CN105808833A - Online parameter identification method of parallel synchronous generator based on multi-data sets - Google Patents

Abstract

This invention discloses an online parameter identification method of a parallel synchronous generator based on multi-data sets. The method comprises the following steps: representing a dynamic process of the generator and determining a discrete parameter identification target through constructing a discrete dynamic equation of multiple data sets; and online identifying the parameter of the generator through calculating the discrete dynamic equation and the Jacobian matrix and the Hessian matrix of the discrete identification target to the parameter of each data set. The method requires no use of an extra field test, protects the generator equipment, has higher parameter identification precision, and overcomes the defect that the traditional online parameter identification method of the generator is not precise. This method mines the loose coupling relationship in the parameter identification of the multi-data sets, and effectively deploys the calculations of different data sets on different calculation units to realize the parallel computing of the multiple data sets; thus, the calculation efficiency is greatly improved; and the method is applicable to the fields of dynamic modelling of a power system, and the like.

Description

A kind of based on many data union of sets row synchronous generator on-line parameter discrimination method

Technical field

The invention belongs to the Identifying Dynamical Parameters field of power system, especially a kind of based on many data union of sets row synchronous generator on-line parameter discrimination method.

Background technology

Power System Dynamic Simulation is a kind of most important instrument for operation of power networks and Electric Power Network Planning.Electric power system dispatching department determines scheduling strategy, Electric Power Network Planning and Unit Combination according to the result of emulation.Therefore the accuracy of dynamic simulation directly affects economy and the safety of this operation of power networks.And an accurate generator model is exactly a simulation result basis accurately, and the extensive use along with measuring units such as phasor measurement units PMU, the Dynamic Signal of synchronous generator can accurate acquisition be sent to dispatching of power netwoks department so that the on-line parameter identification of synchronous generator is increasingly studied widely.But no matter it is the mode using on-the-spot test, or the mode using Power System Disturbances data realizes electromotor on-line parameter identification, they all inevitably suffer a problem that: it is not often very sufficient that power system collects electromotor related data, cause that the identification result of all parameters of electromotor can not well identification.

For above-mentioned technical difficulty, document " OnlineEstimationofSynchronousGeneratorParametersUsingPRB SPerturbations " proposes to apply pseudo random binary signal to automatism voltage control device and carries out the on-line parameter identification of electromotor, but the size of the signal that this method applies needs artificial experience to determine, signal too small parameter is not easy identification, and signal then can cause synchronous generator permanent damage too greatly.Document " PMUBasedGeneratorParameterIdentificationtoImprovetheSyst emPlanningandOperation " proposes the trajectory sensitivity analysis Parameters variation the using parameter impact on output, the parameter of identification difference is fixed as nominal value, thus improving the identification effect of parameter, but this method can not identification these be set as the parameter of nominal value, more can not reflect the situation that parameter is actual.

Although existing many achievements in research, but there is no in synchronous generator on-line identification field not only convenient, but also can the on-line identification computational methods of accurate recognition generator parameter.

Summary of the invention

The technical problem to be solved is the defect overcoming above-mentioned prior art to exist, there is provided a kind of based on many data union of sets row synchronous generator on-line parameter discrimination method, its multiple data acquisition systems gathered according to synchronous generator on-line monitoring system, can efficiently and conveniently all parameters of identification synchronous generator, may be directly applied to synchronous generator on-line parameter identification and analyze.

