CN106844952A - Based on the generator dynamic state estimator method without mark Particle filtering theory - Google Patents
Based on the generator dynamic state estimator method without mark Particle filtering theory Download PDFInfo
- Publication number
- CN106844952A CN106844952A CN201710040475.7A CN201710040475A CN106844952A CN 106844952 A CN106844952 A CN 106844952A CN 201710040475 A CN201710040475 A CN 201710040475A CN 106844952 A CN106844952 A CN 106844952A
- Authority
- CN
- China
- Prior art keywords
- generator
- particle
- state
- filtering
- state estimator
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Control Of Eletrric Generators (AREA)
Abstract
The invention discloses a kind of based on the generator dynamic state estimator method without mark Particle filtering theory, generator quadravalence dynamical equation is used first as the state equation of generator dynamic state estimator, then the metric data such as generator rotor angle, the angular speed of generator are obtained using power system analysis software simulation PMU devices, and sets up the measurement equation of generator.By obtaining the static estimation value of state estimation initial time as the initial value of generator dynamical state Startup time, predecessor is generated near initial value, and filtering is tracked to quantity of states such as generator rotor angle, the angular speed of generator using without mark particle filter algorithm, finally obtain the estimate of Generator Status amount.Method proposed by the present invention reduces the demand to particle, filtering accuracy and computational efficiency are superior to traditional particle filter method, the inventive method increased the dispersiveness of particle simultaneously so that the robustness of method proposed by the present invention is better than traditional particle filter method and Unscented kalman filtering method.
Description
Technical field
Invention is related to a kind of based on the generator dynamic state estimator method without mark Particle filtering theory, belongs to power system
Monitoring, analysis and control technology field.
Background technology
As the increase of power network scale and power system complexity increase, simultaneously because there is measurement in existing measuring apparatus
Error, direct measurement means are difficult to obtain the real status information of power system.Power system time of day is power system point
The important references of analysis, control and decision-making, and state estimation can filter the error of electrical power system metric data, be connect as far as possible
The approximation of nearly system time of day.
Traditional state estimation is obtained frequently with the static state method of estimation iterative with least-squares algorithm as representative
To the state approximation of power system sometime section.Because power system scale is big, complexity is high, electrical power system transient event
Barrier is difficult to avoid that, therefore, it is badly in need of finding precision and computational efficiency method for estimating state higher so as to faster obtain more accurate
Status information, shorten failure occur to take control protection time.Estimate that dynamic state estimator is not compared to static state
Can only filter measurement noise, and with good predictive ability, can be the security evaluation of power system, status predication, pre-
The On-line funchons such as anti-control provide support.And the PMU equipment in WAMS can provide the amount that high accuracy, high frequency refresh
Data are surveyed, also for the real-time of dynamic state estimator provides powerful guarantee.
The wave filter for setting up rational generator dynamic model and selection function admirable is electric system generator dynamic shape
The top priority that state is estimated.For generator dynamic model, different researchers have selected the generating of different orders according to Research Requirements
Motor-driven state establishing equation generator dynamic model, is broadly divided into second order, quadravalence and six rank dynamical equations.And consider to generate electricity motor-driven
States model it is non-linear, multidigit scholar propose respectively based under the framework such as kalman estimate and Bayesian Estimation filtering calculate
Method, main representative has EKF (EKF), Unscented kalman filtering (UKF), volume Kalman filtering (CKF), particle
Filtering (PF) etc..EKF has that linearisation truncated error is big, and UKF and CKF uses different sample modes, using sampling
Point propagates the average and variance of nonlinear equation, without being linearized.However, UKF is sensitive to initial value, parameter is chosen without determination
Principle, though CKF is without selection parameter, filtering lifting is limited.PF filtering accuracies are high, but a large amount of particles of needs carry out computing, count
Calculate efficiency low, and when Chong Die with likelihood function less priori or measurement model precision higher is predicted, filtering may be caused to fail.
The content of the invention:
The technical problems to be solved by the invention are directed to the deficiency of prior art presence and provide a kind of being based on without mark grain
The generator dynamic state estimator method of sub- filtering theory.
The present invention to achieve the above object, is adopted the following technical scheme that:
It is a kind of based on the generator dynamic state estimator method without mark Particle filtering theory, methods described is in a computer
Realize according to the following steps successively:
1) parameter information of generating set in the system of dynamic state estimator needed for obtaining;
2) metric data needed for obtaining state estimation using power system analysis software simulation PMU equipment;
3) state estimator initialization;
4) generator dynamic state estimator model is set up;The state of generator is set up using the quadravalence dynamical equation of generator
Equation, the measurement equation of generator is set up according to the PMU data for obtaining;
5) primary is generated near state initial value, starts filtering algorithm;
6) UKF generation the importance density function and the new sampling particle of passing ratio amendment sampling;
7) update particle weights and normalize;
8) judge whether to need resampling;If so, resampling steps are then gone to, if it is not, then continuing next step.
9) state estimation result at current time is exported.
10) judge whether filtering operation terminates, if so, then exporting last whole state estimation result;If it is not, then going to step
It is rapid 6) to continue next step.
Step 1) in, the parameter information of generator includes machine torque, active and idle rated power, the synchronization of generator
The unit sum of angular rate, damped coefficient, inertia time constant, stator excitation voltage rating and generator.
