CN103872678A - Load model identification method based on transformer substation measurement - Google Patents

Abstract

The invention discloses a load model identification method based on transformer substation measurement. The load model identification method comprises the steps of performing homology cluster division on transformer substations in a power grid, constructing load models of equivalent induction motors, constructing an identification criterion function, optimizing load parameters to be identified in the load models according to a response between input voltage and an output of an actual system by a particle swarm algorithm, and counting the load models of different stations to obtain a regional power grid load model parameter library. According to the load model identification method, an initial active power proportional coefficient Kpm and a rated initial load rate coefficient Mlf are introduced into the identified parameters, so that the influence caused by the time-varying characteristic of a load amplitude value is eliminated, and the load models are more accurate; due to the arrangement of a speed adjustment factor and an inertia factor of the particle swarm algorithm, the convergence precision of the algorithm is improved.

Description

A kind of load model identification method measuring based on transformer station

Technical field

The invention belongs to power system load model technical field, more specifically say, relate to a kind of load model identification method measuring based on transformer station.

Background technology

Electric power system is by generating, transmission of electricity, distribution and the most of organic whole forming of electric load four.Wherein, electric load is the general name of power consumption equipment in electric power networks, sometimes also comprises the power distribution network for connecting power equipment.Because having electric energy, electric power system cannot store in a large number, electrical network is the extremely strong characteristic of relevance to each other, therefore in the time of electrical network normal condition, electrical network links keeps dynamic equilibrium, and in the time that arbitrary portion breaks down in electrical network, remainder likely can be subject to great impact, finally cause electrical network local disorders, even cause the concussion of whole electrical network, cause the great loss of social safety and economic development.Along with socioeconomic sustained and rapid development, it is more and more heavy that electric load also becomes, and this makes the requirement of fail safe to electric power system, reliability stricter.China's electrical network is just towards " large electrical network; superhigh pressure; large unit; future development at a distance " at present, research dynamic load characteristic can effectively be found the fragile position of electrical network, and adopt in time various in-advance measures, and ensureing the safety and reliability of electrical network, this is the fail safe that improves electric power system, the key of reliability.

In recent years, all obtained many achievements in load modeling research field both at home and abroad, but a lot of model is still relatively coarse or even differ very large with actual characteristic.Inaccurate load model seems very inharmonious with accurate other component models of electric power system, has even limited the raising of grid simulation precision.In simulation calculation, use this relatively coarse load model, if simulation result is too conservative, can make the ability of the defeated generating of electrical network can not fully use and cause the wasting of resources; If simulation result too optimistic, can cause design planning unreasonable, reduce the stabilizing power of operation of power networks, even cause the system splitting collapse consequence causing owing to controlling misoperation.In addition the everyway that, Characteristics of Electric Load calculates grid simulation has profound influence: trend calculating, transient stability, small-signal dynamic stability and the voltage stabilization of part throttle characteristics on electric power system all can produce certain impact.Therefore the accuracy of load model is great to the good influence on system operation of electric power system, and it is very necessary that electric load model is carried out to deep research.Aspect model structure research, at present primary study mechanism load model (being mainly static load three rank induction machines in parallel), and difference equation model in non-mechanism formula model; In System Discrimination algorithm, some advanced discrimination methods have also been applied in load modeling as artificial nerve network model has also obtained a large amount of concerns in the application in load modeling field.Although forefathers' research has obtained certain achievement, along with the complexity of electrical network is more and more higher, the accuracy of load model still has much room for improvement.

Summary of the invention

The object of the invention is to overcome the deficiencies in the prior art, a kind of load model identification method measuring based on transformer station is provided, improve the accuracy of load parameter identification, improve the accuracy of load model.