For this, the present invention adopts the following technical scheme that: a kind of synchronous generator on-line parameter discrimination method based on many data acquisition systems, it comprises the steps:

The first step: when electrical network generation disturbance, generator terminal voltage, machine end electric current, machine end active power, reactive power and excitation voltage signal before and after grid disturbance are acquired by the synchronous generator on-line monitoring system utilizing power plant, and are stored in data base as disturbing signal；

Second step: set end voltage signal and excitation voltage signal in the repeatedly disturbing signal data acquisition system occur the disturbing signal data acquisition system currently gathered and history electrical network are divided into input signal u_i, other signal is then divided into output signal y_mi, wherein subscript i represents i-th data acquisition system, i=1,2 ... n_c, n_cRepresent the sum of data acquisition system；

3rd step: the input signal u according to i-th data acquisition system_iWith Generator Parameters variable θ, the subordination principle D under structure i-th data acquisition system_i(u_i, θ) characterize corresponding disturbance and issue motor dynamics process and output state；Output signal y according to i-th data acquisition system_miBuild parameter identification target

4th step: utilize trapezoid formula to subordination principle D_i(u_i, θ) and identification targetCarrying out sliding-model control, the Nonlinear System of Equations after discretization is usedAnd Ψ_iDescribe, wherein i=1,2 ... n_c,Represent the input signal after discretization, simultaneously initialization data collection variable s_iNonlinear System of Equations is made with parametric variable θSet up；

5th step: calculate the data set variable s under different pieces of information set_iWith parametric variable θ to object function Ψ_iAnd Nonlinear System of EquationsJacobian matrix L_iAnd J_i, and extend extra large gloomy matrixI=1,2 ... n_c, the data of different pieces of information set calculate and carry out on independent calculation processing unit；

6th step: by the Jacobian matrix L under different pieces of information set_iAnd J_i, and extend extra large gloomy matrixCollect, and calculate the renewal amount Δ θ of parametric variable by calculating process P1 and calculate the renewal amount Δ s of process P2 calculating data set variable_i；

7th step: if the renewal amount Δ θ of parametric variable is less than permissible range, then algorithmic statement, export identification result；Otherwise algorithm is not restrained, to data collection variable s_iUpdate with parametric variable θ, namely respectively plus the renewal amount Δ s of data set variable_iWith the renewal amount Δ θ of parametric variable, be then back to the 5th step.

The present invention uses history repeatedly noisy data, it is achieved the synchronous generator on-line parameter identification of many data acquisition systems, extends existing Parameter Identification Method of Synchronous Generator.

Further, the subordination principle D in the 3rd described step, under i-th data acquisition system_i(u_i, θ) and there is following form:

$D_i(u_i,\mathrm{\θ}):\left\{\begin{array}{c}0=F_i({\stackrel{\·}{z}}_i,z_i,u_i,\mathrm{\θ})\\ 0=G_i(y_i,z_i,u_i,\mathrm{\θ})\\ z_i\left(t_0\right)=z_{i0}\end{array}\right.,$

Wherein, subscript i represents i-th data acquisition system, z andRepresenting the derivative of the state variable in electromotor dynamic process and state variable, y is the output state of electromotor, and F characterizes the nonlinear function of the dynamic process of synchronous generator, and G is then the nonlinear function of the output defining synchronous generator, t₀For emulating the initial time of period, z_i0Initial value for state variable.

Further, the parameter identification target in the 3rd described step, under i-th data acquisition systemThere is following form:

Wherein, subscript i represents i-th data acquisition system, and y represents the output state of electromotor, y_miRepresenting the output signal measured, W represents the synchronous generator corresponding sampling error of on-line detecting system sample devices in power plant, T_fRepresent the end time of parameter identification, T representing matrix transposition.

Further, in the 4th described step, Nonlinear System of EquationsThere is following form:

$E_i({\hat{u}}_i,s_i,\mathrm{\θ}):\left\{\begin{array}{c}0=C_{i,t}(s_{i,t},s_{i,t-1},{\hat{u}}_{i,t},{\hat{u}}_{i,t-1}\mathrm{\θ})\\ =\left[\begin{array}{c}{\hat{F}}_{i,t}({\hat{z}}_{i,t},{\hat{z}}_{i,t-1},{\hat{u}}_{i,t},{\hat{u}}_{i,t-1}\mathrm{\θ})\\ {\hat{G}}_{i,t}({\hat{y}}_{i,t},{\hat{z}}_{i,t},{\hat{u}}_{i,t},\mathrm{\θ})\end{array}\right]\\ {\hat{z}}_{i,0}=z_{i0}\\ t=1,\mathrm{2...}{n}_t\end{array}\right.,$