Step 2) in metric data include:The absolute generator rotor angle of generator, angular speed changing value, the width of port voltage phasor
Value and phase angle, port electromagnetic power and machine torque.
Step 3) in initialization include |input paramete and metric data, state initial value, setting up procedure noise and amount are set
Noise covariance battle array is surveyed, prediction covariance initial value is set, UPF populations and filtering parameter and sampling interval and sampling week are set
Phase;
Step 4) in the state equation form of generator be:
In formula, δ is the absolute generator rotor angle (radian) of generator amature under dq0 coordinate systems, ω0It is specified synchronous rotational speed (electric arc
Degrees second), Δ ω is generator angular speed variable quantity (perunit value), and TJ is generator time inertia constant, and D is Generator Damping
Coefficient, EfIt is stator excitation voltage, TmIt is the machine torque (perunit value) of generator, Te(Te=Pe/ ω) it is the electromagnetism of generator
Torque (perunit value), ignores stator impedance and assumes ω ≈ 1, obtains Te≈Pe(PeIt is the electromagnetic power of generator), E 'dWith E 'qPoint
Not Wei generator d axles and q axle transient internal voltages, U andIt is the amplitude and phase angle of generator port voltage, XdAnd XqRespectively
The synchronous reactance of generator d axles and q axles, X 'dWith X 'qThe respectively transient state reactance of generator d axles and q axles, T 'd0With T 'q0Respectively
It is generator d axles and the transient state open circuit time constant of q axles;
Generator measurement is respectively the absolute generator rotor angle of generator, angular speed, electromagnetic power, is directly obtained by PMU equipment,
I.e.:Y=[δz,ωz,Pez]T=[y1,y2,y3]T, generator measurement equation form is:
Step 6) in proportion of utilization sampling UKF generation the importance density function include:
The weights of ratio amendment sampling meet:
In formula:λ=α2(L+ κ)-L is fine setting parameter, and for the distance at control point to average, wherein L is state variable
Number, κ is secondary decimation factor, is taken as 3-L;α is ratio modifying factor, is taken as 0.12;β is parameter to be selected, and regulation β improves variance
Precision, takes β=2.
The step of generating the importance density function by UKF includes:
New sampled point is generated by the state filtering value and filtering covariance matrix of each particle of last moment, is substituted into and is generated electricity
The state equation of machine, obtains first step status predication average and first step prediction covariance matrix of each particle in UKF;
New sampled point is generated by the first step predicted value of each particle, the measurement using measurement equation and current time is believed
Breath corrects second step status predication average and second step prediction covariance matrix of each particle in UKF;
According to the importance density function in second step status predication average and second step prediction covariance matrix generation PF
And generate new sampling particle;
Step 8) in the effective particle threshold for judging resampling elect 1/3rd of total population as, method for resampling choosing
For amount of calculation is smaller, residual error resampling of good performance.
Beneficial effect:The present invention is compared with prior art:Generator Status amount in electromechanical transient process can be provided
More accurately estimate, improve the degree of accuracy to the filtering of Generator Status amount.Method proposed by the present invention is reduced to particle
Demand, computational efficiency is better than traditional PF estimators, while increased the dispersiveness of particle so that method proposed by the present invention
Robustness also superior to traditional PF estimators, and method proposed by the present invention can adjust filtering accuracy, filtering according to population
Precision and flexibility are superior to UKF estimators.
Brief description of the drawings:
Fig. 1 is flow chart of the method for the present invention;
Fig. 2 is the node system schematic diagram of tri- machines of WSCC nine;
Fig. 3 is the test system based on the node system of tri- machines of WSCC nine, and the generator rotor angle of generator 1 is estimated under the inventive method
The comparison diagram of curve and BPA simulation softwares actual value curve;
Fig. 4 is the test system based on the node system of tri- machines of WSCC nine, and the angular speed of generator 1 is estimated under the inventive method
The comparison diagram of index contour and BPA simulation softwares actual value curve;
Fig. 5 is the test system based on the node system of tri- machines of WSCC nine, the q axle transient state of generator 1 under the inventive method
The comparison diagram of electromotive force estimation curve and BPA simulation softwares actual value curve;
Fig. 6 is the test system based on the node system of tri- machines of WSCC nine, the d axle transient state of generator 1 under the inventive method
The comparison diagram of electromotive force estimation curve and BPA simulation softwares actual value curve;
Fig. 7 is the test system based on the node system of tri- machines of WSCC nine, and the generator rotor angle of generator 2 is estimated under the inventive method
The comparison diagram of curve and BPA simulation softwares actual value curve;
Fig. 8 is the test system based on the node system of tri- machines of WSCC nine, and the angular speed of generator 2 is estimated under the inventive method
The comparison diagram of index contour and BPA simulation softwares actual value curve;
Fig. 9 is the test system based on the node system of tri- machines of WSCC nine, the q axle transient state of generator 2 under the inventive method
The comparison diagram of electromotive force estimation curve and BPA simulation softwares actual value curve;
Figure 10 is the test system based on the node system of tri- machines of WSCC nine, the d axle transient state of generator 2 under the inventive method
The comparison diagram of electromotive force estimation curve and BPA simulation softwares actual value curve;
Figure 11 is the test system based on the node system of tri- machines of WSCC nine, and the generator rotor angle of generator 3 is estimated under the inventive method
The comparison diagram of index contour and BPA simulation softwares actual value curve;
Figure 12 is the test system based on the node system of tri- machines of WSCC nine, the angular speed of generator 3 under the inventive method
The comparison diagram of estimation curve and BPA simulation softwares actual value curve;
Figure 13 is the test system based on the node system of tri- machines of WSCC nine, the q axle transient state of generator 3 under the inventive method
The comparison diagram of electromotive force estimation curve and BPA simulation softwares actual value curve;
