CN113971259A

CN113971259A - Error reliability parameter identification and correction method for power generation and transmission system element

Info

Publication number: CN113971259A
Application number: CN202111234908.5A
Authority: CN
Inventors: 胡博; 谢开贵; 李春燕; 邵常政; 曹侃; 周鲲鹏; 彭吕斌; 林铖嵘; 卢慧; 李维展
Original assignee: Chongqing University; State Grid Corp of China SGCC; Electric Power Research Institute of State Grid Hubei Electric Power Co Ltd
Current assignee: Chongqing University; State Grid Corp of China SGCC; Electric Power Research Institute of State Grid Hubei Electric Power Co Ltd
Priority date: 2021-10-22
Filing date: 2021-10-22
Publication date: 2022-01-25
Anticipated expiration: 2041-10-22
Also published as: CN113971259B

Abstract

The application discloses a method for identifying and correcting error reliability parameters of power generation and transmission system elements, which comprises the following steps: identifying and correcting error reliability parameters to construct a nonlinear optimization problem which uses the reliability index estimated value and the statistic value deviation to be minimized as a target function, and judging whether error parameters exist in the element reliability parameters of the power generation and transmission system; reliability parameters of the power generation and transmission system are simplified; dividing the reliability parameters into sensitive parameters and non-sensitive parameters according to the sensitivity; estimating parameters of a unit connected with the load nodes, estimating and identifying sensitive parameters, estimating all parameters, identifying error parameters and correcting the error parameters by adopting an interval algorithm. The invention considers the reliability evaluation inverse problem of the power transmission and generation system, and can identify and correct wrong element reliability parameters by starting from the reliability indexes of the power transmission and generation system.

Description

Error reliability parameter identification and correction method for power generation and transmission system element

Technical Field

The invention belongs to the technical field of power system reliability evaluation, and relates to a method for identifying and correcting error reliability parameters of power generation and transmission system elements.

Background

The inverse problem of the reliability of the power system refers to the use of the reliability index of the whole system (or node) to identify and correct the wrong element reliability parameters.

In power system reliability evaluation, the accuracy of element reliability parameters has a significant impact on system reliability evaluation, and inaccurate reliability parameters may result in diametrically opposite evaluation results. When the power system element is signed, the reliability parameter of the element is emphasized, and the parameter is often derived from the historical reliability statistical result of the same type of element. Wrong historical data can mislead bidding results and affect the reliability and safety of future power systems.

The component reliability parameters are often obtained from historical outage records, but the data quality is not uniform. On the one hand, this is because the utility companies need to count huge amounts of data, which is difficult to avoid with erroneous and unusable data; on the other hand, the automatic data acquisition equipment on the component side is not completely put into use, and the error data is difficult to completely record. The wrong device reliability parameters are mixed in a large number of device reliability parameters, which require identification and correction.

Currently, research on the inverse problem of system reliability is mainly focused on a power distribution system, and no research related to the power generation and transmission system exists. Therefore, how to identify and correct the error reliability parameters in the power transmission and generation system to ensure the reliability and safety of the future power system is a technical problem to be solved urgently by those skilled in the art.

With the increasing popularity of intelligent detection devices (such as PMU and smart meters), the reliability index of the whole power system or node is easier to obtain, and the element reliability parameter can be inverted from the reliability index of the system (or node). Based on this, the reliability inverse problem technique is produced.

Disclosure of Invention

In order to overcome the defects in the prior art, the application provides a method for identifying and correcting the error reliability parameters of the elements of the power generation and transmission system.

In order to achieve the above purpose, the invention adopts the following technical scheme:

a method for identifying and correcting error reliability parameters of power generation and transmission system elements comprises the following steps:

step 1: identifying and correcting error reliability parameters to construct a nonlinear optimization problem which uses the reliability index estimated value and the statistic value deviation to be minimized as a target function, judging whether error parameters exist in the reliability parameters of the elements of the power generation and transmission system, if so, entering the step 2, otherwise, outputting the error parameters and the correction value;

step 2: reliability parameters of the power generation and transmission system are simplified;

and step 3: calculating the sensitivity of the reliability parameters according to a multi-measure quantification method, and dividing the reliability parameters into sensitive parameters and non-sensitive parameters according to the sensitivity;

and 4, step 4: estimating parameters of the unit connected with the load nodes to obtain a suspected error unit parameter set S_GECPAnd wherein the estimates of the parameters;

and 5: based on S_GECPAnd estimating and identifying the sensitive parameters by adopting a global optimization algorithm to obtain a suspected error parameter set S_ECPAnd wherein the estimates of the parameters;

step 6: based on S_ECPAnd estimating the parameters by using a CIVR-LSE algorithm to estimate all the parameters, and optimizing a suspected error parameter set S_ECP；

And 7: according to the deviation of the parameter statistic and its estimated value, identifying S_ECPAnd correcting the error parameters by adopting an interval algorithm.

The invention further comprises the following preferred embodiments:

preferably, in step 1, a functional equation system of the reliability index with respect to the element reliability parameter is first constructed, where the functional equation system of the reliability index with respect to the element reliability parameter is:

wherein N is_CIs the number of elements, N_idThe number of known reliability indexes;

is a known system or node reliability index statistic, e₁,e₂,……,e_NidThe mapping relationship of the element reliability parameters to the reliability indexes,

is a parameter of the failure rate of the component,

is an element repair rate parameter;

then, based on the formula (1), identifying and correcting the error reliability parameters to construct a nonlinear optimization problem which takes the reliability index estimation value and the statistic value deviation as a target function in a minimization way:

s.t.x^LB≤x≤x^UB

wherein, x is the parameter vector to be solved in the formula (1); x is the number of^LBAnd x^UBRespectively, reliability parameter value ranges obtained from engineering experience.

Preferably, in step 1, reliability evaluation is performed based on the parameter statistic, and the deviation percentage γ between the reliability index evaluation value and the statistic is calculated_i，i∈S_B；

If any node i exists, let gamma_i>Threshold value gamma⁰Marking nodes of the power generation and transmission system, and turning to the step 2; otherwise, outputting error parameters and a sum correction value;

S_Bthe method comprises the steps of collecting all nodes of a power generation and transmission system;

preferably, step 2 is specifically:

if no injection power exists on a node bus between two power transmission lines connected in series, the two lines are regarded as an equivalent line;

and (4) regarding the reliability parameters of the same-model parallel transmission lines of the same-model units or the same transmission corridor installed at the same node as the same, and expressing the reliability parameters by using the same variable to be solved.

Preferably, step 3 specifically comprises:

step 3.1: reliability parameter sensitivity is quantified using three measures:

measuring one: quantifying the sensitivity to the reliability parameters based on a system reliability index measure;

measure two: measuring the sensitivity of the quantitative reliability parameter based on an objective function f (x) of the reliability index estimation value and the minimum deviation of the statistical value;

measure three: quantifying reliability parameter sensitivity based on apportionment scale measure of elements to system reliability indexes;

step 3.2: based on the reliability parameter sensitivity quantified by the three measures, clustering is carried out on the parameters, and the reliability parameters are divided into sensitive parameters and non-sensitive parameters.