Wherein, subscript i represents i-th data acquisition system, and subscript t represents the t moment,WithRepresent the state variable of discretization, input signal and output state, z respectively_i0For the initial value of state variable, n_tRepresent the sum of the moment after all discretizations,Represent the nonlinear function of the dynamic process characterizing synchronous generator after discretization,Represent the nonlinear function of the output of the definition synchronous generator after discretization, C represent byWithThe discretization nonlinear function constituted, s_iRepresenting i-th data set variable, its form is as follows:

$s_i={\left[\begin{array}{ccccc}{{\hat{z}}_{i,1}}^T& {{\hat{y}}_{i,1}}^T& \mathrm{...}& {{\hat{z}}_{i,n_t}}^T& {{\hat{y}}_{i,n_t}}^T\end{array}\right]}^T.$

Further, in the 4th described step, Nonlinear System of Equations Ψ_iThere is following form:

${\mathrm{\Ψ}}_i=\underset{t=1}{\overset{n_t}{\Σ}}{(y_{i,t}-y_{mi,t})}^T{W}^{-1}(y_{i,t}-y_{mi,t}),$

Wherein, subscript i represents i-th data acquisition system, and subscript t represents the t moment, and y represents the output state of electromotor, y_mRepresenting the output signal measured, W represents the synchronous generator corresponding sampling error of on-line detecting system sample devices in power plant, n_tRepresent the sum of the moment after all discretizations.

Further, in the 5th described step, Jacobian matrix L_iAnd J_iThere is following form:

$L_i=\left[\begin{array}{c}\frac{\∂{\mathrm{\Ψ}}_i}{\∂\mathrm{\θ}}=0\\ \frac{\∂{\mathrm{\Ψ}}_i}{\∂s_i}\end{array}\right],J_i=\left[\begin{array}{cc}\frac{\∂C_i}{\∂\mathrm{\θ}}& \frac{\∂C_i}{\∂s_i}\end{array}\right],C_i=\left[\begin{array}{c}{C}_{i,1}\\ .\\ .\\ .\\ C_{i,t}\\ .\\ .\\ .\\ C_{i,n_t}\end{array}\right],$

Wherein, subscript i represents i-th data acquisition system, and subscript t represents the t moment,Representing partial derivative operator, s represents data set variable, and θ represents that parametric variable, Ψ represent discretization object function, and C represents the discretization nonlinear function characterizing electromotor dynamic characteristic and output characteristics.

Further, in the 5th described step, extend extra large gloomy matrixThere is following form:

${\hat{H}}_i=\left[\begin{array}{cc}{H}_{\mathrm{\θ}\mathrm{\θ},i}& H_{\mathrm{\θ}i}\\ {H_{\mathrm{\θ}i}}^T& H_i\end{array}\right],$

Wherein, subscript i represents i-th data acquisition system, H_θθRepresent discretization nonlinear function C the gloomy matrix in sea to parametric variable, H_θiRepresent the discretization nonlinear function C gloomy matrix in sea to parametric variable and data collection variable, H represent discretization nonlinear function C only the gloomy matrix in sea of data collection variable is only added by the gloomy matrix in sea of data collection variable with discretization object function Ψ and, T representing matrix transposition.

Further, in the 5th described step, independent calculation processing unit refer to core cpu, calculation server node or other there is the computing equipment of complete Floating-point Computation and logic processing capability.