Figure 14 is the test system based on the node system of tri- machines of WSCC nine, the d axle transient state of generator 3 under the inventive method
The comparison diagram of electromotive force estimation curve and BPA simulation softwares actual value curve;
Figure 15 is the test system based on the node system of tri- machines of WSCC nine, and the generator rotor angle of generator 1 is estimated under the inventive method
The contrast schematic diagram of index contour and UKF, PF algorithm;
Figure 16 is the test system based on the node system of tri- machines of WSCC nine, the angular speed of generator 1 under the inventive method
The contrast schematic diagram of estimation curve and UKF, PF algorithm;
Figure 17 is the test system based on the node system of tri- machines of WSCC nine, the q axle transient state of generator 1 under the inventive method
The contrast schematic diagram of electromotive force estimation curve and UKF, PF algorithm;
Figure 18 is the test system based on the node system of tri- machines of WSCC nine, the d axle transient state of generator 1 under the inventive method
The contrast schematic diagram of electromotive force estimation curve and UKF, PF algorithm;
Figure 19 is domestic certain actual bulk power system to there is an inclined noise measuring system, and the inventive method different is made an uproar at four kinds
To the generator rotor angle estimation curve and the contrast schematic diagram of BPA simulation softwares actual value curve of generator in the case of sound;
Figure 20 is domestic certain actual bulk power system to there is an inclined noise measuring system, and the inventive method different is made an uproar at four kinds
To the Attitude rate estimator curve and the contrast schematic diagram of BPA simulation softwares actual value curve of generator in the case of sound;
Figure 21 is domestic certain actual bulk power system to there is an inclined noise measuring system, and the inventive method different is made an uproar at four kinds
To the q axle transient internal voltage estimation curves and the contrast schematic diagram of BPA simulation softwares actual value curve of generator in the case of sound;
Figure 22 is domestic certain actual bulk power system to there is an inclined noise measuring system, and the inventive method different is made an uproar at four kinds
To the d axle transient internal voltage estimation curves and the contrast schematic diagram of BPA simulation softwares actual value curve of generator in the case of sound.
Specific embodiment:
The techniqueflow invented is described in detail below in conjunction with the accompanying drawings:
Dynamic state estimator
Dynamic state estimator carries out data acquisition using power system analysis software BPA simulations PMU equipment first, and will
BPA operation results are by conversion and calculate as actual value, actual value is superimposed random error and is estimated as measuring value feeding state
Gauge.
In recent years, the extensive utilization of PMU equipment brings change to Electrical Power System Dynamic state estimation.PMU equipment can be provided
In high precision, the metric data that high frequency refreshes, for the real-time of dynamic state estimator provides guarantee.Because generator amature is present
Inertia, the quantity of state such as generator rotor angle will not undergo mutation in the dynamic process of generator.The filter of known generators previous moment quantity of state
Wave number, utilization state equation obtains current time predicted value, with reference to PMU measuring values, can obtain the filter value of current time quantity of state,
The recursion on time shaft is realized with this, the dynamic state estimator result of generator is finally given.The state equation and amount of generator
Survey equation form be:
In formula, x, u, y are respectively defined as state variable, control variables and measure variable, and f is Generator Status equation, and h is
Generator measurement equation, w is process noise, and v is to measure noise, generally assumes that w~N (0, Q), v~N (0, R), wherein Q and R point
Not Wei w and v error covariance matrix, w and v is separate and independently of state variable.
The discrete form of Generator Status equation and measurement equation is:
In formula, F (xk,uk,wk)=xk+f(xk,uk,wk) Δ t, k be sampling instant, Δ t is sampling step length.
The model of generator
The dynamic state estimator model of generator is made up of state equation and measurement equation, the number of Generator Status equation
That is the exponent number of generator.Because generator time transient process is of short duration, existing PMU equipment is difficult to the accurate time transient process that obtains and measures
Amount, ignores D, Q winding corresponding with secondary transient process and stator dynamic process, by generator model dimensionality reduction, obtains synchronous generator
Machine Nonlinear Fourth Order model.The phasors such as voltage, the electric current of generator outlet being measured using PMU equipment, generator is connected
External electrical network is decoupled, and Generator Status is carried out dynamic state estimator independently of power network.Generator dynamical state is estimated
The state equation of meter is as follows:
In formula, δ is the absolute generator rotor angle (radian) of generator amature under dq0 coordinate systems, ω0It is specified synchronous rotational speed (electric arc
Degrees second), Δ ω is generator angular speed variable quantity (perunit value), and TJ is generator time inertia constant, and D is Generator Damping
Coefficient, EfIt is stator excitation voltage, TmIt is the machine torque (perunit value) of generator, Te(Te=Pe/ ω) it is the electromagnetism of generator
Torque (perunit value), ignores stator impedance and assumes ω ≈ 1, obtains Te≈Pe(PeIt is the electromagnetic power of generator), E 'dWith E 'qPoint
Not Wei generator d axles and q axle transient internal voltages, U andIt is the amplitude and phase angle of generator port voltage, XdAnd XqRespectively
The synchronous reactance of generator d axles and q axles, X 'dWith X 'qThe respectively transient state reactance of generator d axles and q axles, T 'd0With T 'q0Respectively
It is generator d axles and the transient state open circuit time constant of q axles;
Measurement y=[δz,ωz,Pez]T=[y1,y2,y3]T, the measurement equation of dynamic state estimator is as follows:
The step of the inventive method
Implement dynamic state estimator using UPF algorithms.UPF is based on PF frameworks, and the importance density function generation is generated using UKF
For the importance density function in PF, algorithm is comprised the following steps that:
Step one:Initialization:
The input information such as generator parameter and two side datas, program initialization, starting state estimator makes k=0, first
Beginning state variable x0Predecessor collection is nearby generated,Corresponding weight coefficient
(M is particle number).