For achieving the above object, the present invention is based on the load model identification method that transformer station measures, comprise the following steps:

S1: the transformer station in electrical network is divided according to electric pressure, adopt induction motor static load in parallel to represent the equivalent load model of each transformer station;

S2: equivalent induction motor adopts three order induction motor load model, α=[R _s, X _s, X _m, R _r, X _r, H, A, B, k _pz, k _pi, k _qz, k _qi, K _pm, M _lf] ^tfor the independence parameter vector to be identified of load model, β=[e _x0, e _y0, s ₀, x _s, K _l] ^tfor the identified parameters vector that can try to achieve by motor limit and α, wherein R _sfor stator resistance, X _sfor stator winding leakage reactance, X _mfor excitation reactance, R _rfor rotor resistance, X _rfor rotor leakage reactance, H are that rotor inertia time constant, A, B are electromechanics torque characteristics parameter, k _pz, k _pi, k _qz, k _qifor static characteristic parameter, e _x0, e _y0for the initial transient voltage of motor, s ₀for the revolutional slip of motor, x _sfor the synchronous reactance between stator and rotor, K _lfor load factor; K _pmrepresent initial active power proportionality coefficient, K _pm=P ₀'/P, P ₀' be the initial active power of motor, P by the active power that consumes in transient process of survey load point; M _lffor specified initial load rate coefficient,

s _mBthe rated capacity of induction motor, U _bload reference voltage, U ₀it is the busbar voltage initial value of loading in transient process;

S3: carry out load model parameters identification, concrete steps comprise:

S3.1: input voltage excitation U (k) sampled measurements obtain N+1 actual output response y (t)=[P (t), Q (t)] in actual transformer substation system ^twherein t represents to measure sample sequence number, span is 0≤t≤N, the active power value that P (t) consumes in transient process for the load point of moment t, the reactive power value that Q (t) consumes in transient process for the load point of moment t;

S3.2: by the reality output response discrete-time series in corresponding moment u (0), u (1) ..., u (N) } and input load model, the initial steady state condition of load model is: initial voltage U ₀=u (0), initial active-power P ₀=P (0), initial reactive power Q ₀=Q (0); Set n the initial value of independent parameter vector α to be identified, employing particle cluster algorithm is to following target function optimizing:

$\min J(\mathrm{\α},\mathrm{\β})=\min \underset{t=0}{\overset{N}{\mathrm{\Σ}}}{[y\left(t\right)-y_m\left(t\right)]}^T[y\left(t\right)-y_m\left(t\right)]$

Wherein, y _m(t)=[P _m(t), Q _m(t)] ^tthe output response that during for input u (t), load model obtains;

Speed in particle cluster algorithm is adjusted factor a (m) and is determined according to following formula:

$a\left(m\right)=\frac{1}{F}\underset{i=1}{\overset{D}{\mathrm{\Σ}}}\sqrt{a_{\max }\left(m\right)+a_{\min }\left(m\right)}$

$a_{\max }\left(m\right)=\frac{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{(n_{\mathrm{ij}}^1-c_{\mathrm{ij}}^1)}^2}{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{\left(c_{\mathrm{ij}}^1\right)}^2}$

$a_{\min }\left(m\right)=\frac{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{(n_{\mathrm{ij}}^2-c_{\mathrm{ij}}^2)}^2}{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{\left(c_{\mathrm{ij}}^2\right)}^2}$

Wherein, F is particle number in population, and D is the dimension of independent parameter vector α to be identified,

represent respectively maximal rate and the fitting data corresponding to minimum speed of the parameters arranging,

represent respectively maximal rate and the measurement data calculated value corresponding to minimum speed of the parameters arranging.

S4: the load model of each equivalent induction motor and load parameter identification result thereof articulate the inferior duty value model of bus and load parameter as transformer station, and the load model of different websites is added up, and obtain area power grid load model's databank.

The present invention is based on the load model identification method that transformer station measures, first the transformer station in electrical network is carried out to motor cluster division, build the load model of equivalent induction motor, set up identification criterion function, adopt particle cluster algorithm to carry out optimizing to load parameter to be identified in load model by the output response of input voltage and real system, finally add up the load model of different websites, obtain area power grid load model's databank.

The present invention has following beneficial effect:

(1), for the transformer station of the different electric pressures of area power grid, gather transformer station and go out to survey busbar voltage, current data, to the cluster for the first time of all websites in region, as the priori conditions of load classification;

(2) by introducing initial active power Proportional coefficient K _pmwith specified initial load rate coefficient M _lf, the impact that the time variation of elimination load amplitude brings;

(3) the position update algorithm of employing variable coefficient, the speed that has reset particle cluster algorithm is adjusted the factor and inertial factor, prevents that the appearance of precocious phenomenon from avoiding falling into into local best points simultaneously, improves algorithmic statement precision.