Further, in the 6th described step, calculate process P1 and there is following form:

Δ θ=A^-1b

$A=\underset{i=1}{\overset{n_c}{\Σ}}(H_{\mathrm{\θ}\mathrm{\θ},i}+{\left(\frac{\∂{C_i}^{-1}}{\∂s_i}\frac{\∂C_i}{\∂\mathrm{\θ}}\right)}^T{H}_i{\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}}-{\left({\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}}\right)}^T{H}_{\mathrm{\θ}i}^T-H_{\mathrm{\θ}i}{\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}})$

$b=\underset{i=1}{\overset{n_c}{\Σ}}({\frac{\∂C_i}{\∂\mathrm{\θ}}}^T{\mathrm{\λ}}_i+H_{\mathrm{\θ}i}^T{\frac{\∂C_i}{\∂s_i}}^{-1}{C}_i-{\left({\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}}\right)}^T(\frac{\∂{\mathrm{\Ψ}}_i}{\∂s_i}-{\frac{\∂C_i}{\∂s_i}}^T{\mathrm{\λ}}_i+H_i{\frac{\∂C_i}{\∂s_i}}^{-1}{C}_i\left)\right)$

Wherein, subscript i represents i-th data acquisition system, i=1,2 ... n_c, n_cRepresenting the sum of data acquisition system, C represents that discretization nonlinear function characterizes the discretization nonlinear function of electromotor dynamic characteristic and output characteristics, and Ψ represents discretization object function, and s represents data set variable, and θ represents parametric variable, H_θθRepresent discretization nonlinear function C the gloomy matrix in sea to parametric variable, H_θiRepresent the discretization nonlinear function C gloomy matrix in sea to parametric variable and data collection variable, H represent discretization nonlinear function C only the gloomy matrix in sea of data collection variable is only added by the gloomy matrix in sea of data collection variable with discretization object function Ψ and, λ_iRepresent C_iCorresponding Lagrange multiplier ,-1 representing matrix is inverted,Represent partial derivative operator, T representing matrix transposition.

Further, in the 6th described step, calculate process P2 and there is following form:

${\mathrm{\Δs}}_i=-{\frac{\∂C_i}{\∂s_i}}^{-1}{C}_i-{\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}}\mathrm{\Δ}\mathrm{\θ}.$

Wherein, subscript i represents i-th data acquisition system, i=1,2 ... n_c, C represents that discretization nonlinear function characterizes the discretization nonlinear function of electromotor dynamic characteristic and output characteristics, and s represents data set variable, and θ represents parametric variable, and Δ θ represents the renewal amount of parametric variable, and-1 representing matrix is inverted,Represent partial derivative operator.

Compared with existing technology, the present invention mainly has following improvement:

1. the present invention uses many data acquisition systems to carry out generator parameter on-line identification, it is not necessary to extra on-the-spot test, protects gen-set.

2. present invention can apply to synchronous generator on-line parameter identification field, the energy all parameters of identification, and there is higher parameter identification precision.

3. the parallel computation mode based on many data sets proposed by the invention, has speed-up ratio height, a technical advantage that parallel efficiency is high.

Accompanying drawing explanation

Fig. 1 is the flow chart of the present invention.

Fig. 2 is single data acquisition system and many data acquisition systems Identification Errors scattergram.

Fig. 3 is that the present invention tests example parallel speedup ratio curve.

Detailed description of the invention

Below in conjunction with specification drawings and specific embodiments, the present invention is elaborated.

The synchronous generator on-line parameter discrimination method based on many data acquisition systems as shown in Figure 1, its step is as follows:

The first step: when electrical network generation disturbance, utilizes the synchronous generator on-line monitoring system in power plant that generator terminal voltage, machine end electric current, machine end active power, reactive power and excitation voltage signal before and after grid disturbance are acquired, and is stored in data base；

Second step: set end voltage signal and excitation voltage signal in the repeatedly noisy data set occur the data acquisition system currently gathered and history electrical network are divided into input signal u_i, other signal is then divided into output signal y_mi, wherein subscript i represents i-th data acquisition system, i=1,2 ... n_c, n_cRepresenting the sum of data acquisition system, subscript m represents measures the output signal arrived；

3rd step: the input signal u according to i-th data acquisition system_iWith Generator Parameters variable θ, the subordination principle D under structure i-th data acquisition system_i(u_i, θ) characterize corresponding disturbance and issue motor dynamics process and output state；Output signal y according to i-th data acquisition system_miConstructing variable identification target