Preferably, in step 3.1, a partial derivative of a system reliability index to a reliability parameter of a certain element is measured to characterize the sensitivity, the larger the partial derivative is, the more sensitive the system reliability index is to the change of the element reliability parameter, and the more the element reliability is improved, the more the whole system reliability is easily improved;

measure two-dimensional component j repair rate mu_jOr failure rate lambda_jChanging a small quantity element, calculating the increment of the related reliability index based on f (x), wherein the larger the increment of the reliability index is, the more sensitive the reliability index is to the change of the reliability parameter;

and measuring the third step that the elements distribute the system reliability indexes proportionally, wherein the unreliability of the system is borne by the fault elements, and the normally-running elements do not participate in the distribution of the system reliability indexes.

Preferably, in step 3.2, K-Means clustering is adopted, the reliability parameters are respectively clustered into two groups according to the quantization results of the three measures of sensitivity, and the final sensitive parameter set is the union of the sensitive parameter sets corresponding to the measures.

Preferably, step 4 specifically includes:

step 4.1: recording the load node set of all the connected units as S^NGSet of parameters suspected of being faulty

Setting the threshold value of the index value deviation as gamma^NGThe parameter decision threshold is beta⁰；

According to the sequence that the absolute value of the deviation between the node reliability index evaluation value and the statistic value thereof is from large to small, S is calculated^NGInner nodes are ordered and marked as vector D^NG；

D^NGIs denoted as D^NG(k) Let k equal to 1;

step 4.2: if D is^NG(k)≥α^NGThen, taking the fault rate and the repair rate of all units connected with the node DNG (k) as parameters to be solved, and solving the nonlinear optimization problem taking the reliability index estimation and the statistical value deviation minimization as the objective function by utilizing the load loss probability LOLP, the power shortage frequency LOLF and the expected power loss EENS reliability index value of the node; otherwise, turning to step 4.4;

step 4.3: the set of all the set reliability parameters on the node is recorded as

The maximum deviation d of the estimate of the parameter m from its statistical value_mComprises the following steps:

wherein,

is an estimate of m that is,

is a statistical value of m;

if d is_m≥β⁰Then record

And adding m to S_GECP；

Making k equal to k +1, and turning to step 4.2;

step 4.4: set of outputs S_GECPAnd estimates of parameters therein.

Preferably, the global optimization algorithm in step 5 is a hybrid optimization algorithm based on a PSO algorithm and a TRR, the PSO algorithm is used in an outer layer of the hybrid optimization algorithm, one particle refers to a vector of all reliability parameter values, in an inner layer, for each particle, the reliability parameter value is used as an initial value, the TRR algorithm is used to search a local optimal solution near the particle, the local solution is used to replace an original particle, the objective function value of the local solution is used as the fitness of the particle, and the algorithm marks a sensitive parameter with a large deviation between an estimated value and a statistical value as a suspected error parameter.

Preferably, step 5 specifically includes:

step 5.1: giving all reliability index statistics values and initial value ranges of all parameters;

the set of sensitive reliability parameters is recorded as S^sensiThe suspected error parameter set is

The parameter determination threshold is beta⁰；

The number of iterations of the PSO algorithm is N_{PSO_iter}Population size N_{PSO_size}；

Randomly generating an initial population, wherein a single particle is marked as x;

the iteration number k is 0;

step 5.2: for a particle x, taking x as an initial value, and solving by adopting a confidence domain reflection method to obtain a solution x 'and a fitness value f (x');

let x be x', if k be N_{PSO_iter}Turning to step 5.5;

step 5.3: updating the individual optimum and the global optimum according to the particle fitness;

step 5.4: updating the speed and position of each particle, wherein k is k +1, and turning to step 5.2;

step 5.5: outputting a global optimal solution of the nonlinear optimization problem and an objective function value thereof;

step 5.6: for the parameter m, m ∈ S_sensiCalculating its estimate x in the optimal solution_mDeviation from statistical value d_m；

If d is_m≥β⁰Adding m to the set S_ECPStoring the estimated value x of m_m；

If m is equal to S_ECP∩S_GECPThen the estimate of m needs to be corrected to x_mAnd

and stored.

Preferably, step 6 specifically includes:

step 6.1: setting an initial determination threshold value beta^initMinimum decision threshold β^minNumber of rolling calculations K^rollThe current judgment threshold value is recorded as beta;

note that the initial value of the parameter m in each LSE is

The LSE obtained this time is estimated as x_m；

If m is equal to S_ECPThen will be

Initializing the sensitive parameter estimation value obtained in the step 5;

otherwise, it orders

Let the iteration number k be 1, and β be β^init；

Step 6.2: using the whole parameters as variables, executing LSE algorithm to obtain parameter estimated value x_m；

Step 6.3: calculating the parameter deviation d_mIf d is_mNot less than beta, then

Step 6.4: if K is equal to K^rollThen, thenTurning to step 6.5;

otherwise, let β ═ β - (β)^init-β^min)/(K^roll-1), k ═ k +1, go to step 6.2;

step 6.5: output of

And x_mFor each parameter m, if d_m≥β^minAdding m to S_ECP。

Preferably, step 7 specifically includes:

step 7.1: solving a function equation set of the reliability index related to the element reliability parameter by adopting an interval algorithm;

step 7.2: in step 7.1, if solutions exist in the equation set, all node reliability indexes corresponding to the solutions are respectively calculated, and whether a certain solution exists or not is judged to meet the reliability condition, that is:

judging that a certain solution exists, and for any node i, satisfying gamma_i≤γ⁰If the error parameter identification is finished, the solution is the error parameter correction value, and go to step 7.4;

if the equation set has no solution or each solution cannot meet the reliability condition, turning to step 7.3;

step 7.3: extended set S_ECPFrom S, the parameter suspected to be correct_ECPRemoving, and returning to the step 7.1;

step 7.4: end, output set S_ECPAnd a corresponding correction value.

Preferably, step 7.1 is performed with S_ECPAnd the medium parameter is a variable to be solved, an initial value range is given, reliability indexes with the same number as the variables are selected, a function equation set of the reliability indexes related to the element reliability parameters is constructed, and an interval algorithm is adopted for solving.

Preferably, step 7.3 said expanded set S_ECPThe method comprises the following steps:

will S_ECPSorting the other parameters from large to small, and then adding the parameters and their combinations to S according to the sorting_ECPAnd performing interval check until the check is passedSet S_ECPAnd (5) expanding.

Preferably, the step 7.3 is to determine the plausible correct parameter from S_ECPThe medium exclusion specifically is:

before checking each interval, a confidence domain reflection algorithm is adopted for S_ECPPerforming least square estimation on the medium parameter, and calculating the minimum deviation c between the estimated value of the parameter m and the statistical value thereof_m：

Setting a parameter exit threshold to beta^exitIf c is_m<β^exitThen, consider m as the correct parameter and change it from S_ECPRemoving; if c is_m≥β^exitThen continue to remain at S_ECPIn (1).

The beneficial effect that this application reached:

the invention provides a method for identifying and correcting error parameters of a power transmission system reliability evaluation inverse problem oriented to reliability parameter correction, which considers the reliability evaluation inverse problem of the power transmission system and can identify and correct the error element reliability parameters by starting from the reliability indexes of the power transmission system.

Drawings

FIG. 1 is a flow chart of the method of the present invention;

FIG. 2 is a schematic diagram of a load node of a connection unit according to an embodiment of the present invention;

FIG. 3 is a flowchart of an initial value-varying rolling iteration algorithm according to an embodiment of the present invention;

FIG. 4 is a flowchart illustrating error parameter correction based on interval algorithm according to an embodiment of the present invention;

FIG. 5 is a flowchart illustrating the error reliability parameter identification and correction according to an embodiment of the present invention;

FIG. 6 is a diagram illustrating the variation of the estimation of some parameters during CIVR-LSE iteration according to an embodiment of the present invention.

