CN105868514A - Wide-area power system load model calibration method based on calibration navigation - Google Patents

Abstract

The invention provides a wide area power system load model calibration method based on calibration navigation, which comprises the following steps: the method comprises the steps of establishing a load parameter error index, identifying a system leading calibration area, classifying loads to be calibrated in the system leading calibration area in two stages, uniformly calibrating various loads in the leading calibration area, and calibrating a load model based on a multi-disturbance scene. The load model parameters finally obtained by the wide area power system load model calibration method can accurately and reliably simulate the regional load of the wide area system, the calibration result can be applied to the fields of design, planning, operation and the like of the power system according to different requirements, the related calculation precision and prediction accuracy can be obviously improved, and the system safety and economy are improved.

Description

A kind of wide area power system load model calibration steps based on calibration navigation

Technical field

The present invention relates to a kind of wide area power system load model calibration steps, a kind of based on calibration navigation wide Territory power system load model calibration steps.

Background technology

Owing to can not arbitrarily changing the operating mode of real system and applying disturbance, thus Power System Planning and operation study all from Do not open Digital Simulation.But, Practical Project repeatedly finding, simulation result cannot reappear real dynamic process, being forbidden of emulation Really the decision-making formulated accordingly will be caused dangerous or uneconomical.Model is the principal element causing phantom error, and its effectiveness is commented Estimating and calibrating is the important basic work of power system.Increasing in order to make phantom follow the tracks of system in time is built and is adjusted, and needs Periodically to carry out the calibration of system model.The operating random disturbance of system is that model calibration provides good opportunity.System It is that the two all can obtain identical moving under all excitations that model can become the necessary and sufficient condition of actual wide area system mirror image State responds, and its research must be based on the overall dynamics information of system.WAMS (WAMS) is wide area power system model Assessment and calibration provide with reference to Information base.The real system response curve gathered by WAMS, as the standard compared, is built Diversity factor evaluation index between vertical emulation and actual measurement track, as the object function of model calibration.Wide area power system load number Mesh is various, and therefore, in large scale system simulation process, load bus is usually load, transformator, power supply, compensation device etc. Polymerization, furthermore, it is contemplated that the dispersion of load self and time variation, in analogue system, other element of load model relatively system is more not Credible.If according to the effectiveness of random disturbance assessment load model, and invalid load model can be calibrated in time, then can be to power train System planning, design and run offer reference the most accurately, improve system design and run arrange safety and economy, There is great engineering use value.

Wide area power system node is numerous, in conventional wide area system load model efficiency assessment, utilizes single Local acknowledgement signal reflection system model characteristic, and it is directly based upon responding trajectory itself or its geometric shape feature sets up effectiveness Evaluation index, during model calibration, typically presses unique characteristics classification by system loading, and each type load uses same model ginseng Number, carries out the unified parameters calibration of total system load.The method cannot react the general characteristic of wide area system load model comprehensively, And its index is positioned at inhibited stably in system loading parameter and is unsatisfactory for seriality, the mistake that follow-up load model will be caused to calibrate Sentence.And the method does not investigate the different load node influence degree to response in space during load model is calibrated, will Model error affects result and shares all load buses in space equally, had both been unfavorable for the accurate alignment of model, and had also made parameter to be calibrated There is higher dimension, reduce the computational efficiency of calibration.Utilize the classification of load unique characteristics can minimizing portion to a certain extent Divide parameter to be calibrated, but do not account for load offseting because of the respective response results caused by the difference of its spatial distribution, nothing Method avoids corresponding model calibration error.The most conventional wide area system load model efficiency assessment and calibration steps do not have There is sufficiently high engineering practical value.

Summary of the invention

The technical problem to be solved in the present invention is that existing method cannot react the overall special of wide area system load model comprehensively Levy, or do not account for each load bus difference because of spatial distribution to responding the difference of influence degree and respective response results Offset, it is impossible to avoid corresponding model calibration error.