4th step: utilize trapezoid formula to subordination principle D_i(u_i, θ) and identification targetCarrying out sliding-model control, the Nonlinear System of Equations after discretization is usedAnd Ψ_iDescribe, wherein i=1,2 ... n_c,Represent the input signal after discretization, simultaneously initialization data collection variable s_iNonlinear System of Equations is made with parametric variable θSet up；

5th step: calculate the data set variable s under different pieces of information set_iWith parametric variable θ to object function Ψ_iAnd Nonlinear System of EquationsJacobian matrix L_iAnd J_i, and extend extra large gloomy matrixI=1,2 ... n_c, the data of different pieces of information set calculate and carry out on independent calculation processing unit；

6th step: by the Jacobian matrix L under different pieces of information set_iAnd J_i, and extend extra large gloomy matrixCollect, and calculate the renewal amount Δ θ of parameter by calculating process P1 and calculate the renewal amount Δ s of process P2 calculating data set variable_i；

7th step: if the renewal amount Δ θ of parametric variable is less than permissible range, then algorithmic statement, export identification result；Otherwise algorithm is not restrained, to data collection variable s_iUpdate with parametric variable θ, namely respectively plus the renewal amount Δ s of data set variable_iWith parametric variable renewal amount Δ θ, be then back to the 5th step.

Application examples

Use the exploitation of Matlab programming language to achieve the calculation procedure of synchronous generator on-line parameter discrimination method (i.e. the present invention) based on many data acquisition systems, and use one to be equipped withThe PC of X56502.67HzCPU and 24GB internal memory completes test and the checking of the present invention.

In implementation process, the set end voltage signal gather the synchronous generator on-line detecting system in power plant and excitation voltage signal are as the input data of electromotor, and other gather signal, such as machine end electric current, active power, reactive power is as the output data of electromotor.The synchronous generator that test uses is furnished with corresponding excitation unit and arrangements for speed regulation, use the complex synchronization generator model on 4 axle 6 rank for describing the output procedure of the dynamic process of electromotor in synchronous machine modeling process, the information of the synchronous generator used by test that table 1 represents.

Table 1: the Generator Parameters nominal value that test uses

Generator parameter	Parameter declaration	Parametric nominal value
			Xd	D-axis reactance	1.8000 5 -->
Xq	Quadrature axis reactance	1.7500
			X’d	D-axis transient state reactance	0.3000
X’q	Quadrature axis transient state reactance	0.4700
			X”	Subtranient reactance	0.2300
T’d0	Direct-axis transient time constant	4.8000
			T”d0	Direct-axis subtransient time constant	0.0350
T’q0	Quadrature axis time constant	1.5000
			T”q0	Quadrature axis time time constant	0.0700
Tj	Rotary inertia	6.4000

For the synchronous generator of test, current single noisy data set discrimination method and the many data collection approach in conjunction with 23 noisy data set of history generation is taked to carry out Parameter Estimation of Synchronous Machines respectively.Table 2 illustrates the result at single data acquisition system discrimination method and many data acquisition systems discrimination method.

Table 2: Parameter Estimation of Synchronous Machines result

Preliminary from the above it can be seen that the identification effect of transient state in single data acquisition system identification and time time constant is than the weak effect of many data acquisition systems identification, but other parameter identification difference are inconspicuous.For unified Analysis Identification Errors result, Fig. 2 illustrates the relative error distribution of two kinds of methods.Therefrom can finding out it is no matter electromotor reactance parameter or electromotor transient state and time transient state time parameter intuitively, the effects gathering identifications are better than the identification effect of singleton more.

Further, the example of the data acquisition system of varying number is for computational efficiency and the parallel speedup ratio of testing the present invention.Table 3 gives 12 data acquisition systems and 24 data are integrated into the calculating time under different core cpu quantity.It is observed that on the one hand along with the increase that data acquisition system uses, the calculating time is to increase in table, CPU calculates the increase of core on the other hand, and the used time of parameter identification is in minimizing trend.