Detailed Description

The present application is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present application is not limited thereby.

The invention relates to a method for identifying and correcting error reliability parameters of elements of a power generation and transmission system, which checks or obtains the reliability parameters of the elements by starting from known reliability indexes, belongs to the reliability inverse problem, and the related concept of the reliability inverse problem is shown in a table 1:

TABLE 1 concept relating to inverse problem of reliability

The invention can use the known node/system reliability index statistics and the function of the reliability index on the element reliability parameter (CRP) to search and correct the error reliability parameter, e₁,e₂,……,e_NidThe general expression of the mapping relation from the element reliability parameters to each reliability index is as follows:

is a parameter of the failure rate of the component,

is a component repair rate parameter.

In the formula (1), the reliability parameter of each element includes two parameters of a failure rate and a repair rate. Thus, in the presence of N_CIn a system of individual elements, there is 2N_CA reliability parameter. Suppose 2N_CAmong the reliability parameters is N_EAn error parameter and N_E<N_idI.e. the number of unknowns is less than the number of equations. It is noted that, the equation set (1) can be converted into a nonlinear optimization problem which minimizes the reliability index estimation value and the statistical value deviation as an objective function, and the optimization problem is solved under the constraint of the actual reliability parameter value range of the engineering, so that the estimation value of the reliability parameter can be obtained. Suspected erroneous parameters may then be identified and corrected by comparing the parameter estimates to their statistical values. The compact form of the specific optimization problem is:

s.t.x^LB≤x≤x^UB

wherein, x is the parameter vector to be solved in the formula (1); x is the number of^LBAnd x^UBRespectively, reliability parameter value ranges obtained from engineering experience, e_i(x) Is a vector (e)₁,e₂,……,e_Nid)'. Theoretically, the overall reliability parameter truth value constitutes the global optimal solution of the formula.

As can be seen from equation (2), the reliability index is a multidimensional, nonlinear function with respect to the reliability parameter, and the degree of dimensionality and nonlinearity thereof increases as the number of elements increases. The number of variables to be solved in the optimization problem (2) is recorded as N_xThen N is_x＝2N_C. In general, N_xMuch greater than N_id. Taking IEEE-RTS as an example, N_id＝54，N_x140. Thus, in theory, equation (2) may have multiple globally optimal solutions.

Due to the fact that variable dimensionality and nonlinearity degree are too high, a local optimal solution is obtained no matter a deterministic mathematical programming algorithm (such as an interior point method, a confidence domain reflection method and the like) or an intelligent optimization algorithm is adopted to directly solve a formula. In a locally optimal solution, both the wrong and the correct parameters deviate from their true values, which makes accurate correction of the parameters difficult. Although the interval optimization algorithm can theoretically obtain a global optimal solution, the interval search times are exponentially increased along with the variable dimension, so that the efficiency is extremely low in solving the equation.

In order to obtain a true value of the parameter, when the formula (2) is solved, the parameter estimation is divided into two steps of sensitive parameter estimation and overall parameter estimation, good estimation values of the sensitive parameters are obtained firstly and are used as initial values to participate in the subsequent overall parameter estimation. This is because when the initial value error of the sensitive parameter is large, the accuracy of the overall parameter estimation is greatly affected. In the overall parameter estimation, the estimation result is corrected by using the parameter statistical value. When the deviation between the parameter estimation value and the statistic value is small, the statistic value is still used as the initial value of the subsequent parameter estimation; only if the estimated value of a parameter deviates significantly from its statistical value will it be identified as a wrong parameter and the estimated value will replace its statistical value as the initial value for the subsequent parameter estimation. This is because the vector of true parameters is in the vicinity of the vector of statistical parameter values in the search space, assuming that only a small number of parameter errors are present. By searching around the parameter statistic vector, the above correction strategy can improve the possibility that the parameter estimation converges to the true value, and prevent the estimation from departing from the true value, which will be developed in detail below.

Since the error reliability parameter correction of a power generation and transmission system is a high-dimensional and strong nonlinear problem, it is difficult to find a method capable of identifying all error parameters and accurately correcting the error parameters. A good parameter correction result should not only enable its corresponding reliability index evaluation value to accurately fit the reliability index statistical value, but also be sufficiently close to the true parameter value. Further, from the aspects of errors between the corrected values and the true values, and the like, the invention provides a plurality of indexes for evaluating the corrected results, which are specifically shown in table 2.

TABLE 2 index of the effect of parameter correction

As shown in fig. 1, in practice, the present invention comprises the following steps:

step 1: performing reliability evaluation based on the parameter statistic value, judging whether error parameters exist in the reliability parameters of the elements of the power transmission system, if so, entering the step 2, otherwise, outputting the error parameters and the correction value;

in specific implementation, all the node sets are marked as S_BSet of suspected error parameters

Setting a threshold value gamma⁰。

Reliability evaluation is carried out based on the parameter statistic value, and the deviation percentage gamma of the reliability index evaluation value and the statistic value is calculated_i，i∈S_B(ii) a For example, the reliability index evaluation value is f_eThe statistical value is f_sPercent deviation is

step 2: the method for simplifying the reliability parameters of the power generation and transmission system specifically comprises the following steps:

And step 3: calculating the sensitivity of the reliability parameters according to a multi-measure quantification method, and dividing the reliability parameters into sensitive parameters and non-sensitive parameters according to the sensitivity, wherein the method specifically comprises the following steps:

step 3.1: reliability parameter sensitivity is quantified using three measures:

further, for a power transmission and generation system, common reliability evaluation indexes comprise system LOLP, LOLF and EENS indexes, and other indexes can be obtained through deduction of the three indexes; common component reliability parameters include failure rate λ, repair rate μ, and component availability rate a, among others. The system reliability index is a function of the component reliability parameter.

The system is provided with n elements, wherein the engineering interesting elements are m, and the availability rates are A respectively₁,A₂,…,A_mThen the relationship between the system LOLP, LOLF and EENS metrics and the component availability is expressed as:

LOLP＝f₁(A₁,A₂,…,A_m)

LOLF＝f₂(A₁,A₂,…,A_m)

EENS＝f₃(A₁,A₂,…,A_m)

wherein f is₁,f₂,…,f₃Is about A₁,A₂,…,A_mIs a non-linear function of (a).

The measure is characterized by the partial derivative of the system reliability index on the reliability parameter of a certain element, for example, the sensitivity of the system reliability indexes LOLP, LOLF and EENS on the reliability parameter of element l is:

each component has two states of normal and fault, and the engineering interesting component has 2^mAnd (4) a combined state.

Any one of the combination states j can be further divided into working element state sets

And failed element state set

I (l) is an indicator variable satisfying:

the larger the partial derivative is, the more sensitive the system reliability index is to the change of the element reliability parameter, and the more easily the whole system reliability is improved by improving the element reliability.

Measure two: measure the sensitivity of the quantitative reliability parameter based on an objective function f (x) of equation (2);

the second measure is calculated by the formula:

wherein j is 1,2, … …, N_C。

The sensitivity based on the above formula is a partial derivative of the function to the independent variable at a certain point, and cannot reflect the function change condition in a large range of the independent variable. In addition, the values of the parameters under investigation may themselves be incorrect.