Accompanying drawing explanation

Fig. 1 is a kind of embodiment flow chart that the present invention is based on the load model identification method of transformer station's measurement;

Fig. 2 is the flow chart of particle cluster algorithm in the present invention.

Embodiment

Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described, so that those skilled in the art understands the present invention better.Requiring particular attention is that, in the following description, in the time that perhaps the detailed description of known function and design can desalinate main contents of the present invention, these are described in here and will be left in the basket.

Embodiment

Fig. 1 is a kind of embodiment flow chart that the present invention is based on the load model identification method of transformer station's measurement.As shown in Figure 1, the present invention is based on transformer station measure load model identification method comprise the following steps:

S101: divide motor cluster, equivalence is induction motor:

Power System Interconnection is integrated, it is analyzed and calculates the support that needs whole network data, and the operation of electric power system, management and scheduling are again layering and zonings, Real-time Obtaining whole network data is extremely difficult and loaded down with trivial details, in order to build the electrical network dynamic equivalent model of clear layer, the present invention adopts the hierarchical coordinative mode between electrical network to calculate.Therefore first the transformer station in electrical network is carried out to motor cluster division, adopt equivalent induction motor and static load to represent each motor cluster.Concrete grammar is: transformer station is carried out to cluster according to electric pressure, for the transformer station of same electric pressure according to motor shared ratio reclassification in load.Ripe clustering method has much at present, for example K mean algorithm, FCM Algorithms, hard mean algorithm etc.

S102: build load model:

The present invention considers that network load direct screening is connected on low-voltage bus bar side, adopts three order induction motor load model, and rotor state equation and the stator output equation of induction motor are as follows:

$\left\{\begin{array}{c}\frac{{{\mathrm{dE}}_d}^{\′}}{\mathrm{dt}}=-\frac{1}{T_0^{\′}}[{E_d}^{\′}+(X-X^{\′})I_q]+(\mathrm{\ω}-{\mathrm{\ω}}_r){E_d}^{\′}\\ \frac{{{\mathrm{dE}}_q}^{\′}}{\mathrm{dt}}=-\frac{1}{T_0^{\′}}[{E_q}^{\′}+(X-X^{\′})I_d]+(\mathrm{\ω}-{\mathrm{\ω}}_r){E_q}^{\′}\\ \frac{{\mathrm{d\ω}}_r}{\mathrm{dt}}=\frac{1}{2H}[({E_d}^{\′}{I}_d+{E_q}^{\′}{I}_q)-T_L({{\mathrm{A\ω}}_r}^2+{\mathrm{B\ω}}_r+C))]\end{array}\right.---\left(1\right)$

$\left\{\begin{array}{c}{I}_d=\frac{1}{{R_S}^2+{X_S}^{\′2}}[R_S(U_d-{E_d}^{\′})+X^{\′}(U_q-{E_q}^{\′})]\\ I_q=\frac{1}{{R_S}^2+{X_S}^{\′2}}[R_S(U_q-{E_q}^{\′})-X^{\′}(U_d-{E_d}^{\′})]\end{array}\right.---\left(2\right)$

Wherein E _d' expression rotor d axle electromotive force, E _q' representing rotor q axle electromotive force, ω represents the synchronous speed of stator field, ω _rrepresent the rotating speed of rotor, T ₀'=(X _r+ X _m)/R _r, A, B, C are electromechanics torque characteristics parameter, A+B+C=1.H is rotor inertia time constant, and X and X ' are respectively induction motor steady-state reactance and transient state reactance, X=X _s+ X _m, X'=X _s+ (X _mx _r)/(X _m+ X _r).I _d, I _qbe respectively the cross, straight axle component that flows through induction motor electric current.X _sfor stator winding leakage reactance, X _mfor excitation reactance, X _rfor rotor leakage reactance, R _rfor rotor resistance, R _sfor stator resistance, U _d, U _qbe respectively the cross, straight axle component value of the voltage on load bus.