Table 3: the parallel computation time of Parameter Estimation of Synchronous Machines

Calculated performance data in table 3 are carried out visualization processing further, obtains parallel speedup ratio curve, as shown in Figure 3.Fig. 3 illustrates the test effect of varying number data acquisition system, it can be seen that along with the increase of CPU number use, the increasing degree of parallel speedup ratio is slack-off due to the communication between CPU core, but it is as the increase of data acquisition system, parallel speedup ratio effect increases substantially, illustrates that the present invention has good parallel scalability.

Claims (10)

1., based on many data union of sets row synchronous generator on-line parameter discrimination method, it comprises the steps:

The first step: when electrical network generation disturbance, generator terminal voltage, machine end electric current, machine end active power, reactive power and excitation voltage signal before and after grid disturbance are acquired by the synchronous generator on-line monitoring system utilizing power plant, and are stored in data base as disturbing signal；

Second step: set end voltage signal and excitation voltage signal in the repeatedly disturbing signal data acquisition system occur the disturbing signal data acquisition system currently gathered and history electrical network are divided into input signal u_i, other signal is then divided into output signal y_mi, wherein subscript i represents i-th data acquisition system, i=1,2 ... n_c, n_cRepresent the sum of data acquisition system；

3rd step: the input signal u according to i-th data acquisition system_iWith Generator Parameters variable θ, the subordination principle D under structure i-th data acquisition system_i(u_i, θ) characterize corresponding disturbance and issue motor dynamics process and output state；Output signal y according to i-th data acquisition system_miBuild parameter identification target

4th step: utilize trapezoid formula to subordination principle D_i(u_i, θ) and identification targetCarrying out sliding-model control, the Nonlinear System of Equations after discretization is usedAnd Ψ_iDescribe, wherein i=1,2 ... n_c,Represent the input signal after discretization, simultaneously initialization data collection variable s_iNonlinear System of Equations is made with parametric variable θSet up；

5th step: calculate the data set variable s under different pieces of information set_iWith parametric variable θ to object function Ψ_iAnd Nonlinear System of EquationsJacobian matrix L_iAnd J_i, and extend extra large gloomy matrixI=1,2 ... n_c, the data of different pieces of information set calculate and carry out on independent calculation processing unit；

6th step: by the Jacobian matrix L under different pieces of information set_iAnd J_i, and extend extra large gloomy matrixCollect, and calculate the renewal amount Δ θ of parametric variable by calculating process P1 and calculate the renewal amount Δ s of process P2 calculating data set variable_i；

7th step: if the renewal amount Δ θ of parametric variable is less than permissible range, then algorithmic statement, export identification result；Otherwise algorithm is not restrained, to data collection variable s_iUpdate with parametric variable θ, namely respectively plus the renewal amount Δ s of data set variable_iWith the renewal amount Δ θ of parametric variable, be then back to the 5th step.

2. synchronous generator on-line parameter discrimination method according to claim 1, it is characterised in that: the subordination principle D in the 3rd described step, under i-th data acquisition system_i(u_i, θ) and there is following form:

$D_i(u_i,\mathrm{\θ}):\left\{\begin{array}{c}0=F_i({\stackrel{\·}{z}}_i,z_i,u_i,\mathrm{\θ})\\ 0=G_i(y_i,z_i,u_i,\mathrm{\θ})\\ z_i\left(t_0\right)=z_{i0}\end{array}\right.,$

Wherein, subscript i represents i-th data acquisition system, z andRepresenting the derivative of the state variable in electromotor dynamic process and state variable, y is the output state of electromotor, and F characterizes the nonlinear function of the dynamic process of synchronous generator, and G is then the nonlinear function of the output defining synchronous generator, t₀For emulating the initial time of period, z_i0Initial value for state variable.