Therefore, the classification of the parameters may be misleading based only on the partial derivatives of the current values of the parameters.In response to this deficiency, the present invention measures dual perturbation methods to characterize sensitivity. In order to analyze the influence of the element reliability parameter change value on the system reliability index, the repair rate mu of the element j is determined_jOr failure rate lambda_jChanging a small quantity epsilon, and calculating the increment size of the related reliability index, wherein the increment size can be defined as:

wherein ε represents the relative change of an argument and is generally 0.1. The greater the reliability index increment, the more sensitive the reliability index is to changes in the reliability parameter.

The first measure of sensitivity may also be re-quantified in a similar way.

The first measure of the sensitivity reflects the degree of influence of the parameter on the reliability index of the system/node;

and the second measure reflects the influence of the parameter on the deviation of the reliability index evaluation value and the statistical value. In the study it was found that a parameter that is incorrectly valued generally has no significant effect on its first sensitivity and may cause a large change in its second sensitivity. This is because the second measure is more susceptible to parameter value than the first measure. For example, as the parameter values approach the local optimal solution (the point where the gradient is zero), the second sensitivity of each parameter will gradually approach 0, while the first sensitivity will not change significantly. The two sensitivities can complement each other, and the effect of error parameter identification is improved.

for the third measure of sensitivity, the apportionment ratio of the element to the reliability index can reflect the influence degree of the element reliability level on the reliability index, and is positively correlated with the first measure.

Measure three, the element distributes the apportionment of the system reliability index according to the proportion, wherein, the unreliability of the system is born by the fault element, and the normal operation element does not participate in the apportionment of the system reliability index;

further, illustrated as a two-element system, a failure event U of the system is defined by element X₁And X₂Caused by a fault. Correlation reliability index f (X)₁,X₂,…,X_n) It can be broken down into 3 parts:

the first part being associated only with element X₁Relevant part f₁(X₁)；

Second partial element X₂Related event f₂(X₂)；

The third part is and divides by X₁And X₂Relating to other elements than

Therefore, the temperature of the molten metal is controlled,

let f (k → 1) be the faulty element X₁Corresponding sensitivity index, f (k → 1) is the failed component X₂Corresponding sensitivity index. Then, define the sensitivity index as

Step 3.2: after the sensitivity calculation is completed, clustering is carried out on the parameters based on the reliability parameter sensitivity quantified by the three measures, and the reliability parameters are divided into sensitive parameters and non-sensitive parameters.

The clustering algorithm can be selected from K-Means clustering, fuzzy C-Means clustering and other algorithms. Because the clustering task is simpler, the K-Means clustering method is adopted. According to the three measures of sensitivity, the reliability parameters are respectively grouped into two groups. The final sensitivity parameter set is the union of the sensitivity parameter sets corresponding to the respective measures.

And 4, step 4: estimating parameters of the unit connected with the load node to obtain suspected error unit parametersSet of numbers S_GECPAnd wherein the estimates of the parameters;

and 4-5, calculating the estimation value of the sensitive parameter by adopting a global optimization algorithm so as to provide a good initial value for the subsequent overall parameter estimation.

Note that when the near load shedding policy is adopted after an accident occurs in the system, if a machine group connected to a load node fails, the load of the node is first shed. In other words, the reliability index of the node is greatly influenced by the reliability of the connected unit. Therefore, before the sensitive parameters are integrally estimated, the reliability parameters of the connected units can be estimated by using the reliability indexes of the load nodes.

The concrete steps are sequentially explained as follows:

(1) parameter estimation of unit connected with load node under consideration of nearby load shedding strategy

For a certain load node connected with a unit, if the reliability index evaluation value and the statistical value are significantly different, the probability that the reliability parameter of the unit connected with the node is wrong is high. Therefore, the reliability index of the load node can be used for estimating the reliability parameter of the unit connected with the node.

The parameter estimation (algorithm 1) of the unit connected with the load node comprises the following steps:

step 4.1: the set of load nodes of all the connected units is marked as S^NGSet of parameters suspected of being faulty

The threshold value of the index value deviation is recorded as gamma^NGAnd the parameter decision threshold is recorded as beta⁰. According to the sequence that the absolute value of the deviation between the node reliability index evaluation value and the statistic value thereof is from large to small, S is calculated^NGInner nodes are ordered and marked as vector D^NG。D^NGIs denoted as D^NG(k) In that respect Let k equal to 1. FIG. 2 is a schematic view of a load node of a connecting setG1-G4 are machine sets.

Step 4.2: if D is^NG(k)≥α^NGThen with node D^NG(k) Establishing an optimization problem as (2) and solving by using LOLP, LOLF and EENS index values of the node, wherein the fault rates and repair rates of all connected units are parameters to be solved; otherwise, go to step 4.4.

wherein,

is the estimate of m in the present algorithm,

is a statistical value of m. If d is_m≥β⁰Then record

And adding m to S_GECP. Let k be k +1, go to step 4.2.

Step 4.4: the algorithm is finished and the set S is output_GECPAnd estimates of parameters therein.

It should be noted that: because the reliability index information of the local nodes is utilized, compared with the sensitive parameter estimation and identification method based on the global optimization algorithm provided in the following step (2), the error of the parameter estimation at this point may be larger. Therefore, the suspected error parameter should be determined by comprehensively considering the subsequently obtained parameter estimation values.

Further, with respect to S obtained here_GECPShould a certain parameter in (2) be not judged as an error parameter in the sensitive parameter estimation and identification based on the global optimization algorithm proposed later, it should be S_GECPAnd (4) excluding.

(2) Sensitive parameter estimation and identification based on global optimization algorithm

The method utilizes all reliability index statistical values to establish an optimization problem as shown in a formula (2) and estimate sensitive parameters.

Although only estimates of sensitive parameters are of interest here, the optimization variables need to be global reliability parameters, not just sensitive parameters. This is because, if the optimized variables only contain sensitive parameters, the wrong parameters in the non-sensitive parameters may interfere with the sensitive parameter estimation, resulting in a large error in the sensitive parameter estimation.

The local search precision of the conventional mathematical programming algorithm is high, but the quality of the solution depends on the initial value, and the solution can be converged only when the initial value is close to the true value. The swarm intelligence optimization algorithm can search in the whole space and has certain global convergence characteristics, however, the local search capability is weak, and the calculation precision cannot meet the identification requirement.

The method adopts a mixed optimization algorithm based on Particle Swarm Optimization (PSO) algorithm and trust domain reflectometry (TRR) algorithm to obtain the estimation value of the sensitive parameter. The algorithm fully combines the global search characteristic of the group intelligent algorithm and the local search capability of the conventional algorithm, and can give approximate values of sensitive parameters.

The PSO algorithm is adopted in the outer layer of the hybrid optimization algorithm, and one particle refers to one vector of all reliability parameter values. In the inner layer, for each particle, the reliability parameter value is used as an initial value, a TRR algorithm is adopted to search a local optimal solution near the particle, the original particle is replaced by the local solution, and the objective function value of the local solution is used as the fitness of the particle. And after the algorithm is completed, marking the sensitive parameters with larger deviation between the estimated value and the statistical value as suspected error parameters.

The sensitive parameter initial value estimation algorithm (algorithm 2) comprises the following steps:

step 5.1: and giving all reliability index statistics values and initial value ranges of all parameters.