In three order induction motor load model, static part is:

$\left\{\begin{array}{c}\frac{P}{P_0}=k_{\mathrm{Pz}}\left(\frac{U}{U_0}\right)+k_{\mathrm{Pi}}\left(\frac{U}{U_0}\right)+k_{\mathrm{Pp}}\\ \frac{Q}{Q_0}=k_{\mathrm{Qz}}\left(\frac{U}{U_0}\right)+k_{\mathrm{Qi}}\left(\frac{U}{U_0}\right)+k_{\mathrm{Qp}}\end{array}\right.---\left(3\right)$

Wherein, k _pz, k _pi, k _ppfor static characteristic parameter, k _pz+ k _pi+ k _pp=1; k _qz, k _qi, k _qpfor static characteristic parameter, k _qz+ k _qi+ k _qp=1; P ₀, Q ₀be respectively with P, Q active power and the reactive power value that surveyed load point consumes in the time that transient process starts and in transient process; U, U ₀be respectively magnitude of voltage and the initial value of the bus of loading in transient process.

In order to adapt to the time variation of load, realize the self adaptation of load capacity, the present invention introduces two parameter K _pmand M _lf, wherein, K _pmrepresent initial active power proportionality coefficient, M _lffor specified initial load rate coefficient, formula is:

K _pm＝P ₀′/P (4)

$M_{\mathrm{lf}}=\left(\frac{{P_0}^{\′}}{S_{\mathrm{MB}}}\right)/\left(\frac{U_0}{U_B}\right)---\left(5\right)$

Wherein, P ₀' be the initial active power of motor, S _mBthe rated capacity of induction motor, U _bit is load reference voltage.

By introducing parameter K _pmand M _lfcan guarantee to eliminate completely the impact that the time variation of load amplitude brings, if i.e.: load constituent proportions constant, no matter how payload changes, and can carry out all loads of matching by one group of parameter, simplify to a great extent load modeling process, increased the practicality of model.

Induction motor model is nonlinearized Mathematical Modeling, in power system analysis, is a very difficult problem to the identification analysis of nonlinear model.The parameter identification method of non linear system is mostly that its main process is the parameter vector of finding one group of optimum take optimization method as basis at present, makes predetermined error target function value reach minimum, and error target function is the function that needs the parameter of identification.Because the analytic relationship of this function is difficult to represent, solution space is often also quite complicated, has multiple limits, and in numerous load model parameters discrimination methods, particle cluster algorithm has good robustness and convergence rate faster.But also existence is as poor in ability of searching optimum, and while iterating to the later stage, the flying speed of particle is too small, easily sinks into local extremum, the not high defect of convergence precision.Therefore the present invention, by particle cluster algorithm is improved, promotes convergence precision.

S103: load parameter identification:

The identification criterion function of load model is as shown in the formula shown in (4):

$\min J(\mathrm{\α},\mathrm{\β})=\min \underset{t=0}{\overset{N}{\mathrm{\Σ}}}{[y\left(t\right)-y_m\left(t\right)]}^T[y\left(t\right)-y_m\left(t\right)]---\left(4\right)$

Wherein t=0 is for sampling the zero hour, and N is the moment that sampling finishes, and samples N+1 time altogether.。α=[R _s, X _s, X _m, R _r, X _r, H, A, B, K _pm, M _lf, k _pz, k _pi, k _qz, k _qi] ^tfor the parameter vector to be identified of load model, β=[e _x0, e _y0, s ₀, x _s, K _l] ^tfor the identified parameters vector that can try to achieve by motor limit and α, wherein e _x0, e _y0for the initial transient voltage of motor, s ₀for the revolutional slip of motor, x _sfor the synchronous reactance between stator and rotor, K _lfor load factor.Y (t)=[P (t), Q (t)] ^tand y _m(t)=[P _m(t), Q _m(t)] ^tfor the output response of the actual measurement of moment t and the output response that identification model obtains.

Concrete identification process is:

S3.1: input voltage excitation U (k) sampled measurements obtain N+1 actual output response y (t)=[P (t), Q (t)] in actual transformer substation system ^twherein t represents to measure sample sequence number, span is 0≤t≤N, the active power value that P (t) consumes in transient process for the load point of moment t, the reactive power value that Q (t) consumes in transient process for the load point of moment t.