In order to solve above-mentioned technical problem, the invention provides a kind of wide area power system load mould based on calibration navigation Type calibration steps, comprises the steps:

Step 1, the foundation of load parameter error criterion, WAMS wide area measurement and numerical integration respectively obtain the most same Disturbance Dis_r(r=1,2 ..., R) under the actual measurement track of system and simulated response, system is real to use extended equal area criterion to determine Survey the system dominant mode of track, and it is abundant to calculate the angle stability that under this system dominant mode, real system and analogue system are respectively put Degree, calculating parameter error D according still further to formula (1) is:

$D=N^{-1}\underset{l=1}{\overset{N}{\Σ}}\left|{\mathrm{\η}}_{Ml}-{\mathrm{\η}}_{Sl}\right|---\left(1\right)$

In formula (1), η_MlAnd η_SlIt is respectively actual measurement track and the angle stability of simulation track l pendulum under system dominant mode Nargin, N=min{N_M,N_S, N_MAnd N_SIt is respectively merit angle actual measurement track and total pendulum of simulation track in watch window, if D > error threshold ε, then enter step 2 and start load model calibration, otherwise waits for noisy data next time and repeats step 1；

Step 2, system dominates calibration region identification, computer sim-ulation system angle stability nargin under system dominant mode ForAccording still further to angle stability nargin under formula (2) calculating system dominant mode about node load ParameterParametric sensitivity be:

$\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^j}=\frac{{\mathrm{\η}}_S({\mathrm{\α}}_i^j+{\mathrm{\Δ\α}}_i^j)-{\mathrm{\η}}_S\left({\mathrm{\α}}_i^j\right)}{{\mathrm{\Δ\α}}_i^j}---\left(2\right)$

In formula (2),Expression system jth (j=1,2 ..., n) individual load bus i-th (i=1,2 ..., m) individual model Parameter,For node load parameterPerturbation, n is all load bus numbers to be calibrated of system, and m is each load Contained model parameter number on node, wherein i-th load model parameters is α_i, then calculate angle stability under system dominant mode Nargin is about each load model parameters α_iThe spatial distribution of sensitivity absolute value, and this spatial distribution is joined as load model Number α_iCalibration navigation under system dominant modeFor:

${\mathrm{CN}}_{{\mathrm{\α}}_i}=\left\{\right|\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^1}|,|\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^2}|,...,|\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^j}|,...,|\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^n}\left|\right\}---\left(3\right)$

Calculate under system dominant mode angle stability nargin again about the load model parameters α of each node_iSensitivity symbol Number spatial distribution, and using this spatial distribution as load model parameters α under system dominant mode_iStable spy on each node LevyFor:

${\mathrm{CS}}_{{\mathrm{\α}}_i}=\{\mathrm{sgn}\left(\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^1}\right),\mathrm{sgn}\left(\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^2}\right),\mathrm{...},\mathrm{sgn}\left(\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^j}\right),\mathrm{...},\mathrm{sgn}\left(\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^n}\right)\}---\left(4\right)$

Set about load model parameters α_iLeading calibration node ratio θ_i, rightIn each element by depending on from big to small Secondary arrangement, takes front n θ_iIndividual node is about load model parameters α_iLeading calibration node set, by all m load model ginsengs Number α_i(i=1,2 ..., m) corresponding leading calibration node set composition system dominates calibration region；

Step 3, system is the two phase classification of load to be calibrated in dominating calibration region, travels through all m load model ginsengs Number α_i, it is met formula (6) and makes calibration navigateTake maximum CN_MAXParameter, and define this parameter be system indicate parameter α_s, CN_MAXComputing formula be:

${\mathrm{CN}}_{MAX}=max\{max\left({\mathrm{CN}}_{{\mathrm{\α}}_1}\right),max\left({\mathrm{CN}}_{{\mathrm{\α}}_2}\right),...,max\left({\mathrm{CN}}_{{\mathrm{\α}}_s}\right),...,max\left({\mathrm{CN}}_{{\mathrm{\α}}_m}\right)\}---\left(6\right)$