Work as k>When 0, k=k+1 is made, the one group of new particle collection that will be obtained in upper step iterationUnder substitution
One step.Step 2:The importance density function and sampling particle are generated using UKF:
IfWithIt is the average and covariance of the quantity of state at each particle k moment, is sampled using the amendment of sigma point ratios
Mode, constructs weights and sigma sampled points:
In formula:λ=α2(L+ κ)-L is fine setting parameter, and for the distance at control point to average, wherein L is state variable
Number, κ is secondary decimation factor, is taken as 3-L;α is ratio modifying factor, is taken as 0.12;β is parameter to be selected, and regulation β improves variance
Precision, β=2.
In formula,It is matrixThe i-th row (P=ATA, A are that P is obtained by the decomposition of Collins's base
Low order triangular matrix) or i row (P=AAT)。
Each particle is predicted, quantity of state one-step prediction value is calculatedAnd its prediction covariance
According toWithEach particle sigma sampled points are constructed again:
Each particle state and variance are updated:
So far, the estimate of the current time particle is obtained using UKF algorithms to each particleAnd covarianceFrom
And obtain the importance density function and be:
Wherein N () is the probability density function of Gaussian distributed.
Then the sampling particle in importance density function is:
Step 3:Calculate weights and normalize:
Calculate weights:
Normalization:
Step 4:Judge whether to need to carry out resampling, if so, then enter resampling steps, if not into next step:
CalculateIf(NtIt is the threshold value of setting, is typically taken as M/3) then illustrate particle
Weights have been degenerated serious, it is necessary to carry out resampling, are otherwise directly entered next step.During resampling, by weights height replicate with
Particle is reset, the number of duplication is directly proportional to its weights, reject the small particle of weights, and the particle after treatment is mapped as etc. to weigh
M particle of weight, i.e.,
Step 5:Output state estimateWith covariance matrix Pk+1:
Step 6:Judge whether circulation terminates, if so, then exporting whole results, filtering terminates, if it is not, being transferred to step 2.
The advantage of the inventive method
Because EKF truncated errors are big, and UKF processes nonlinear function using sampled point, although remain the non-thread of function
Property, filtering accuracy is high, but UKF initial values choose it is inaccurate convergence may be caused not restrain even slowly, and the selection of UKF parameters is scarce
Weary strict theoretical proof, its filter effect is restricted by parameter, it is impossible to which different according to filtering occasion require adjustment filtering accuracy,
Lack flexibility.PF based on bayesian theory can change population adjustment filtering accuracy, and with the increase of population, estimate
Value can infinite approach actual value.But PF is scanned for using substantial amounts of particle to state space, increase computation burden, and PF with
Prior probability for suggestion be distributed, when likelihood function just prior probability distribution afterbody or likelihood function it is too narrow (due to
Spike caused by error in measurement very little) when, filtering is easy for dissipating or fails.
Therefore, the inventive method has done following improvement:
First, the inventive method obtains generator metric data using PMU equipment, realizes generator and external electrical network solution
Coupling, reduces transmission error influence.Secondly, UPF be devoted to solve Kalman's framework under filtering method to strong nonlinearity non-gaussian
The filtering accuracy of system is limited, and tradition PF easily particle occurs and moves back in sampling process because the importance density function is inaccurate
The problem of change.Using the importance density function in UKF generations PF, it possesses optimal the importance density function to the inventive method
Produce the two major features of prediction particle:One is the particle information at existing current time, and two is using newest metric data.Cause
This, the inventive method is Chong Die more with true posterior probability density function support using the importance density function that UKF is produced, by
During the importance density function is generated, the inventive method has used newest measurement information, can be by particle transfer extremely
Likelihood region high, reduces the particle demand of description posterior probability density, improves filtering accuracy and filtration efficiency.Additionally,
The inventive method be easy to sampling realize, again introduce resampling, increased the weight of effective particle, estimated accuracy is high, can accurately with
The change of track quantity of state, and there is robustness to noise.
