeated identical by two steps card
Operation.
Generator stator insulation residual breakdown strength is predicted using the prediction model after verifying, it is more smart in order to obtain
True prediction result uses Radial basis kernel function, function expression when predicting generator stator insulation residual breakdown strength
It is as follows:
K(x,xi)=exp (- γ | | x-xi0||2), γ > 0;
In formula, x is former data group, xi0For kernel function center, γ is kernel function width parameter, control function radial effect model
It encloses.
The advantages of basic principles and main features and technical solution of the invention can be brought has been shown and described above.
It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, described in above embodiments and description
Only basic principle of the invention, under the premise of not departing from spirit of that invention and principle, the present invention can also have various change
And replacement, these variations and alternatives can all be fallen within the protection scope that claims of the present invention and its equivalent are defined.
Claims (7)
1. a kind of generator stator insulation residual breakdown strength prediction technique based on particle group optimizing, which is characterized in that including step
It is rapid:
Step 1: generator degradation the data obtained is carried out correlation analysis first, the low non-destructive of correlation is rejected
Electric parameter, the high electric parameter of retention relationship;Then existing dimension difference between different data is eliminated by normalization;
Data after normalization are classified, sample set data and inspection set data are divided into;
Step 2: sample set data are obtained optimal penalty parameter c and kernel functional parameter g using particle swarm optimization algorithm;
Step 3: optimal penalty parameter c and kernel functional parameter g are substituted into support vector machines generator stator insulation residual breakdown strength
In the modeling of prediction model, support vector machines generator stator insulation residual breakdown strength prediction model is obtained;
Step 4: by sample set data and inspection set data to obtained support vector machines generator stator insulation residue breakdown field
Strong prediction model carries out double verification, is carried out using the prediction model after verifying to generator stator insulation residual breakdown strength pre-
It surveys.
2. the generator stator insulation residual breakdown strength prediction technique based on particle group optimizing according to claim 1, special
Sign is that step 1 carries out correlation analysis to generator degradation the data obtained using Wilson's relevance formula:
In above formula, Cov (X, Y) indicates the covariance of X and Y, and Var [X] is the variance of X, and Var [Y] is the variance of Y, and X represents non-
Destructive electric parameter, Y indicate generator stator insulation residual breakdown strength.
3. the generator stator insulation residual breakdown strength prediction technique based on particle group optimizing according to claim 1, special
Sign is that normalized function expression described in step 1 is as follows:
In formula, xiIndicate former data group, yiData after indicating normalization, xmaxFor a data maximum in former data group,
xminFor a data the smallest in former data group, n is the data amount check in former data group.
4. the generator stator insulation residual breakdown strength prediction technique based on particle group optimizing according to claim 1, special
Sign is, particle swarm optimization algorithm described in step 2, and particle more new-standard cement is as follows:
Wherein, i=1,2...n, n are particle number, and 1≤d≤D, D are space dimensionality, and k is the number of iterations, wkIt is inertia coeffeicent,
Indicate the weight to speed control;Accelerator coefficient is referred to as recognized, indicates the study to itself history optimum point;It is referred to as social
Accelerator coefficient indicates the study to all particle history optimum points;WithIt is the random value on [0,1], Mei Yibu
Iteration all generates at random;Iteration particle updates particle position by individual extreme value pbest and global extremum gbest each time;
The update rule such as following formula of each coefficient:
In above formula, wini、wfin、c1ini、c1fin、c2ini、c2finRespectively inertia coeffeicent, cognition accelerator coefficient and society accelerate
The initial value and final value of coefficient, kmaxFor the iterative steps upper limit.
5. the generator stator insulation residual breakdown strength prediction technique based on particle group optimizing according to claim 4, special
Sign is: taking w in the coefficient update rule of particle swarm optimization algorithmini=0.9, wfin=0.4, c1ini=c2fin=2.5, c1fin
=c2ini=0.5.
6. the generator stator insulation residual breakdown strength prediction technique based on particle group optimizing according to claim 4, special
Sign is, Radial basis kernel function is used when predicting generator stator insulation residual breakdown strength, and function expression is as follows:
K(x,xi)=exp (- γ | | x-xi0||2), γ > 0;
In formula, x is former data group, xi0For kernel function center, γ is kernel function width parameter, control function radial effect range.