3. synchronous generator on-line parameter discrimination method according to claim 1, it is characterised in that: the parameter identification target in the 3rd described step, under i-th data acquisition systemThere is following form:

Wherein, subscript i represents i-th data acquisition system, and y represents the output state of electromotor, y_miRepresenting the output signal measured, W represents the synchronous generator corresponding sampling error of on-line detecting system sample devices in power plant, T_fRepresent the end time of parameter identification, T representing matrix transposition.

4. synchronous generator on-line parameter discrimination method according to claim 1, it is characterised in that: in the 4th described step, Nonlinear System of EquationsThere is following form:

$E_i({\hat{u}}_i,s_i,\mathrm{\θ}):\left\{\begin{array}{c}0=C_{i,t}(s_{i,t},s_{i,t-1},{\hat{u}}_{i,t},{\hat{u}}_{i,t-1}\mathrm{\θ})\\ =\left[\begin{array}{c}{\hat{F}}_{i,t}({\hat{z}}_{i,t},{\hat{z}}_{i,t-1},{\hat{u}}_{i,t},{\hat{u}}_{i,t-1}\mathrm{\θ})\\ {\hat{G}}_{i,t}({\hat{y}}_{i,t},{\hat{z}}_{i,t},{\hat{u}}_{i,t},\mathrm{\θ})\end{array}\right]\\ {\hat{z}}_{i,0}=z_{i0}\\ t=1,2...n_t\end{array}\right.,$

Wherein, subscript i represents i-th data acquisition system, and subscript t represents the t moment,WithRepresent the state variable of discretization, input signal and output state, z respectively_i0For the initial value of state variable, n_tRepresent the sum of the moment after all discretizations,Represent the nonlinear function of the dynamic process characterizing synchronous generator after discretization,Represent the nonlinear function of the output of the definition synchronous generator after discretization, C represent byWithThe discretization nonlinear function constituted, s_iRepresenting i-th data set variable, its form is as follows:

$s_i={\left[\begin{array}{ccccc}{{\hat{z}}_{i,1}}^T& {{\hat{y}}_{i,1}}^T& ...& {{\hat{z}}_{i,n_t}}^T& {{\hat{y}}_{i,n_t}}^T\end{array}\right]}^T.$

5. synchronous generator on-line parameter discrimination method according to claim 1, it is characterised in that: in the 4th described step, Nonlinear System of Equations Ψ_iThere is following form:

${\mathrm{\Ψ}}_i=\underset{t=1}{\overset{n_t}{\Σ}}{(y_{i,t}-y_{mi,t})}^T{W}^{-1}(y_{i,t}-y_{mi,t}),$

Wherein, subscript i represents i-th data acquisition system, and subscript t represents the t moment, and y represents the output state of electromotor, y_mRepresenting the output signal measured, W represents the synchronous generator corresponding sampling error of on-line detecting system sample devices in power plant, n_tRepresent the sum of the moment after all discretizations.

6. synchronous generator on-line parameter discrimination method according to claim 1, it is characterised in that: in the 5th described step, Jacobian matrix L_iAnd J_iThere is following form:

$L_i=\left[\begin{array}{c}\frac{\∂{\mathrm{\Ψ}}_i}{\∂\mathrm{\θ}}\mathrm{=0}\\ \frac{\∂{\mathrm{\Ψ}}_i}{\∂s_i}\end{array}\right],J_i=\left[\begin{array}{cc}\frac{\∂C_i}{\∂\mathrm{\θ}}& \frac{\∂C_i}{\∂s_i}\end{array}\right],C_i=\left[\begin{array}{c}{C}_{i,1}\\ \·\\ \·\\ \·\\ C_{i,t}\\ \·\\ \·\\ \·\\ C_{i,n_t}\end{array}\right],$

Wherein, subscript i represents i-th data acquisition system, and subscript t represents the t moment,Representing partial derivative operator, s represents data set variable, and θ represents that parametric variable, Ψ represent discretization object function, and C represents the discretization nonlinear function characterizing electromotor dynamic characteristic and output characteristics.