The set of sensitive reliability parameters is denoted S^sensiSet of suspected error parameters

The parameter determination threshold is denoted as β⁰。

Number of iterations N of the PSO algorithm_{PSO_iter}Population size N_{PSO_size}。

An initial population was randomly generated, with individual particles noted as x.

The iteration number k is 0.

Step 5.2: and for the particle x, taking x as an initial value, and solving by adopting a confidence domain reflection method to obtain a solution x 'and a fitness value f (x').

Let x be x'. If k is equal to N_{PSO_iter}Go to step 5.5.

Step 5.3: and updating the individual optimum and the global optimum according to the particle fitness.

Step 5.4: and updating the speed and the position of each particle, wherein k is k + 1. Go to step 5.2.

Step 5.5: and outputting the global optimal solution of the formula (2) and the objective function value thereof.

Step 5.6: for parameter m (m ∈ S)_sensi) Calculating its estimate x in the optimal solution using equation (5)_mDeviation from statistical value d_m。

and stored.

S derived from the above algorithm_ECPThe estimated value of the parameter is used as the initial value of the subsequent estimation algorithm.

Step 6: based on S_ECPAnd wherein the parameters are estimated by CIVR-LSEThe algorithm estimates all parameters and optimizes a suspected error parameter set S_ECP；

Step 6, starting from the estimation of the error sensitive parameters, estimating all reliability parameters, and providing a rolling estimation algorithm for correcting the initial value solution of the rolling estimation algorithm by adopting a parameter statistic value to estimate all parameters, wherein the algorithm can improve the accuracy of parameter estimation, and specifically comprises the following steps:

by using the parameter statistics, a Changing initial value rolling iterative least square estimation (CIVR-LSE) algorithm is provided.

The algorithm performs multiple rounds of LSE by scrolling in an attempt to obtain a solution that is sufficiently close to the true value. The suspected erroneous parameters are then identified by comparing the deviations of the parameter statistics and estimates.

Since only a small number of error parameters exist in the system, the parameter statistic vector is closer to the true value vector in the search space of the optimization problem. Thus, in parameter estimation, the parameter statistics should be "trusted". And only when the deviation between the parameter estimated value and the statistical value is large, the parameter estimated value is judged as a suspected error parameter, the estimated value is used as an initial value of the parameter in the next round of LSE, and otherwise, the statistical value is still used as the initial value. In this way, the algorithm both corrects suspected erroneous parameters and ensures that the true values of the parameters are searched for near their statistical values. As the number of scrolls increases, the solution for CIVR-LSE gradually approaches the true value.

Maximum deviation d of the estimated value of the parameter m from its statistical value in the CIVR-LSE process_mComprises the following steps:

wherein S is^paraIs the global reliability parameter set.

For equation (6), the decision threshold is β, if d_mBeta or more, the parameter m is judged to be wrong, and the estimation value x is used_mAs an initial value for the lower round LSE.

The flow chart of CIVR-LSE algorithm (Algorithm 3) is shown in FIG. 3, and the steps are as follows:

step 6.1: setting an initial determination threshold value beta^initMinimum decision threshold β^minNumber of rolling calculations K^rollThe current decision threshold is denoted as β. The initial value of the parameter m in each LSE is recorded as

The LSE obtained this time is estimated as x_m. If m is equal to S_ECPThen will be

Initializing the sensitive parameter estimation value obtained in the step 5; otherwise, it orders

Let the iteration number k be 1, and β be β^init。

Step 6.2: taking the whole parameters as variables, executing LSE to obtain parameter estimation value x_m。

Step 6.3: calculating the parameter deviation d according to equation (5)_m. If d is_mNot less than beta, then

Step 6.4: if K is equal to K^rollTurning to step 6.5; otherwise, let β ═ β - (β)^init-β^min)/(K^roll-1), k ═ k + 1. Go to step 6.2.

Step 6.5: output of

And x_m. For each parameter m, if d_m≥β^minAdding m to S_ECP。

Research shows that after parameter estimation in the steps 4 and 5Most error parameters can be identified, which meets the actual requirement of engineering for improving the accuracy of parameters, but some error parameters with less influence on the reliability index may be omitted. In order to form a closed-loop method system for identifying and correcting error parameters, the invention provides a parameter identification completion degree inspection method based on an interval algorithm, and whether all error parameters are identified is inspected. If the checking is passed, the solution of the interval algorithm is the correction value of the error parameter; otherwise, enumerating parameters or their combination according to the sequence of sensitivity of parameters from large to small, and adding them into the set S_ECPThe test is performed until the test is passed.

Parameter identification completion degree detection and correction based on interval algorithm

Based on the analysis of the steps 1 to 6, a suspected error parameter set S is obtained_ECP。

However, S cannot be ascertained_ECPIf all error parameters are included, a check is required. With S_ECPAnd (3) establishing an LSE problem shown in the formula (2) by using all reliability indexes and solving the LSE problem by using the medium error parameters as variables.

If S_ECPIf all error parameters are included, the objective function value approaches 0; otherwise, the function value is not 0. The premise of ensuring the validity of the checking thought is that a global optimal solution can be obtained by a solving algorithm. However, the conventional non-linear optimization algorithm cannot judge whether the result is the optimal solution. Therefore, it is impossible to determine S by whether the objective function value is close to 0 or not_ECPWhether all error parameters are included.

The interval algorithm can judge the existence of the solution in the interval to be researched, and can ensure that the global optimum is obtained. Because the interval optimization algorithm has low calculation efficiency, an interval iteration algorithm aiming at the equation set is adopted.

With S_ECPAnd (3) setting an initial value range of the medium parameter as a variable to be solved, selecting the reliability indexes with the number equal to that of the variable, constructing an equation set as the formula (1) and solving the equation set.

And if the interval algorithm has a solution, performing reliability evaluation based on the solution. If all system/node reliability indicators match their statistical values, it can be determined that all error parameters have been identified, and the solution is the correction value for the error parameter. Otherwise, it can be determined that there are still erroneous parameters missing.

At this time, the subsequent second and third methods are needed to search for the missing error parameters.

Missing error parameter search based on sensitivity ranking

Enumeration of S_ECPOther parameters or combinations thereof, respectively, to S_ECPAnd performing interval check until the check is passed. When the number of the parameters is too large, the key for improving the search efficiency is how to determine the priority order of the parameters participating in the interval check.

Based on the sensitivity thought, according to the system reliability index change caused by the deviation of the parameter estimated value and the statistic value thereof, S is calculated_ECPThe other parameters are sorted from large to small. Then, parameters and their combinations are added to S in this order_ECP。

Method for eliminating correct reliability parameter

Set S since some correct parameters will be misjudged as incorrect parameters_ECPMay contain several correct parameters. If S_ECPToo many parameters will affect the calculation efficiency of the interval algorithm, so S should be reduced as much as possible_ECPThe number of parameters in (1).

Setting a parameter exit threshold to beta^exitIf c is_m<β^exitThen m can be considered as the correct parameter and be taken from S_ECPRemoving; if c is_m≥β^exitThen continue to remain at S_ECPIn (1). Threshold beta^exitShould be small to avoid excluding the wrong parameters.

Fig. 4 shows a simple error parameter correction flow based on the interval algorithm.

Namely, step 7 specifically includes:

step 7.1: according to the method of the first step, an interval algorithm is adopted to solve an equation set as the formula (1);

step 7.3: according to the method of (2) above, the set S is expanded_ECP(ii) a According to the third method, the suspected correct parameters are selected from S_ECPRemoving, and returning to the step 7.1;

step 7.4: end, output set S_ECPAnd a corresponding correction value.