7. the generator stator insulation residual breakdown strength prediction technique based on particle group optimizing according to claim 1, special
Sign is that the step four first punctures the generator stator insulation residue that sample set data substitution prediction model is predicted
Field strength compares with actual generator stator insulation residual breakdown strength, judges whether in the error range of permission, if not existing
Then return step two in the error range of permission change seeking for population initial velocity or penalty parameter c and kernel functional parameter g
Excellent range to penalty parameter c and kernel functional parameter g optimizing, carries out second step verifying again if to the extent permitted by the error,
Inspection set data are substituted into support vector machines generator stator insulation residual breakdown strength prediction model, repeat identical operation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811637846.0A CN109799454A (en) | 2018-12-29 | 2018-12-29 | Generator stator insulation residual breakdown strength prediction technique based on particle group optimizing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811637846.0A CN109799454A (en) | 2018-12-29 | 2018-12-29 | Generator stator insulation residual breakdown strength prediction technique based on particle group optimizing |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109799454A true CN109799454A (en) | 2019-05-24 |
Family
ID=66558267
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811637846.0A Pending CN109799454A (en) | 2018-12-29 | 2018-12-29 | Generator stator insulation residual breakdown strength prediction technique based on particle group optimizing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109799454A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111382546A (en) * | 2020-04-07 | 2020-07-07 | 东方电气集团东方电机有限公司 | Method for predicting service life of generator insulation system based on support vector machine modeling |
CN113884794A (en) * | 2021-09-28 | 2022-01-04 | 南方电网调峰调频发电有限公司检修试验分公司 | Method for evaluating main insulation aging state of motor wire bar |
CN113884839A (en) * | 2021-11-04 | 2022-01-04 | 国网四川省电力公司电力科学研究院 | Multi-parameter insulation state evaluation method and system for capacitor voltage transformer |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1402015A (en) * | 2002-09-13 | 2003-03-12 | 西安交通大学 | Method and device for diagnosis of insulation ageing of electric machine based on small wave conversion |
CN102680893A (en) * | 2012-05-22 | 2012-09-19 | 李朝晖 | Main insulation fault real-time online early warning device and method for generator |
CN103678941A (en) * | 2013-12-31 | 2014-03-26 | 武汉大学 | Prediction method for electrode air gap breakdown voltage |
CN109063400A (en) * | 2018-10-17 | 2018-12-21 | 苏州科技大学 | Generator stator bar insulating thermal aging lifetime estimation method based on statistical check |
-
2018
- 2018-12-29 CN CN201811637846.0A patent/CN109799454A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1402015A (en) * | 2002-09-13 | 2003-03-12 | 西安交通大学 | Method and device for diagnosis of insulation ageing of electric machine based on small wave conversion |
CN102680893A (en) * | 2012-05-22 | 2012-09-19 | 李朝晖 | Main insulation fault real-time online early warning device and method for generator |
CN103678941A (en) * | 2013-12-31 | 2014-03-26 | 武汉大学 | Prediction method for electrode air gap breakdown voltage |
CN109063400A (en) * | 2018-10-17 | 2018-12-21 | 苏州科技大学 | Generator stator bar insulating thermal aging lifetime estimation method based on statistical check |
Non-Patent Citations (2)
Title |
---|
CHUNGUANG HOU 等: "Prediction of synchronous closing time of permanent magnetic actuator for vacuum circuit breaker based on PSO-BP", 《 IEEE TRANSACTIONS ON DIELECTRICS AND ELECTRICAL INSULATION》 * |
舒胜文: "基于电场特征集和支持向量机的空气间隙起晕和击穿电压预测研究", 《中国博士学位论文全文数据库》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111382546A (en) * | 2020-04-07 | 2020-07-07 | 东方电气集团东方电机有限公司 | Method for predicting service life of generator insulation system based on support vector machine modeling |
CN113884794A (en) * | 2021-09-28 | 2022-01-04 | 南方电网调峰调频发电有限公司检修试验分公司 | Method for evaluating main insulation aging state of motor wire bar |
CN113884839A (en) * | 2021-11-04 | 2022-01-04 | 国网四川省电力公司电力科学研究院 | Multi-parameter insulation state evaluation method and system for capacitor voltage transformer |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109799454A (en) | Generator stator insulation residual breakdown strength prediction technique based on particle group optimizing | |
CN110400231B (en) | Failure rate estimation method for electric energy metering equipment based on weighted nonlinear Bayes | |
CN104166790A (en) | Lithium-ion battery performance evaluation method based on TOPSIS theory | |
CN111628494B (en) | Low-voltage distribution network topology identification method and system based on logistic regression method | |
CN104915638A (en) | Least squares support vector machine electric shock current detection method based on parameter optimization | |
CN109298225B (en) | Automatic identification model system and method for abnormal state of voltage measurement data | |
CN108336739A (en) | A kind of Probabilistic Load Flow on-line calculation method based on RBF neural | |
Moradifar et al. | Intelligent localisation of multiple non‐linear loads considering impact of harmonic state estimation accuracy | |
CN109742788A (en) | A kind of grid-connected Performance Evaluating Indexes modification method of new energy power station | |
CN109359665A (en) | A kind of family's electric load recognition methods and device based on support vector machines | |
Guo et al. | Parameter identification for static var compensator model using sensitivity analysis and improved whale optimization algorithm | |
CN103324858A (en) | Three-phase load flow state estimation method of power distribution network | |
CN112070118A (en) | Station area phase sequence identification method based on data clustering fitting analysis | |
CN111046913A (en) | Load abnormal value identification method | |
CN109921426A (en) | Wind-electricity integration system probability load flow calculation method based on CV-KDE | |
Zhao et al. | Research on state of health for the series battery module based on the Weibull distribution | |
CN108574290B (en) | Method, device, terminal and readable storage medium for positioning oscillation source of forced oscillation | |
Zhu et al. | A novel fast estimation and regroup method of retired lithium‐ion battery cells | |
CN109858667A (en) | It is a kind of based on thunder and lightning weather to the short term clustering method of loading effects | |
CN113408192A (en) | Intelligent electric meter error prediction method based on GA-FSVR | |
He et al. | Power system frequency situation prediction method based on transfer learning | |
Luo | [Retracted] Application of Reinforcement Learning Algorithm Model in Gas Path Fault Intelligent Diagnosis of Gas Turbine | |
CN116502149A (en) | Low-voltage power distribution network user-transformation relation identification method and system based on current characteristic conduction | |
CN110969214A (en) | Transient state security domain online construction method based on support vector machine comprehensive model | |
Liu et al. | A Density-Based Spatial Clustering of Application with Noise Algorithm and its Empirical Research |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190524 |
|
RJ01 | Rejection of invention patent application after publication |