Embodiment
The example of present invention test is the node system of tri- machines of WSCC nine and certain 224 node Hainan Power Grid system.The machines of WSCC tri-
Nine node system node metric data are that the true value superposition random noise simulation of BPA simulation softwares is obtained, and generator is used during emulation
Quadravalence model, and consider that the water turbine governors of generator 2 are acted on, obtain the actual value curve of the machine torque of generator 2.Simulated fault
The 40th cycle to the 45th cycle (i.e. 0.8~0.9s) is set to, three-phase metallicity occurs short at Bus5-Bus7 branch road line outlets
Road, then fault clearance.Generator through being transitioned into new stable state after a period of time.A length of 6s during emulation.
In order to verify the performance of the inventive method, tracking velocity UKF and filtering accuracy PF conducts higher faster is selected
The control methods of the method for the present invention, the performance of three kinds of algorithms is contrasted.
In order to more intuitively compare three kinds of filter effects of algorithm, average relative error is chosenAnd maximum absolute error
xmAs the Performance Evaluating Indexes that algorithm is contrasted.I.e.:
In formula, N is the number of times of sampling, and it is the actual value of i & lt sampling to be taken as 300, xi,It is the estimation of i & lt filtering
Value.
Fig. 3~Fig. 5 is the quantity of state estimate and BPA simulation results pair of three generators in the node system of tri- machines of WSCC nine
Than figure.It can be seen that at 0~40 cycle (during systematic steady state), UPF algorithms being capable of Fast Convergent, real-time tracking, when system occurs suddenly
Transient fault, each state variable of generator when UPF remains to accurate tracking failure and after failure, estimated accuracy meets will
Ask.
Fig. 7~Figure 10 is that filtering of four state variables of generator 1 under the inventive method and UKF and PF algorithms is bent
Line.Wherein in UKF algorithms, its filtering covariance initial value is set to unit matrix, using initial value during Generator Stable as state variable
Initial value.κ=3-L in UKF, α=0.12, β=2.Parameter and covariance set identical with UKF in UPF, in this test UPF and
PF uses residual error resampling method, population to be 100.It can be seen that, UPF algorithms filter start after can Fast Convergent, therefore
Barrier can accurately track Generator Status amount before and after occurring, and filtering accuracy is high.UKF algorithms are restrained relatively slowly after startup is filtered, therefore
Barrier can not be tracked accurately after occurring, and filtering accuracy is less than UPF.PF algorithms can be compared with rapid convergence, in fault moment after startup is filtered
Also can real-time tracking, but filtering accuracy be less than UPF.
Three kinds of achievement datas of algorithm understand that UPF is in average relative error and maximum absolute error index in comparison sheet 1
Filter effect is significantly better than that UKF and PF.UPF filtering accuracies are high, can accurately track generator state variables, in same parameter
Lower filtering accuracy is set and is better than UKF, filtering accuracy is also significantly better than PF under same population.
To study influence of the different populations to the inventive method filter effect, selection numbering is in actual electric network system
The Generator Status of HAID4G are tracked and filter, and study the inventive method filtering performance under different populations.By table 2
It can be seen that, the filtering accuracy of the inventive method is as the increase of population is in the excellent trend of change, but it is while filtering accuracy is improved
There is a problem that amount of calculation constantly increases.The inventive method can be according to the different particle of different filtering accuracy requirement selections
Number, i.e., choose a small amount of particle to reduce computation burden in the case where filtering accuracy is relatively low, higher and right in filtering accuracy
Calculating the little occasion of time requirement can suitably increase population guarantee precision, and flexibility is strong.PF is reaching the inventive method
Need to use more particles during identical filtering accuracy, amount of calculation is also accordingly increased, computation burden becomes weight.Test shows, phase
Than in PF, in the case where same accuracy requirement is reached, the amount of calculation of the inventive method is about the half of PF, than PF efficiency high.
In order to verify the inventive method variance increase have inclined noise under filter effect, in the actual electric network of Hainan
It is that the Generator Status of HAID4G are tracked and filter to choose numbering, and by UKF and filtering of the PF under the conditions of same noise
Performance is contrasted with the inventive method.UPF is 100 with PF populations in this test.Certain hair is set in 50 cycle
Motor egress line three phase short circuit fault.The Gauss that measurement noise is set to the white Gaussian noise and average non-zero for incrementally increasing makes an uproar
Sound, is following 4 kinds of situations:Case1, average is 0, and standard deviation is δ0;Case2, average is 0, and standard deviation is 3 δ0;Case3, average
It is E0, standard deviation is δ0;Case4, average is 3E0, standard deviation is 3 δ0。
From table 3, as the increase of noise criteria difference or noise average displaced from zero yardstick increase, under three kinds of filtering methods
The average relative evaluated error of four state variables of generator is incrementally increased, wherein, UPF filtering effects are presented in first three situation
Fruit is better than PF, situation of the PF filter effects better than UKF, and works as noise average displaced from zero farther out, and error to standard deviation is also increased to
At 3 times of primary standard difference, filtering easily occurs in the PF under 100 particles can not accurately be tracked, and filtering part curve is bent into constant
Line, filtering accuracy is drastically deteriorated, now filtering failure.This be due to alternator failure during process noise especially therefore
Process noise during the process noise of barrier initial time period is much larger than stabilization and after failure vanishes.PF underestimates in failure initial time period
The process-noise variance of prediction step, causes the covariance of prior state amount process noise less than real processes noise covariance,
Particle it is dispersed weak, noise have partially and the double influence of process noise deviation under, the PF filter effect urgency under 100 particles
Drastic change is bad.The filter effect of PF can accordingly be improved by the population for increasing the process noise covariance of PF and increase PF.UPF
Newest measurement information is make use of by selecting UKF to generate the importance density function, filtering accuracy is higher, and is filtered in UPF
Cheng Zhong, sigma point samplings are carried out to each particle, increased the dispersiveness of particle, therefore are made an uproar by process in failure process
The influence of sound is smaller.Found by three kinds of contrasts of method, the filtering accuracy of UPF is in standard deviation increase or noise average non-zero etc.