Reutilization system instruction parameter alpha_sInvariant featureBy invariant featureIn each load bus positive and negative will System is dominated the load bus of calibration region and is divided into two classes, further according to load nature of electricity consumed composition X of load bus_j, and utilize mould Paste C means clustering algorithm, respectively to two type load node subseries again, obtains load classification number total in system dominates calibration region Mesh is C, c_k(c_k=1,2 ..., C) type load is WhereinFor belonging to c_kThe load number of class,It is c_kIn type loadIndividual load；

Step 4, the unified calibration of each type load, each load model of system non-dominant calibration region in leading calibration region Without checking, system is dominated all load buses of same class load in calibration region and is set to identical model parameter, then unifies Model parameter on calibration C type load node, if c_kCalibration is needed on type load nodeIndividual model parameter is with vector It is expressed asThen total parameter of the required calibration of system isUse again Formula (1) is as the object function of parametric calibration, then using default parameter value as the initial value of calibration, utilizes gradient descent method meter Calculation obtains the parameter vector set that min (D) is correspondingFor parametric calibration result；

Step 5, load model based on many disturbances scene is calibrated, and persistently records system disturbance track, utilizes each disturbance Under measurement data, repeat step 1-4, obtain respectively correspondence system under each disturbance dominate calibration region load model calibration knot Really, it is updated to simulation system data storehouse, along with the continuous accumulation of disturbance record, calibrates and update whole system load model.

Angle stability nargin reflection wide area system load model under the dominant pattern that employing system overall situation phase-swing curves calculates Response characteristic, can reflect the impact of whole wide area system load, makes the global measuring of WAMS be fully used；Stability margin At the Continuous property of parametric stability region boundary, enable the object function accurate instruction load parameter calibration set up accordingly；With Under dominant pattern angle stability nargin about the sensitivity of each node load parameter as the school of wide area power system load model Quasi-navigation, with leading calibration region and the invariant feature of load bus of its compartment system load model, model calibration process is only Load for leading calibration region launches, and greatly reduces the dimension of wide area system load model calibration object, shields non- The parameter impact of leading calibration region, improves model calibration precision；Stablizing according to each load bus in leading calibration region Feature and load nature of electricity consumed composition carry out two phase classification, and same class load bus carries out unifying school by identical model parameter Standard, reduces load model parameter to be calibrated the most further, improves efficiency and the precision of the calibration of wide area power system load model.

Limit in scheme, step 2 further as the present invention, willIn the corresponding position of leading calibration node Element puts 1, and remaining element is non-dominant node,Corresponding position element set to 0, obtain about load model parameters α_i's Pilot bus flag sequenceThe Pilot bus flag sequence DBS of system is by each load model parameters α_iPilot bus Flag sequenceOr computing obtain.Pass throughCalculating, each load model parameters α_iAll it has the leading of correspondence Calibration node set, the leading calibration node set of each parameter typically will not be identical, the Pilot bus corresponding by each parameter Flag sequence or the Pilot bus flag sequence DBS of system that computing obtains can fully count and all loads on each node The impact that system is responded by parameter.

Limit in scheme, step 3 further as the present invention, load nature of electricity consumed composition X_j=(x_j1,x_j2,x_j3, x_j4), wherein x_j1、x_j2、x_j3And x_j4Represent respectively shared by industry on jth load bus, business, agricultural and other load Electricity consumption proportion.

Limit in scheme, step 3 further, by invariant feature as the present inventionIn each load bus positive and negative will System is dominated the load of calibration region and is divided into two classes to be respectively as follows: on a type load node, stability margin under system dominant mode η_SParameter alpha is indicated about this category node_sPartial derivative be nonnegative value；On another kind of load bus, steady under system dominant mode Determine nargin η_SParameter alpha is indicated about this category node_sPartial derivative be negative value.