7. synchronous generator on-line parameter discrimination method according to claim 1, it is characterised in that: in the 5th described step, extend extra large gloomy matrixThere is following form:

${\hat{H}}_i=\left[\begin{array}{cc}{H}_{\mathrm{\θ}\mathrm{\θ},i}& H_{\mathrm{\θ}i}\\ {H_{\mathrm{\θ}i}}^T& H_i\end{array}\right],$

Wherein, subscript i represents i-th data acquisition system, H_θθRepresent discretization nonlinear function C the gloomy matrix in sea to parametric variable, H_θiRepresent the discretization nonlinear function C gloomy matrix in sea to parametric variable and data collection variable, H represent discretization nonlinear function C only the gloomy matrix in sea of data collection variable is only added by the gloomy matrix in sea of data collection variable with discretization object function Ψ and, T representing matrix transposition.

8. synchronous generator on-line parameter discrimination method according to claim 1, it is characterized in that: in the 5th described step, independent calculation processing unit refer to core cpu, calculation server node or other there is the computing equipment of complete Floating-point Computation and logic processing capability.

9. synchronous generator on-line parameter discrimination method according to claim 1, it is characterised in that: in the 6th described step, calculate process P1 and there is following form:

Δ θ=A^-1b

$A=\underset{i=1}{\overset{n_c}{\mathrm{\Σ}}}(H_{\mathrm{\θ}\mathrm{\θ},i}+{\left({\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}}\right)}^T{H}_i{\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}}-{\left({\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}}\right)}^T{H}_{\mathrm{\θ}i}^T-H_{\mathrm{\θ}i}{\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}})$

$b=\underset{i=1}{\overset{n_c}{\mathrm{\Σ}}}({\frac{\∂C_i}{\∂\mathrm{\θ}}}^T{\mathrm{\λ}}_i+H_{\mathrm{\θ}i}^T{\frac{\∂C_i}{\∂s_i}}^{-1}{C}_i-{\left({\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}}\right)}^T(\frac{\∂{\mathrm{\Ψ}}_i}{\∂s_i}-{\frac{\∂C_i}{\∂s_i}}^T{\mathrm{\λ}}_i+H_i{\frac{\∂C_i}{\∂s_i}}^{-1}{C}_i\left)\right)$

Wherein, subscript i represents i-th data acquisition system, i=1,2 ... n_c, n_cRepresenting the sum of data acquisition system, C represents that discretization nonlinear function characterizes the discretization nonlinear function of electromotor dynamic characteristic and output characteristics, and Ψ discretization object function, s represents data set variable, and θ represents parametric variable, H_θθRepresent discretization nonlinear function C the gloomy matrix in sea to parametric variable, H_θiRepresent the discretization nonlinear function C gloomy matrix in sea to parametric variable and data collection variable, H represent discretization nonlinear function C only the gloomy matrix in sea of data collection variable is only added by the gloomy matrix in sea of data collection variable with discretization object function Ψ and, λ_iRepresent C_iCorresponding Lagrange multiplier ,-1 representing matrix is inverted,Represent partial derivative operator, T representing matrix transposition.

10. synchronous generator on-line parameter discrimination method according to claim 1, it is characterised in that: in the 6th described step, calculate process P2 and there is following form:

${\mathrm{\Δs}}_i=-{\frac{\∂C_i}{\∂s_i}}^{-1}{C}_i-{\frac{\∂C_i}{\∂s_i}}^{-1}\frac{\∂C_i}{\∂\mathrm{\θ}}\mathrm{\Δ}\mathrm{\θ},$

Wherein, subscript i represents i-th data acquisition system, i=1,2 ... n_c, C represents that discretization nonlinear function characterizes the discretization nonlinear function of electromotor dynamic characteristic and output characteristics, and s represents data set variable, and θ represents parametric variable, and Δ θ represents the renewal amount of parametric variable, and-1 representing matrix is inverted,Represent partial derivative operator.