Remain to meet filtering accuracy and convergence requirement under noise situations, there is stronger robustness to noise.
Generator dynamic state estimator evaluation index under the algorithms of different of table 1
The different populations of table 2 issue motor dynamics state estimation evaluation index
The lower 3 kinds of algorithms dynamic state estimator evaluation index of different noises of table 3
Claims (7)
1. a kind of based on the generator dynamic state estimator method without mark Particle filtering theory, it is characterised in that:
1) parameter information of generating set in the system of dynamic state estimator needed for obtaining;
2) metric data needed for obtaining state estimation using power system analysis software simulation PMU equipment;
3) state estimator initialization;
4) generator dynamic state estimator model is set up;The state side of generator is set up using the quadravalence dynamical equation of generator
Journey, the measurement equation of generator is set up according to the PMU data for obtaining;
5) primary is generated near state initial value, starts filtering algorithm;
6) UKF generation the importance density function and the new sampling particle of passing ratio amendment sampling;
7) update particle weights and normalize;
8) judge whether to need resampling;If so, resampling steps are then gone to, if it is not, then continuing next step;
9) state estimation result at current time is exported;
10) judge whether filtering operation terminates, if so, then exporting last whole state estimation result;If it is not, then going to step 6)
Continue next step.
2. according to claim 1 based on the generator dynamic state estimator method without mark Particle filtering theory, its feature
It is:Step 1) in, the parameter information of generator includes the machine torque of generator, active and idle rated power, synchronous electricity
The unit sum of angular speed, damped coefficient, inertia time constant, stator excitation voltage rating and generator.
3. according to claim 1 based on the generator dynamic state estimator method without mark Particle filtering theory, its feature
It is:Step 2) in metric data include:The absolute generator rotor angle of generator, angular speed changing value, the amplitude of port voltage phasor
With phase angle, port electromagnetic power and machine torque.
4. according to claim 1 based on the generator dynamic state estimator method without mark Particle filtering theory, its feature
It is:Step 3) in state estimator initialization include that |input paramete initialization and metric data are initialized, at the beginning of setting state
Value, setting up procedure noise and measurement noise covariance battle array, set prediction covariance initial value, set UPF populations and filtering parameter
And sampling interval and sampling period;
5. according to claim 1 based on the generator dynamic state estimator method without mark Particle filtering theory, its feature
It is:Step 4) in the state equation form of generator be:
In formula, δ is the absolute generator rotor angle of generator amature under dq0 coordinate systems, ω0It is specified synchronous rotational speed, Δ ω is generator angle speed
Degree variable quantity, TJ is generator time inertia constant, and D is Generator Damping coefficient, EfIt is stator excitation voltage, TmIt is generator
Machine torque, Te(Te=Pe/ ω) it is the electromagnetic torque of generator, ignore stator impedance and assume ω ≈ 1, obtain Te≈Pe, its
Middle PeIt is the electromagnetic power of generator, E 'dWith E 'qRespectively the d axles of generator and q axle transient internal voltages, U andIt is generator end
The amplitude and phase angle of mouth voltage, XdAnd XqThe respectively synchronous reactance of generator d axles and q axles, X 'dWith X 'qRespectively generator d
The transient state reactance of axle and q axles, T 'd0With T 'q0The respectively transient state open circuit time constant of generator d axles and q axles;
Generator measurement is respectively the absolute generator rotor angle of generator, angular speed, electromagnetic power, is directly obtained by PMU equipment, i.e.,:Y=
[δz,ωz,Pez]T=[y1,y2,y3]T, generator measurement equation form is:
6. according to claim 1 based on the generator dynamic state estimator method without mark Particle filtering theory, its feature
It is:Step 6) in proportion of utilization sampling UKF generation the importance density function include:
The weights of ratio amendment sampling meet:
In formula:λ=α2(L+ κ)-L is fine setting parameter, and for the distance at control point to average, wherein L is state variable number, and κ is
Secondary decimation factor, is taken as 3-L;α is ratio modifying factor, is taken as 0.12;β is parameter to be selected, and regulation β improves variance precision,
Take β=2;
The step of generating the importance density function by UKF includes:
New sampled point is generated by the state filtering value and filtering covariance matrix of each particle of last moment, generator is substituted into
State equation, obtains first step status predication average and first step prediction covariance matrix of each particle in UKF;
New sampled point is generated by the first step predicted value of each particle, is repaiied using the measurement information at measurement equation and current time
Just second step status predication average and second step of each particle in UKF predicts covariance matrix;