The beneficial effects of the present invention is: (1) uses angle stability under the dominant pattern that system overall situation phase-swing curves calculates Nargin reflection wide area system load model response characteristic, can reflect the impact of whole wide area system load, make the overall situation of WAMS Measurement is fully used；(2) stability margin is at the Continuous property of parametric stability region boundary, makes the object function set up accordingly Can the calibration of accurate instruction load parameter；(3) with sensitive about each node load parameter of angle stability nargin under dominant pattern Spend the calibration as wide area power system load model to navigate, with leading calibration region and the load of its compartment system load model The invariant feature of node, model calibration process is launched only for the load of leading calibration region, greatly reduces wide area system and bears The dimension of lotus model calibration object, shields the parameter impact of non-dominant calibration region, improves model calibration precision；(4) leading Leading in calibration region invariant feature and load nature of electricity consumed composition according to each load bus and carry out two phase classification, same class is born Lotus node carries out unifying calibration by identical model parameter, reduces load model parameter to be calibrated the most further, improves wide area The efficiency of power system load model calibration and precision.

Accompanying drawing explanation

Fig. 1 is the method flow diagram of the present invention.

Detailed description of the invention

As it is shown in figure 1, the wide area power system load model calibration steps based on calibration navigation that the present invention proposes combines Accompanying drawing describes detailed description of the invention in detail.

As it is shown in figure 1, the wide area power system load model calibration steps based on calibration navigation of the present invention specifically includes Following steps:

Step 1, the foundation of load parameter error criterion, WAMS wide area measurement and numerical integration respectively obtain same disturbance Dis_r(r=1,2 ..., R) under the actual measurement track of system and simulated response, system surveys rail to use extended equal area criterion to determine The dominant pattern of mark, and calculate the angle stability nargin that under this dominant pattern, real system and analogue system are respectively put, according to formula (1) Computation model parameter error D, step-up error threshold epsilon, if D > ε, then start load model and be calibrated into step 2；Otherwise, wait Noisy data repeats step 1 next time；

$D=N^{-1}\underset{l=1}{\overset{N}{\Σ}}\left|{\mathrm{\η}}_{Ml}-{\mathrm{\η}}_{Sl}\right|---\left(1\right)$

In formula (1), η_MlAnd η_SlIt is respectively actual measurement track and the merit angle of simulation track l pendulum under actual measurement track dominant pattern Stability margin；N=min{N_M,N_S, N_MAnd N_SIt is respectively merit angle to survey in watch window and total pendulum of simulation track；

Step 2, system is dominated calibration region identification, is calculated electric system simulation system under actual measurement track dominant pattern Angle stability narginPerturbation jth (j=1,2 ..., n) individual node load parameterAccording to Under formula (2) calculating system dominant mode, angle stability nargin is about the sensitivity of this parameter, abundant with angle stability under dominant pattern Spend the load model parameters α about each node_iThe spatial distribution of sensitivity absolute value as parameter calibration in this mode Navigation(Calibration Navigator), as shown in formula (3)；By angle stability nargin under dominant pattern about each joint The load model parameters α of point_iThe spatial distribution of sensitivity symbol as parameter invariant feature on each node under this pattern(Characteristic of Stability), as shown in formula (4)；

$\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^j}=\frac{{\mathrm{\η}}_S({\mathrm{\α}}_i^j+{\mathrm{\Δ\α}}_i^j)-{\mathrm{\η}}_S\left({\mathrm{\α}}_i^j\right)}{{\mathrm{\Δ\α}}_i^j}---\left(2\right)$

${\mathrm{CN}}_{{\mathrm{\α}}_i}=\left\{\right|\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^1}|,|\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^2}|,...,|\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^j}|,...,|\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^n}\left|\right\}---\left(3\right)$

${\mathrm{CS}}_{{\mathrm{\α}}_i}=\{\mathrm{sgn}\left(\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^1}\right),\mathrm{sgn}\left(\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^2}\right),\mathrm{...},\mathrm{sgn}\left(\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^j}\right),\mathrm{...},\mathrm{sgn}\left(\frac{\∂{\mathrm{\η}}_S}{\∂{\mathrm{\α}}_i^n}\right)\}---\left(4\right)$

Wherein,Expression system jth (j=1,2 ..., n) individual load bus i-th (i=1,2 ..., m) individual model ginseng Number,For node load parameterPerturbation, n is all load bus numbers to be calibrated of system, m be each load joint The number of contained model parameter on point, wherein i-th load model parameters is α_i；