S3.2: by the reality output response discrete-time series in corresponding moment u (0), u (1) ..., u (N) } and input load model, the initial steady state condition of load model is: initial voltage U ₀=u (0), initial active-power P ₀=P (0), initial reactive power Q ₀=Q (0); Set the initial value of independent parameter vector α to be identified, adopt particle cluster algorithm to following target function optimizing:

$\min J(\mathrm{\α},\mathrm{\β})=\min \underset{t=0}{\overset{N}{\mathrm{\Σ}}}{[y\left(t\right)-y_m\left(t\right)]}^T[y\left(t\right)-y_m\left(t\right)]---\left(5\right)$

Wherein, y _m(t)=[P _m(t), Q _m(t)] ^tthe output response that during for input u (t), load model obtains.

Visible, according to independent parameter vector α to be identified and initial steady state condition, can calculate identified parameters vector β, thus the optimizing of substitution target function.

Particle cluster algorithm is the optimal solution of being searched particle in plain space and determined function by use, and all particle objects are all by an adaptive value being determined by majorized function with determine that the speed of their headings forms.The operation behavior of particle cluster algorithm is:

$\left\{\begin{array}{c}{v}_{\mathrm{id}}^{k+1}={\mathrm{wv}}_{\mathrm{id}}^k+\mathrm{Random}(0,c_1)(p_{\mathrm{id}}^k-x_{\mathrm{id}}^k)+\mathrm{Random}(0,c_2)(p_{\mathrm{gd}}^k-x_{\mathrm{id}}^x)\\ x_{\mathrm{id}}^{k+1}=x_{\mathrm{id}}^k+a\left(m\right)v_{\mathrm{id}}^{k+1}\\ v_{d\min }\≤v_{\mathrm{id}}^k\≤v_{d\max }\\ x_{d\max }\≤x_{\mathrm{id}}^k\≤x_{\mathrm{dmx}}\end{array}\right.---\left(6\right)$

In note population, particle number is F, and the dimension of the independent parameter vector α to be identified of each particle is D, and the position of i particle can be expressed as the vector x of a D dimension _i=(x _i1, x _i2..., x _id..., x _iD), i=1,2 ..., F, d=1,2 ..., D.By x _isubstitution target function just can calculate adaptive value, just can weigh x by adaptive value _iquality.The flying speed of particle is also a D dimensional vector, is designated as v _i=(v _i1, v _i2..., v _id..., v _iD).The optimal location that i particle searches is up to now p _i=(p _i1, p _i2..., p _id..., p _iD), the optimal location that whole population searches is up to now p _g=(q _g1, q _g2..., q _gd..., q _gD).W is inertial factor, c ₁, c ₂be the study factor, Random () is random function, the number of times that subscript k is current iteration.A (m) is that speed is adjusted the factor, conventionally gets 1.But the position that has redefined a (m) in the present invention and come convergent-divergent particle of future generation, adopts annular to follow the example of, and speed interval number is D-1, m=1,2 ..., D-1.The size of a (m) is determined by many curvilinear finite differences fitting process.Concrete grammar is as follows:

If the measurement data of j parameter is c in i particle _ij, fitting data is n _ij, the target function of many curvilinear finite differences fitting process is:

$a\left(m\right)=\frac{1}{F}\underset{i=1}{\overset{F}{\mathrm{\Σ}}}\left(\frac{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{(n_{\mathrm{ij}}-c_{\mathrm{ij}})}^2}{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{c_{\mathrm{ij}}}^2}\right)---\left(7\right)$

When the speed adjustment factor is finally determined, require " flight " speed of particulate can not depart from actual maximal rate and minimum speed value, therefore, in the time determining target function, need to consider both, therefore can be respectively at maximal rate and fitting data corresponding to minimum speed that parameters is set

measurement data calculated value is respectively

prerequisite under, the speed of particle maximal rate and minimum speed is adjusted the factor and can be made as respectively:

$a_{\max }\left(m\right)=\frac{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{(n_{\mathrm{ij}}^1-c_{\mathrm{ij}}^1)}^2}{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{\left(c_{\mathrm{ij}}^1\right)}^2}---\left(8\right)$