The importance density function and life in second step status predication average and second step prediction covariance matrix generation PF
The sampling particle of Cheng Xin;
7. according to claim 1 based on the generator dynamic state estimator method without mark Particle filtering theory, its feature
It is:Step 8) in the effective particle threshold for judging resampling elect 1/3rd of total population as, method for resampling is elected as
Amount of calculation is smaller, residual error resampling of good performance.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710040475.7A CN106844952A (en) | 2017-01-20 | 2017-01-20 | Based on the generator dynamic state estimator method without mark Particle filtering theory |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710040475.7A CN106844952A (en) | 2017-01-20 | 2017-01-20 | Based on the generator dynamic state estimator method without mark Particle filtering theory |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106844952A true CN106844952A (en) | 2017-06-13 |
Family
ID=59119215
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710040475.7A Pending CN106844952A (en) | 2017-01-20 | 2017-01-20 | Based on the generator dynamic state estimator method without mark Particle filtering theory |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106844952A (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107425548A (en) * | 2017-09-11 | 2017-12-01 | 河海大学 | A kind of interpolation H ∞ EKFs generator dynamic state estimator method |
CN109327026A (en) * | 2018-09-28 | 2019-02-12 | 河海大学 | A kind of low-voltage network state of section estimation method |
CN109754013A (en) * | 2018-12-31 | 2019-05-14 | 浙江大学 | A kind of electric system hybrid measurement fusion method based on Unscented kalman filtering |
CN110162871A (en) * | 2019-05-16 | 2019-08-23 | 东北大学 | A kind of Electrical Power System Dynamic estimation method based on Unscented kalman particle filter |
CN110441691A (en) * | 2018-05-02 | 2019-11-12 | 西南科技大学 | It is a kind of based on the SOC estimation method for simplifying particle Unscented transform |
CN110619142A (en) * | 2019-07-25 | 2019-12-27 | 国网福建省电力有限公司电力科学研究院 | Generator dynamic characteristic excitation degree evaluation method based on measurement data quality inspection |
CN110707693A (en) * | 2019-10-18 | 2020-01-17 | 燕山大学 | Ensemble Kalman filtering dynamic state estimation method based on AMI full-scale measuring point partition |
CN111797974A (en) * | 2020-06-01 | 2020-10-20 | 武汉大学 | Power system state estimation method combining particle filtering and convolutional neural network |
CN112507597A (en) * | 2020-11-02 | 2021-03-16 | 中国南方电网有限责任公司超高压输电公司广州局 | Multi-cluster particle filter based multi-terminal flexible direct current transmission system state evaluation method |
CN112688596A (en) * | 2020-12-04 | 2021-04-20 | 西安理工大学 | Position sensorless control method based on UPF permanent magnet linear synchronous motor |
JP2021193563A (en) * | 2020-06-08 | 2021-12-23 | アー・ベー・ベー・パワー・グリッズ・スウィツァーランド・アクチェンゲゼルシャフトAbb Power Grids Switzerland Ag | Method for executing predicted soundness analysis for asset and computing system |
CN114880912A (en) * | 2022-05-24 | 2022-08-09 | 郑州大学 | Robust dynamic state estimation method for electromechanical transient process of generator |
US12001180B2 (en) | 2020-06-08 | 2024-06-04 | Hitachi Energy Ltd | Condition-based method for malfunction prediction |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140244189A1 (en) * | 2013-02-27 | 2014-08-28 | Bigwood Technology, Inc. | System For Achieving Real-Time Monitoring and State Estimation in Power Distribution Networks |
CN104777426A (en) * | 2015-04-17 | 2015-07-15 | 河海大学 | Power generator dynamic state estimation method based on unscented transformation strong tracking filtering |
CN105403834A (en) * | 2015-12-22 | 2016-03-16 | 华北电力大学 | Dynamic state evaluation method of generator |
-
2017
- 2017-01-20 CN CN201710040475.7A patent/CN106844952A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140244189A1 (en) * | 2013-02-27 | 2014-08-28 | Bigwood Technology, Inc. | System For Achieving Real-Time Monitoring and State Estimation in Power Distribution Networks |
CN104777426A (en) * | 2015-04-17 | 2015-07-15 | 河海大学 | Power generator dynamic state estimation method based on unscented transformation strong tracking filtering |
CN105403834A (en) * | 2015-12-22 | 2016-03-16 | 华北电力大学 | Dynamic state evaluation method of generator |
Non-Patent Citations (2)
Title |
---|
付存利: "《目标跟踪中粒子滤波及其改进算法研究》", 《万方数据》 * |
陈亮等: "《基于容积卡尔曼滤波的发电机动态状态估计》", 《中国电机工程学报》 * |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107425548A (en) * | 2017-09-11 | 2017-12-01 | 河海大学 | A kind of interpolation H ∞ EKFs generator dynamic state estimator method |
CN107425548B (en) * | 2017-09-11 | 2020-06-16 | 河海大学 | Interpolation H∞Dynamic state estimation method for extended Kalman filter generator |
CN110441691A (en) * | 2018-05-02 | 2019-11-12 | 西南科技大学 | It is a kind of based on the SOC estimation method for simplifying particle Unscented transform |
CN109327026B (en) * | 2018-09-28 | 2021-08-31 | 河海大学 | Low-voltage distribution network interval state estimation method |
CN109327026A (en) * | 2018-09-28 | 2019-02-12 | 河海大学 | A kind of low-voltage network state of section estimation method |
CN109754013A (en) * | 2018-12-31 | 2019-05-14 | 浙江大学 | A kind of electric system hybrid measurement fusion method based on Unscented kalman filtering |