According toWithThe load model parameters α of each node of system under dominant pattern can be judged respectively_iTo being The spatial distribution of system dynamic effect size and Orientation, suitably chooses about load model parameters α_iLeading calibration node ratio θ_i, TakeIn the maximum front n θ of each element_iIndividual node is about load model parameters α_iLeading calibration node, by the correspondence of CN Position element puts 1, and remaining element is non-dominant node, and the corresponding position element of CN sets to 0, and obtains about load model parameters α_iPilot bus mark(Dominant Buses Symbol) sequence；

The Pilot bus flag sequence DBS of system by each parameter Pilot bus flag sequence or computing obtain, if being Altogether there is m parameter (α to be calibrated₁,...,α_m), then shown in the Pilot bus flag sequence such as formula (5) of system；

In order to improve computational efficiency, by controlling θ_iValue, by the leading calibration region size limit of system necessarily In the range of, DBS will exist more null value；

Step 3, the two phase classification of load to be calibrated in leading calibration areas, travel through all m parameter to be calibrated, expired Foot formula (6) makes calibration navigation take maximum CN_MAXParameter, be defined as system instruction parameter alpha_s；

${\mathrm{CN}}_{MAX}=max\{max\left({\mathrm{CN}}_{{\mathrm{\α}}_1}\right),max\left({\mathrm{CN}}_{{\mathrm{\α}}_2}\right),...,max\left({\mathrm{CN}}_{{\mathrm{\α}}_s}\right),...,max\left({\mathrm{CN}}_{{\mathrm{\α}}_m}\right)\}---\left(6\right)$

Utilize the invariant feature of system instruction parameterPressIn each load bus positive and negative by leading for system calibration The load in region is divided into two classes, utilize Fuzzy C-Means Cluster Algorithm respectively in all kinds of loading zones further according to load bus Load nature of electricity consumed composition X_j=(x_j1,x_j2,x_j3,x_j4) classify, wherein x_j1,x_j2,x_j3,x_j4Represent that jth is born respectively The upper industry of lotus point, business, agricultural and the electricity consumption proportion shared by other load, obtain load classification total in system dominates calibration areas Number is C, c_k(c_k=1,2 ..., C) type load is WhereinFor belonging to c_kThe load number of class,It is c_kIn type loadIndividual load；

Step 4, the unified calibration of each type load, each load model of system non-dominant calibration region in leading calibration region Without checking, being maintained as the parameter value of analogue system in step 1, in system dominates calibration region, all of same class load are born Lotus node is set to identical model parameter, and the model parameter of C type load node is calibrated in unification, if c_kType load node needs school AccurateIndividual parameter with vector representation isThe then Headquarters of the General Staff of the required calibration of system Number isAgain by formula (1) as the object function of parametric calibration, then using default Parameter Typical as at the beginning of calibration Initial value, utilizes gradient descent method to be calculated parameter vector set corresponding to min (D) For parametric calibration result；

Step 5, load model based on many disturbances scene is calibrated, and such as r < R, imports next group disturbance measurement data, repeats Above-mentioned steps 1-4, the load model parameters of the leading calibration region of calibration correspondence under each disturbance, along with constantly tiring out of disturbance record Long-pending, calibrate and update whole system load model；Such as r >=R, then load model calibration terminates.

The technical characterstic of the present invention and beneficial effect: with angle stability under the dominant pattern of system overall situation phase-swing curves calculating Nargin reflection wide area system load model response characteristic, can reflect the impact of whole wide area system load, make the overall situation of WAMS Measurement is fully used；Stability margin, at the Continuous property of parametric stability region boundary, makes the object function energy set up accordingly Enough accurate instruction load parameter calibrations；Under dominant pattern angle stability nargin about each node load parameter sensitivity as The calibration navigation of wide area power system load model, with the leading calibration region of its compartment system load model and load bus Invariant feature, model calibration process is launched only for the load of leading calibration region, is greatly reduced wide area system load model The dimension of calibration object, shields the parameter impact of non-dominant calibration region, improves model calibration precision；In leading calibration region The middle invariant feature according to each load bus and load nature of electricity consumed composition carry out two phase classification, same class load bus phase Same model parameter carries out unifying calibration, reduces load model parameter to be calibrated the most further, and raising wide area power system is born The efficiency of lotus model calibration and precision.Progressively the load model of wide area system is calibrated based on many disturbances scene, finally give Load model parameters can simulate wide area system region load accurately, reliably, and calibration result can be according to different demand application In power system design, plan, the field such as operation, relevant computational accuracy and prediction accuracy will be significantly improved, improve system System safety and economy.