Examples

The following describes the technical solution of the present invention with reference to fig. 5, and verifies the method for identifying and correcting the error value of the reliability parameter of the present invention by using modified RBTS, IEEE-RTS, and 91-node power system (hereinafter abbreviated as CS system).

The general flow of the method for identifying and correcting the error reliability parameters is shown in FIG. 5, which comprises the following steps:

and 4, step 4:estimating parameters of the unit connected with the load nodes to obtain a suspected error unit parameter set S_GECPAnd wherein the estimates of the parameters;

In the raw data of the test system, the reliability parameters of the same type of elements are the same and independent of each other. For example, there are four types of power transmission branches in an IEEE-RTS transmission system, and the original reliability parameters are shown in table 3.

TABLE 3 Primary reliability parameters for IEEE-RTS System legs

Note: the repair time of two 138kV cables in IEEE-RTS is different and is respectively 16h and 35 h. The branch numbers corresponding to the branch types are shown in table 4.

TABLE 4 IEEE-RTS system Branch and Branch type correspondence

Branch type	Corresponding branch number
		138kV cable

		1,10
138kV overhead line			2-6,8-13
	138/230kV transformer	7,14-17
230kV overhead line			18-38

In order to reflect the difference of the reliability levels of the same type of elements of the power generation and transmission system, the invention sets reliability parameters with different values for the same type of power transmission elements. For example, for 138kV overhead lines of an IEEE-RTS system, a group of random numbers which are subjected to normal distribution are generated as the true value of the unit fault rate of each line of the type by taking the original value of the unit-length fault rate of the overhead line as the mean value and 1/10 of the unit-length fault rate as the variance.

In order to simulate error parameters in engineering practice, error parameter embodiments are set by modifying the reliability parameter values of part of elements of the test system in advance. And taking the value of each parameter before modification as a true value, and taking the value after modification as a parameter statistic value. Before parameter modification, the system and node reliability indexes are calculated based on the true values and serve as the statistical values of the reliability indexes in the calculation examples. And after the parameters are modified, reliability evaluation is carried out based on the parameter statistic value or the estimated value in the parameter identification and correction process, and the obtained reliability index is used as the reliability index evaluation value.

In connection with the actual situation of component reliability in the power system, step 2 of the algorithm of fig. 5 makes the following simplifying assumption:

a) the reliability parameters of the units of the same type installed on the same node are the same;

b) the reliability parameters of the same-type parallel transmission lines connected between the same two nodes are the same.

The component reliability parameters are the same, meaning that their failure rates (or repair rates) share a variable to be solved in the parameter correction problem. These two assumptions reduce the likelihood that multiple solutions exist.

Assume that a parameter whose error between the parameter statistic and its true value exceeds 10% is taken as an error parameter. Determination threshold value γ of reliability index ⁰1%, decision threshold β in initial value estimation of sensitive parameter⁰Initial decision threshold β in CIVR-LSE, 20%^init20%, minimum decision threshold β^min10%. Correct parameter decision threshold beta ^exit1% of the total weight. The value range of each parameter can be obtained by engineering experience, and the value range is assumed to be 0.5-2 times of the true value of the parameter.

Reliability parameter identification and correction results

(1) Testing of RBTS and IEEE-RTS systems

Error parameter examples Case 1-Case 2 and Case 3-Case 7 are set here for RBTS and IEEE-RTS, respectively, to verify the validity of the proposed method. Error parameters for the various embodiments are shown in Table 5

TABLE 5 component reliability parameter error values for RBTS and IEEE-RTS

Note: lambda [ alpha ]_G1-20Which refers to the failure rate of all 20MW units installed on node 1, and L7 represents leg 7.

The results of the error reliability parameter identification of cases 1 to 7 are shown in Table 6.

TABLE 6 ERROR RELIABILITY PARAMETER IDENTIFICATION RESULTS FOR THE EMBODIMENTS OF RBTS AND IEEE-RTS SYSTEMS

Comparing the recognition results of the embodiments with the set error parameters, it can be seen that the recognition results of each embodiment all include all the error reliability parameters. Table 7 lists the reliability parameter correction values for some of the examples.

TABLE 7 RBTS and IEEE-RTS System partial embodiment reliability parameter correction results

It can be seen from the results in table 7 that the interval algorithm can accurately correct the error values of the reliability parameters, all the correction intervals of the error parameters include the true values of the parameters, and the maximum error between the interval boundary and the true values does not exceed 0.1%.

(2) Test of 91-node system (CS System)

Here the validity of the proposed method is verified using the 91-node CS system, the error parameters of Case 8 are as follows:

TABLE 8 element reliability parameter error values for CS systems

Case 8 includes 10 error parameters, involving 23 components, which account for 9.7% of the total number of system components. The result of the false reliability parameter identification of Case 8 is as follows:

Case 8：λ_G52-45、λ_L1、λ_L40、λ_L139、μ_G1-550、μ_G23-100、μ_G33-175、μ_G66-200and mu_G42-330。

Comparing the Case 8 recognition result with the error parameters set in table 8, it can be seen that: the method provided by the invention identifies 9 parameters, and omits the parameter lambda_L3All sensitive error parameters are included, although not all error parameters are covered. The reliability parameter correction results are shown in the following table.

TABLE 9 reliability parameter correction results for CS systems

As can be seen from the results in Table 9, although the Case 8 identification result does not contain all error parameters, the identified error parameters can be verified by the interval algorithm, and the corrected value is very close to the true value. The result indicates that the missing error parameters do not affect the accuracy of the correction of the other parameters, which may be because the missing error parameters are non-sensitive parameters, which have very little effect on the reliability index.

Reliability parameter identification and correction process analysis

Taking Case 7 of the IEEE-RTS system as an example, the error reliability parameter identification and correction process is shown in steps, including parameter estimation of a set connected with the load node, initial value estimation of a sensitive parameter, overall parameter estimation and identification, and parameter correction based on an interval algorithm.

Step 1-step 3:

see Table 5 for Case 7 data, this example includes 6 error parameters, involving 11 components, which account for 15.7% of the total system components, and involving 14.7% of the total installed capacity.

Before identification and correction, the reliability parameters are first clustered.

The reliability parameters are divided into sensitive parameters and non-sensitive parameters by comprehensively considering three measures of sensitivity. The set of sensitivity parameters is as follows:

S^sensi＝{λ_G18-400，λ_G21-400，λ_G23-350，λ_G13-197，λ_G23-155，λ_G1-76，λ_G2-76，λ_G16-155，λ_G22-50，λ_G1-20，μ_G18-400，μ_G21-400，μ_G23-350，μ_G13-197，μ_G23-155，μ_G15-12}。

and according to the clustering result, sensitive parameters of the IEEE-RTS system are set fault rate and repair rate. It should be noted that a failure of any unit directly reduces power supply redundancy, which in turn leads to reduced system reliability. Therefore, if the total number of the units of the system is less, the reliability parameters of all the units can be added into the sensitive parameter set.

And 4, step 4: parameter estimation of unit connected with load node

The corresponding algorithm for parameter estimation of the unit connected with the load node is algorithm 1. And sequencing the load nodes according to the sequence that the deviation of the reliability index evaluation value and the statistic value thereof is from large to small. As only one rough estimation is carried out on the unit parameters, the step mainly focuses on the load nodes with larger reliability index deviation, and therefore, a deviation threshold value gamma is set^NG20% by weight. The following table gives the index value deviation greater than the threshold value gamma^NGThe load node of (1).