$a_{\min }\left(m\right)=\frac{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{(n_{\mathrm{ij}}^2-c_{\mathrm{ij}}^2)}^2}{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{\left(c_{\mathrm{ij}}^2\right)}^2}---\left(8\right)$

On this basis, the final goal function of speed adjustment factor a (m) can be defined as:

$a\left(m\right)=\frac{1}{F}\underset{i=1}{\overset{D}{\mathrm{\Σ}}}\sqrt{a_{\max }\left(m\right)+a_{\min }\left(m\right)}---\left(10\right)$

Because speed is adjusted the change curve more complicated of the factor, if expect, utility function approaches to obtain proper data and needs to improve number of times, increases amount of calculation thereby increase undetermined coefficient.Consider that collection point is discrete point, difference equation does not require first derivative, and amount of calculation is less, and the scope of application is wider.

Inertial factor w can have larger impact to the performance of particle cluster algorithm in formula (6): the convergence rate that larger w can boosting algorithm, the precision that less w can boosting algorithm.The present invention proposes a kind of in iterative process w according to circumstances carry out the method for adaptive correction, along with moving ahead of calculating, reduce w by index gradually, its formula is:

$w\left(k\right)=[2/(1+e^{\∂k/k_{\max }})]w^0---\left(11\right)$

Wherein

represent normal number, k _maxfor the greatest iteration number of particle cluster algorithm; w ⁰for the upper limit of w (k); K is the number of times of current iteration.

All two " extreme value " of sensing because the direction of particle flight is the same with former algorithm, but in improvement algorithm used in the present invention, particle position is subject to elastic wave velocity adjust the impact of the factor and dwindle adaptively or amplify, while having avoided having arrived iteration end, whole colony is all around drawn close to same extreme point, has therefore prevented that the appearance of precocious phenomenon from avoiding falling into into local best points simultaneously.Can also adopt self adaptation inertial factor, avoid population in iterative process to occur violent " homoplasy ", promote the convergence precision of particle cluster algorithm.

Fig. 2 is the flow chart of particle cluster algorithm in the present invention:

S201: within the scope of solution space, set primary group.

S202: the adaptive value of calculating each particle according to target function;

S203: judge whether adaptive value reaches optimum requirement, if reached, enter step S208; If do not reached, enter step S204;

S204: to each particle, if the adaptive value of current location is better than individual historical optimal value, historical individuality adaptive optimal control value and optimal location are replaced by the current situation of particle; To each particle, if the adaptive value of current location is better than the historical optimal value of colony, historical colony adaptive optimal control value and optimal location are replaced by the current situation of particle;

S205: judge whether iterations reaches maximum iteration time, if reached, enter step S208; If do not reached, enter step S206:

S206: computational speed is adjusted factor a (m) and inertial factor w;

S207: the more position of new particle, return to step S202;

S208: export historical optimum particle position, i.e. particle position corresponding to historical adaptive optimal control value, algorithm finishes.

S104: the load of different qualities carries out load model parameters identification work separately according to above-mentioned steps, the load model of each equivalent induction motor and load parameter identification result thereof articulate the inferior duty value model of bus and load parameter as transformer station, the load model of different websites is added up, obtained area power grid load model's databank.

Although above the illustrative embodiment of the present invention is described; so that those skilled in the art understand the present invention; but should be clear; the invention is not restricted to the scope of embodiment; to those skilled in the art; as long as various variations appended claim limit and definite the spirit and scope of the present invention in, these variations are apparent, all utilize innovation and creation that the present invention conceives all at the row of protection.