CN110162871A (en) * | 2019-05-16 | 2019-08-23 | 东北大学 | A kind of Electrical Power System Dynamic estimation method based on Unscented kalman particle filter |
CN110162871B (en) * | 2019-05-16 | 2023-01-31 | 东北大学 | Power system dynamic estimation method based on unscented Kalman particle filter |
CN110619142A (en) * | 2019-07-25 | 2019-12-27 | 国网福建省电力有限公司电力科学研究院 | Generator dynamic characteristic excitation degree evaluation method based on measurement data quality inspection |
CN110619142B (en) * | 2019-07-25 | 2023-04-07 | 国网福建省电力有限公司电力科学研究院 | Generator dynamic characteristic excitation degree evaluation method based on measurement data quality inspection |
CN110707693A (en) * | 2019-10-18 | 2020-01-17 | 燕山大学 | Ensemble Kalman filtering dynamic state estimation method based on AMI full-scale measuring point partition |
CN110707693B (en) * | 2019-10-18 | 2023-06-16 | 燕山大学 | Aggregate Kalman filtering dynamic state estimation method based on AMI full measurement point partition |
CN111797974A (en) * | 2020-06-01 | 2020-10-20 | 武汉大学 | Power system state estimation method combining particle filtering and convolutional neural network |
JP2021193563A (en) * | 2020-06-08 | 2021-12-23 | アー・ベー・ベー・パワー・グリッズ・スウィツァーランド・アクチェンゲゼルシャフトAbb Power Grids Switzerland Ag | Method for executing predicted soundness analysis for asset and computing system |
JP7297816B2 (en) | 2020-06-08 | 2023-06-26 | ヒタチ・エナジー・スウィツァーランド・アクチェンゲゼルシャフト | Method and Computing System for Performing Asset Predictive Health Analysis |
US12001180B2 (en) | 2020-06-08 | 2024-06-04 | Hitachi Energy Ltd | Condition-based method for malfunction prediction |
CN112507597A (en) * | 2020-11-02 | 2021-03-16 | 中国南方电网有限责任公司超高压输电公司广州局 | Multi-cluster particle filter based multi-terminal flexible direct current transmission system state evaluation method |
CN112507597B (en) * | 2020-11-02 | 2021-10-22 | 中国南方电网有限责任公司超高压输电公司广州局 | Multi-cluster particle filter based multi-terminal flexible direct current transmission system state evaluation method |
CN112688596A (en) * | 2020-12-04 | 2021-04-20 | 西安理工大学 | Position sensorless control method based on UPF permanent magnet linear synchronous motor |
CN114880912A (en) * | 2022-05-24 | 2022-08-09 | 郑州大学 | Robust dynamic state estimation method for electromechanical transient process of generator |
CN114880912B (en) * | 2022-05-24 | 2024-02-27 | 郑州大学 | Robust dynamic state estimation method for electromechanical transient process of generator |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106844952A (en) | Based on the generator dynamic state estimator method without mark Particle filtering theory | |
US20200021133A1 (en) | Voltage stability assessment, control and probabilistic power flow based on multi-dimensional holomorphic embedding techniques | |
CN105403834B (en) | A kind of generator dynamic state estimator method | |
CN102377180B (en) | Power system load modeling method based on electric energy quality monitoring system | |
CN107577870B (en) | Power distribution network voltage power sensitivity robust estimation method based on synchronous phasor measurement | |
CN102135569B (en) | Fluctuation quantity method-based user side harmonic emission level practicality estimation method | |
US10884060B1 (en) | Dynamic parameter estimation of generators | |
US20220399721A1 (en) | Method for grid impedance and dynamics estimation | |
CN105048468A (en) | Power transmission-distribution network integrating voltage stability assessment method based on distributed calculation | |
CN105184027A (en) | Power load modeling method based on interactive multi-model algorithm | |
CN101807798A (en) | Section data integration method for power system safety and stability online analysis | |
CN104182644A (en) | Power distribution network state estimation method fusing distributed generation characteristics | |
Samson et al. | Square-root sigma-point filtering approach to state estimation for wind turbine generators in interconnected energy systems | |
CN106372440B (en) | A kind of adaptive robust state estimation method of the power distribution network of parallel computation and device | |
CN103887800B (en) | Based on the voltage on-line coordination control method of energy margin constraint | |
CN109698505B (en) | Regulation and control quantitative mapping calculation method for large power grid static voltage stability online prevention and control | |
CN104143837A (en) | Alternating-voltage-sensor-free control method for inverter with parameter adaptive characteristic | |
CN107765179A (en) | It is a kind of to be applied to measure the generator dynamic state estimator method lost | |
CN112511056A (en) | Robust generator dynamic state estimation method based on phasor measurement | |
Rao | Exploration of a scalable holomorphic embedding method formulation for power system analysis applications | |
CN110266036A (en) | A kind of dimension reduction method of current transformer multidimensional frequency domain impedance | |
CN116488234A (en) | Power system node inertia calculation method and system considering power grid structure parameters | |
CN116094060A (en) | Main and distribution network integrated state estimation method based on multi-source measurement | |
CN113553538A (en) | Recursive correction hybrid linear state estimation method | |
CN111725810B (en) | State evaluation method and terminal of alternating current-direct current hybrid power grid system |
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: 20170613 |
|
RJ01 | Rejection of invention patent application after publication |