TABLE 10 deviation of reliability index evaluation values from statistical values for partial load nodes

Node numbering	18	1	3	19	6
						Deviation of	32.41％	27.27％	25.96％	21.70％	21.24％

The index value deviation of the node 18 is the largest, and the node is connected with a 400MW unit. The LOLP, LOLF and EENS indexes of the node 18 are used to obtain the estimation values of the fault rate and the repair rate of the unit, and the results are shown in Table 11.

Table 11 estimation of reliability parameters of units connected to node 18

Setting a decision threshold beta⁰20% deviation greater than beta⁰The parameter(s) of (1) is marked as a suspected erroneous parameter and its estimate is recorded. As can be seen from Table 11, λ_G18-400Is 46.00%, should be flagged as a suspected error parameter and its estimate 7.751 recorded. Parameter mu_G18-400Does not exceed the threshold value beta⁰The statistical value 58.4 is still used as its initial value in the subsequent parameter estimation. At this time, the suspected error parameter set S_GECP＝{λ_G18-400}. Next, the second node 1 in the index value deviation degree ranking is estimated, and the parameters of the connected set are estimated.

Table 12 estimation of reliability parameters of unit connected to node 1

Parameter(s)	Statistical value	Valuation	Deviation of estimated and statistical values
				λ_G1-20	12.978	20.119	55.02％
λ_G1-76	4.469	6.0335	35.01％
				μ_G1-20	175.2	187.97	7.29％
μ_G1-76	219	188.03	16.47％

As can be seen from Table 12, λ_G1-20And λ_G1-76Should be marked as suspected error parameters and their estimates recorded. At this time, S_GECP＝{λ_G18-400,λ_G1-20,λ_G1-76}. The subsequent nodes in table 11, i.e.

nodes

3, 19 and 6, are not connected to the crew and therefore the parameter estimation here ends. Otherwise, as stated in (1) of step 4, if the current S_GECPIf a parameter in (1) is not determined to be an error parameter in the subsequent steps, the parameter should be excluded.

And 5: initial estimation and identification of sensitive reliability parameters

The initial value of the sensitive reliability parameter is estimated and identified corresponding to algorithm 2. Setting a parameter decision threshold beta⁰The population size of the PSO algorithm is 30% with a maximum number of iterations of 200, 20%. After the algorithm is executed, the suspected error is identifiedThe parameters and their estimates are shown in table 13.

TABLE 13 list of suspected error parameters in sensitive parameters

Parameter(s)	Statistical value	Valuation	Deviation of estimated and statistical values
				λ_G1-20	12.978	21.397	60.28％
λ_G18-400	5.309	8.950	64.84％
				μ_G15-12	112.308	159.949	55.60％

At this time, S_ECP＝{λ_G18-400,λ_G1-20,μ_G15-12}. According to the algorithm flow step 6, three reliability indexes of the LOLP, EENS and LOLF of the system are selected, and an interval algorithm is adopted to solve S_ECPThe parameter (1). However, the interval algorithm has no solution, which means that not all are currently recognizedAn error parameter.

Step 6: overall reliability parameter estimation and identification

The global reliability parameter estimation and identification corresponds to algorithm 3. And using the estimated value of the suspected error parameter obtained in the previous step as an initial value to participate in initial value rolling least square estimation (CIVR-LSE). Setting an initial determination threshold beta^init20%, minimum decision threshold β^min10%, number of rolling iterations K^roll15. When the first LSE of CIVR-LSE is completed, there are 11 estimated and statistical values with a deviation greater than beta^min. The estimated changes of these 11 parameters during the CIVR-LSE scroll iteration are shown in FIG. 6. The vertical axis represents the percentage of parameter true values, statistical values and estimates relative to the respective true values, and the horizontal axis is the number of 11 parameters. As can be seen from FIG. 6, the estimates of these 11 parameters gradually regress to true values during the CIVR-LSE iteration. This variation illustrates that the solution for CIVR-LSE will gradually approach the true value.

After the CIVR-LSE is finished, the following suspected error parameter set S is obtained_ECP：

S_ECP＝{λ_G1-20,λ_G18-400,λ_L7,μ_G15-12,μ_L11}

By comparison with the error parameter list of Table 5, except for the parameter λ_L23At present, S_ECPAll error parameters have just been included. Solving the set S by adopting an interval algorithm according to the step 6 of the algorithm flow_ECPMedium parameter, algorithm has no solution. This means that not all error parameters have been identified, and therefore, the search for missing error parameters needs to be continued.

And 7: error reliability parameter correction based on interval algorithm

According to the algorithm flow in fig. 5, step 7.1 to step 7.3 are repeated. By enumeration and verification, when the parameter λ_L23Adding S_ECPIn the middle time, the interval algorithm has a solution and can pass the verification of all the reliability indexes, all error parameters can be recognized, and the identification and correction process is finished. Set of final outputs S_ECPComprises the following steps: s_ECP＝{λ_G1-20,λ_G18-400,λ_L7,λ_L23,μ_G15-12,μ_L11}。

It can be seen that S_ECPJust containing all error parameters. In this case, the solution of the interval algorithm is the correction value of the error parameter. The parameter correction values of this example (Case 7) are shown in Table 14.

Comparison of the method of the invention with conventional algorithms

For comparison, the parameter identification results of the above four steps are shown in Table 14, and the first row of the table lists 6 actual error parameters.

Table 14 identification of actual error parameters from each step

Note: "check" indicates that the parameter is included in the suspected error parameter set when the corresponding step is executed, and "x" indicates that the parameter is not included.

As can be seen from table 14, only two error parameters are identified when

steps

4 and 5 are completed; when step 6 is completed, only one parameter is left without being identified; after the algorithm execution is completed, all 6 parameters are identified. As can be seen from the identification process of this embodiment, the 4 main steps set by the proposed method are organically combined, and all the error parameters are effectively identified layer by layer.

In order to prove the effectiveness of the proposed method, 4 sections of evaluation indexes are adopted, and the Case 7 is taken as an example, and the correction effect of the proposed method and a conventional algorithm is compared. The following four algorithms were tested:

the method A comprises the following steps: starting from the parameter statistics value, directly executing LSE;

the method B comprises the following steps: starting from the parameter statistic value, skipping the initial value estimation of the sensitive parameter, and executing CIVR-LSE;

the method C comprises the following steps: after the initial value of the sensitive parameter is estimated, executing single LSE; without performing CIVR-LSE;

method D (inventive method): the initial values of the sensitive parameters are estimated, and then CIVR-LSE is executed.

It should be noted that the method D does not include a correction process based on an interval algorithm. The comparison of method B with method D can reveal the effect of the initial value estimation of the sensitive parameter, and the comparison of method C with method D can reveal the effect of the CIVR-LSE algorithm.

The identification results of each method are shown in Table 14, where the first row of the table is the 6 actual error parameters.

TABLE 15 identification of actual error parameters by each method

Note: "check" indicates that the parameter is included in the suspected error parameter set obtained by the corresponding method, and "x" indicates that the parameter is not included.

As can be seen from table 15, for 6 error parameters, method a identified 4 of them, while the other three identified 5, which indicates that performing LSE directly is less effective. Some evaluation indexes for the identification and correction effects of each method are shown in Table 16.