Claims (4)

1. the load model identification method measuring based on transformer station, is characterized in that, comprises the following steps:

S1: the transformer station in electrical network is divided according to electric pressure, adopt induction motor static load in parallel to represent the equivalent load model of each transformer station;

S2: equivalent induction motor adopts three order induction motor induction motor load model, α=[R _s, X _s, X _m, R _r, X _r, H, A, B, k _pz, k _pi, k _qz, k _qi, K _pm, M _lf] ^tfor the independence parameter vector to be identified of load model, β=[e _x0, e _y0, s ₀, x _s, K _l] ^tfor the identified parameters vector that can try to achieve by motor limit and α, wherein R _sfor stator resistance, X _sfor stator winding leakage reactance, X _mfor excitation reactance, R _rfor rotor resistance, X _rfor rotor leakage reactance, H are that rotor inertia time constant, A, B are electromechanics torque characteristics parameter, k _pz, k _pi, k _qz, k _qifor static characteristic parameter, e _x0, e _y0for the initial transient voltage of motor, s ₀for the revolutional slip of motor, x _sfor the synchronous reactance between stator and rotor, K _lfor load factor; K _pmrepresent initial active power proportionality coefficient, K _pm=P ₀'/P, P ₀' be the initial active power of motor, P by the active power that consumes in transient process of survey load point; M _lffor specified initial load rate coefficient,

s _mBthe rated capacity of induction motor, U _bload reference voltage, U ₀it is the busbar voltage initial value of loading in transient process;

S3: carry out load model parameters identification, concrete steps comprise:

S3.1: input voltage excitation U (k) sampled measurements obtain N+1 actual output response y (t)=[P (t), Q (t)] in actual transformer substation system ^twherein t represents to measure sample sequence number, span is 0≤t≤N, the active power value that P (t) consumes in transient process for the load point of moment t, the reactive power value that Q (t) consumes in transient process for the load point of moment t;

S3.2: by the reality output response discrete-time series in corresponding moment u (0), u (1) ..., u (N) } and input load model, the initial steady state condition of load model is: initial voltage U ₀=u (0), initial active-power P ₀=P (0), initial reactive power Q ₀=Q (0); Set n the initial value of independent parameter vector α to be identified, employing particle cluster algorithm is to following target function optimizing:

$\min J(\mathrm{\α},\mathrm{\β})=\min \underset{t=0}{\overset{N}{\mathrm{\Σ}}}{[y\left(t\right)-y_m\left(t\right)]}^T[y\left(t\right)-y_m\left(t\right)]$

Wherein, y _m(t)=[P _m(t), Q _m(t)] ^tthe output response that during for input u (t), load model obtains;

Speed in particle cluster algorithm is adjusted factor a (m) and is determined according to following formula:

$a\left(m\right)=\frac{1}{F}\underset{i=1}{\overset{D}{\mathrm{\Σ}}}\sqrt{a_{\max }\left(m\right)+a_{\min }\left(m\right)}$

$a_{\max }\left(m\right)=\frac{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{(n_{\mathrm{ij}}^1-c_{\mathrm{ij}}^1)}^2}{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{\left(c_{\mathrm{ij}}^1\right)}^2}$

$a_{\min }\left(m\right)=\frac{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{(n_{\mathrm{ij}}^2-c_{\mathrm{ij}}^2)}^2}{\underset{j=1}{\overset{D}{\mathrm{\Σ}}}{\left(c_{\mathrm{ij}}^2\right)}^2}$

Wherein, F is particle number in population, and D is the dimension of independent parameter vector α to be identified,

represent respectively maximal rate and the fitting data corresponding to minimum speed of the parameters arranging,

represent respectively maximal rate and the measurement data calculated value corresponding to minimum speed of the parameters arranging.

S4: the load model of each equivalent induction motor and load parameter identification result thereof articulate the inferior duty value model of bus and load parameter as transformer station, and the load model of different websites is added up, and obtain area power grid load model's databank.

2. load model identification method according to claim 1, it is characterized in that, the concrete grammar that in described step S1, Coherent Generator Group is divided is: transformer station is carried out to cluster according to electric pressure, for the transformer station of same electric pressure according to motor shared ratio reclassification in load.

3. load discrimination method according to claim 1, is characterized in that, in described step S3.2, the inertial factor w in particle cluster algorithm determines according to following formula:

$w\left(k\right)=[2/(1+e^{\∂k/k_{\max }})]w^0$

Wherein represent normal number, k _maxfor the greatest iteration number of particle cluster algorithm; w ⁰for the upper limit of w (k); K is the number of times of current iteration.

4. load model identification method according to claim 1, is characterized in that, the dynamic equivalent in described step S4 adopts people having the same aspiration and interest equivalence method.

Images