Claims (4)

1. a wide area power system load model calibration steps based on calibration navigation, it is characterised in that comprise the steps:

Step 1, the foundation of load parameter error criterion, WAMS wide area measurement and numerical integration respectively obtain current same disturbance Dis_r(r=1,2 ..., R) under the actual measurement track of system and simulated response, system surveys rail to use extended equal area criterion to determine The system dominant mode of mark, and calculate the angle stability nargin that under this system dominant mode, real system and analogue system are respectively put, Calculating parameter error D according still further to formula (1) is:

$D=N^{-1}\underset{l=1}{\overset{N}{\&Sigma;}}|{\mathrm{\&eta;}}_{Ml}-{\mathrm{\&eta;}}_{Sl}|---\left(1\right)$

In formula (1), η_MlAnd η_SlIt is respectively actual measurement track and the angle stability nargin of simulation track l pendulum, N under system dominant mode =min{N_M,N_S, N_MAnd N_SIt is respectively merit angle actual measurement track and total pendulum of simulation track in watch window, if D > error Threshold epsilon, then enter step 2 and start load model calibration, otherwise waits for noisy data next time and repeats step 1；

Step 2, system dominates calibration region identification, and computer sim-ulation system angle stability nargin under system dominant mode isCalculate angle stability nargin under system dominant mode according still further to formula (2) to join about node load NumberParametric sensitivity be:

$\frac{\&part;{\mathrm{\&eta;}}_S}{\&part;{\mathrm{\&alpha;}}_i^j}=\frac{{\mathrm{\&eta;}}_S({\mathrm{\&alpha;}}_i^j+{\mathrm{\&Delta;\&alpha;}}_i^j)-{\mathrm{\&eta;}}_S\left({\mathrm{\&alpha;}}_i^j\right)}{{\mathrm{\&Delta;\&alpha;}}_i^j}---\left(2\right)$

In formula (2),Expression system jth (j=1,2 ..., n) individual load bus i-th (i=1,2 ..., m) individual model parameter,For node load parameterPerturbation, n is all load bus numbers to be calibrated of system, and m is each load bus Upper contained model parameter number, wherein i-th load model parameters is α_i, then calculate angle stability nargin under system dominant mode About each load model parameters α_iThe spatial distribution of sensitivity absolute value, and using this spatial distribution as load model parameters α_i Calibration navigation under system dominant modeFor:

${\mathrm{CN}}_{{\mathrm{\&alpha;}}_i}=\left\{\left|\frac{\&part;{\mathrm{\&eta;}}_S}{\&part;{\mathrm{\&alpha;}}_i^1}\right|,\left|\frac{\&part;{\mathrm{\&eta;}}_S}{\&part;{\mathrm{\&alpha;}}_i^2}\right|,...,\left|\frac{\&part;{\mathrm{\&eta;}}_S}{\&part;{\mathrm{\&alpha;}}_i^j}\right|,...,\left|\frac{\&part;{\mathrm{\&eta;}}_S}{\&part;{\mathrm{\&alpha;}}_i^n}\right|\right\}---\left(3\right)$

Calculate under system dominant mode angle stability nargin again about the load model parameters α of each node_iThe sky of sensitivity symbol Between be distributed, and using this spatial distribution as load model parameters α under system dominant mode_iInvariant feature on each node For:

${\mathrm{CS}}_{{\mathrm{\&alpha;}}_i}=\{\mathrm{sgn}\left(\frac{\&part;{\mathrm{\&eta;}}_S}{\&part;{\mathrm{\&alpha;}}_i^1}\right),\mathrm{sgn}\left(\frac{\&part;{\mathrm{\&eta;}}_S}{\&part;{\mathrm{\&alpha;}}_i^2}\right),...,\mathrm{sgn}\left(\frac{\&part;{\mathrm{\&eta;}}_S}{\&part;{\mathrm{\&alpha;}}_i^j}\right),...,\mathrm{sgn}\left(\frac{\&part;{\mathrm{\&eta;}}_S}{\&part;{\mathrm{\&alpha;}}_i^n}\right)\}---\left(4\right)$

Set about load model parameters α_iLeading calibration node ratio θ_i, rightIn each element by arranging the most successively Row, take front n θ_iIndividual node is about load model parameters α_iLeading calibration node set, by all m load model parameters α_i (i=1,2 ..., m) corresponding leading calibration node set composition system dominates calibration region；

Step 3, system is the two phase classification of load to be calibrated in dominating calibration region, travels through all m load model parameters α_i, Being met formula (6) makes calibration navigateTake maximum CN_MAXParameter, and define this parameter be system indicate parameter alpha_s, CN_MAXComputing formula be:

${\mathrm{CN}}_{MAX}=max\{max\left({\mathrm{CN}}_{{\mathrm{\&alpha;}}_1}\right),max\left({\mathrm{CN}}_{{\mathrm{\&alpha;}}_2}\right),...,max\left({\mathrm{CN}}_{{\mathrm{\&alpha;}}_s}\right),...,max\left({\mathrm{CN}}_{{\mathrm{\&alpha;}}_m}\right)\}---\left(6\right)$

Reutilization system instruction parameter alpha_sInvariant featureBy invariant featureIn each load bus positive and negative by system The load bus of leading calibration region is divided into two classes, further according to load nature of electricity consumed composition X of load bus_j, and utilize Fuzzy C Means clustering algorithm to two type load node subseries again, obtains load classification number total in system dominates calibration region respectively For C, c_k(c_k=1,2 ..., C) type load is WhereinFor belonging to c_kThe load number of class,It is c_kIn type loadIndividual load；

Step 4, the unified calibration of each type load in leading calibration region, each load model of system non-dominant calibration region without Checking, system is dominated all load buses of same class load in calibration region and is set to identical model parameter, more unified calibration C Model parameter on type load node, if c_kCalibration is needed on type load nodeIndividual model parameter with vector representation isThen total parameter of the required calibration of system isAgain by formula (1) As the object function of parametric calibration, then using default parameter value as the initial value of calibration, gradient descent method is utilized to calculate To the parameter vector set that min (D) is correspondingFor parametric calibration result；

Step 5, load model based on many disturbances scene is calibrated, and persistently records system disturbance track, utilizes under each disturbance Measurement data, repeats step 1-4, obtains correspondence system under each disturbance respectively and dominate the load model calibration result of calibration region, It is updated to simulation system data storehouse, along with the continuous accumulation of disturbance record, calibrates and update whole system load model.

Wide area power system load model calibration steps based on calibration navigation the most according to claim 1, its feature exists In, in step 2, willIn the corresponding position element of leading calibration node put 1, remaining element is non-dominant node, Corresponding position element set to 0, obtain about load model parameters α_iPilot bus flag sequenceThe master of system Lead node label sequence D BS by each load model parameters α_iPilot bus flag sequenceOr computing obtain.

Wide area power system load model calibration steps based on calibration navigation the most according to claim 1 and 2, its feature It is, in step 3, load nature of electricity consumed composition X_j=(x_j1,x_j2,x_j3,x_j4), wherein x_j1、x_j2、x_j3And x_j4Represent jth respectively Industry on individual load bus, business, agricultural and the electricity consumption proportion shared by other load.

Wide area power system load model calibration steps based on calibration navigation the most according to claim 1 and 2, its feature It is, in step 3, by invariant featureIn the positive and negative load that system is dominated calibration region of each load bus be divided into two classes It is respectively as follows: on a type load node, stability margin η under system dominant mode_SParameter alpha is indicated about this category node_sPartial derivative It is nonnegative value；On another kind of load bus, stability margin η under system dominant mode_SParameter alpha is indicated about this category node_s's Partial derivative is negative value.