TABLE 16 partial evaluation index for the identification and correction effects of each method

As can be seen from table 16: each index of method D is optimal, while method A is worst and methods B and C are in between. As can be seen from the selectivity index, although the number of error parameters successfully identified by the methods B and C is the same as that of the method D, the methods B and C misjudge a part of correct parameters as error parameters, and thus the two methods have low selectivity. The suspected error parameters identified by method D are exactly the actual error parameters. The method A misjudges more correct parameters, and the selection degree is the lowest. From the mean error index of the estimates, method D is significantly better than the other three methods, with errors of 7.4%, 25.4%, and 15.4% of the other three methods, respectively. As can be seen from the comparison of the evaluation indexes of the methods B, C and D, the lack of any one of the two steps of sensitive parameter initial value estimation and CIVR-LSE can obviously reduce the effect of error parameter identification and correction, which confirms the effectiveness of the method D, namely the method of the present invention.

Due to the omission of partial component shutdown records, errors exist in the statistical values of the reliability parameters of the partial components, and the error reliability parameters need to be identified and corrected. Under the background, the invention establishes an inverse problem model for error reliability parameter correction and provides a solving method. Since the error reliability parameters are mixed in the overall reliability parameters, the overall parameters need to be estimated. Because the estimation of the sensitive parameters is usually more accurate and the influence on the estimation precision of the whole parameters is larger, the invention provides a two-step parameter estimation strategy. Firstly, sensitive parameters are estimated by adopting a particle swarm optimization algorithm with a local search mechanism. Then, a rolling estimation is performed on the overall parameters by using a varying initial value algorithm. In the two-step parameter estimation process, error parameters are identified according to the deviation between the parameter estimation value and the statistic value; for the error parameter, its estimated value is used instead of its statistical value as the initial value of the subsequent estimation. And finally, judging whether the identification process of the error parameters is finished or not by using an interval algorithm, and correcting the parameters. Aiming at the error parameter embodiment Case 7 of IEEE-RTS, compared with three algorithms such as conventional LSE and the like, the algorithm realizes accurate identification of the error parameter and has optimal selectivity; the parameter estimation errors of the proposed algorithm are only 7.4%, 25.4% and 15.4% of the other three methods, respectively; compared with the true value of the error parameter, the error of the correction value obtained by the method does not exceed 1 percent. The method has important significance for improving the reliability and safety of a future power system.

The present applicant has described and illustrated embodiments of the present invention in detail with reference to the accompanying drawings, but it should be understood by those skilled in the art that the above embodiments are merely preferred embodiments of the present invention, and the detailed description is only for the purpose of helping the reader to better understand the spirit of the present invention, and not for limiting the scope of the present invention, and on the contrary, any improvement or modification made based on the spirit of the present invention should fall within the scope of the present invention.

Claims

1. A method for identifying and correcting error reliability parameters of power generation and transmission system elements is characterized by comprising the following steps:

the method comprises the following steps:

2. The method according to claim 1, wherein the method comprises the steps of:

in step 1, firstly, a function equation set of the reliability index related to the element reliability parameter is constructed, wherein the function equation set of the reliability index related to the element reliability parameter is as follows:

is a parameter of the failure rate of the component,

is an element repair rate parameter;

s.t.x^LB≤x≤x^UB

3. The method according to claim 1 or 2, wherein the method comprises the steps of:

in step 1, reliability evaluation is carried out based on the parameter statistic value, and reliability index evaluation is calculatedPercent deviation gamma of the estimated value from its statistical value_i，i∈S_B；

S_Bis the complete node set of the power generation and transmission system.

4. The method according to claim 3, wherein the method comprises the steps of:

the step 2 specifically comprises the following steps:

5. The method according to claim 2, wherein the method comprises the steps of:

the step 3 specifically comprises the following steps:

step 3.1: reliability parameter sensitivity is quantified using three measures:

6. The method according to claim 5, wherein the method comprises the steps of:

in step 3.1, measuring a partial derivative of a system reliability index to a reliability parameter of a certain element to characterize the sensitivity, wherein the larger the partial derivative is, the more sensitive the system reliability index is to the change of the element reliability parameter, and the more easily the reliability of the element is improved, the more easily the reliability of the whole system is improved;

7. The method according to claim 5, wherein the method comprises the steps of:

in step 3.2, K-Means clustering is adopted, reliability parameters are respectively clustered into two groups according to three measurement quantization results of sensitivity, and the final sensitive parameter set is a union set of sensitive parameter sets corresponding to all measurements.

8. The method according to claim 2, wherein the method comprises the steps of:

the step 4 specifically comprises the following steps:

According to the deviation between the node reliability index evaluation value and its statistical valueFor the sequence of values from big to small, the S is added^NGInner nodes are ordered and marked as vector D^NG；

D^NGIs denoted as D^NG(k) Let k equal to 1;

wherein,

is an estimate of m that is,

is a statistical value of m;

if d is_m≥β⁰Then record

And adding m to S_GECP；

Making k equal to k +1, and turning to step 4.2;

step 4.4: set of outputs S_GECPAnd estimates of parameters therein.

9. The method according to claim 2, wherein the method comprises the steps of:

and 5, the global optimization algorithm is a mixed optimization algorithm based on a PSO algorithm and a TRR (true-false-rate regression) algorithm, the PSO algorithm is adopted on the outer layer of the mixed optimization algorithm, one particle refers to a vector of all reliability parameter values, the reliability parameter value of each particle is used as an initial value on the inner layer, the TRR algorithm is adopted to search a local optimal solution near the particle, the local solution is used for replacing the original particle, the objective function value of the local solution is used as the fitness of the particle, and the sensitive parameter with larger deviation between the estimated value and the statistical value is marked as a suspected error parameter by the algorithm.

10. The method according to claim 9, wherein the method comprises the steps of:

the step 5 specifically comprises the following steps:

The parameter determination threshold is beta⁰；

the iteration number k is 0;

let x be x', if k be N_{PSO_iter}Turning to step 5.5;

and stored.

11. The method according to claim 2, wherein the method comprises the steps of:

the step 6 specifically comprises the following steps:

note that the initial value of the parameter m in each LSE is

The LSE obtained this time is estimated as x_m；

If m is equal to S_ECPThen will be

Initializing the sensitive parameter estimation value obtained in the step 5;

otherwise, it orders

Let the iteration number k be 1, and β be β^init；

Step 6.4: if K is equal to K^rollTurning to step 6.5;

step 6.5: output of

And x_mFor each parameter m, if d_m≥β^minAdding m to S_ECP。

12. The method according to claim 2, wherein the method comprises the steps of:

the step 7 specifically comprises the following steps:

step 7.4:end, output set S_ECPAnd a corresponding correction value.

13. The method according to claim 12, wherein the method comprises the steps of:

step 7.1 with S_ECPAnd the medium parameter is a variable to be solved, an initial value range is given, reliability indexes with the same number as the variables are selected, a function equation set of the reliability indexes related to the element reliability parameters is constructed, and an interval algorithm is adopted for solving.

14. The method according to claim 12, wherein the method comprises the steps of:

step 7.3 the extended set S_ECPThe method comprises the following steps:

will S_ECPSorting the other parameters from large to small, and then adding the parameters and their combinations to S according to the sorting_ECPAnd carrying out interval check until the check is passed, and finishing the set S_ECPAnd (5) expanding.

15. The method according to claim 12, wherein the method comprises the steps of:

step 7.3 from S_ECPThe medium exclusion specifically is: