A method and system for evaluating weak lines of a wind power system

Abstract

The application discloses a kind of weak line evaluation method and system of wind power system containing wind power, consider the system aggregate frequency response model of wind power access, further consider the quasi-steady cascading failure model of wind power ultra-short-term output correlation, uncertainty and low inertia characteristics, overcome the defects that existing quasi-steady cascading failure model is insufficient to consider the uncertainty of wind power output, cannot consider the dynamic frequency characteristics of system when disturbed, to improve the accuracy of wind power system containing wind power cascading failure simulation.The application is based on cascading failure network method and the weak line of cascading failure propagation stage is identified based on the obtained cascading failure data set, overcome the defects that existing weak line identification method can only be identified for a certain normal state, cannot consider subsequent cascading failure characteristics.Enhance the ability of wind power system containing wind power to deal with wind power output uncertainty and resist cascading failure.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of new energy power system reliability evaluation, and particularly relates to a weak line evaluation method and system for a wind power system. BACKGROUND

[0002] Cascade failure is the main cause of power system blackouts. The occurrence and propagation of power system cascading failures are closely related to the failure of some components. These components have higher criticality than other components and are the key components of the system. Therefore, identifying these key components and reinforcing them is an effective means to prevent the occurrence of cascading failures and resist the propagation of cascading failures. One of the main features of new power systems is the high proportion of wind power and other new energy sources connected to the system. The output uncertainty and low inertia and weak frequency response characteristics of wind power may change the distribution of key components of the power system and affect the characteristics of system cascading failures. It is necessary to study the weak lines of wind power systems in view of the output uncertainty and low inertia of wind power, and to improve the ability of wind power systems to cope with uncertainty and prevent and resist cascading failures.

[0003] At present, there have been many researches on weak lines of power systems. The main ideas can be roughly divided into two categories: the first category is based on power system state analysis and uses reliability theory to identify weak lines of the system from the perspective of operating state; the second category is based on complex system theory and mainly considers the topological structure and static power flow distribution of the system to identify weak lines. However, the above methods only evaluate a complete system state before the failure, and mainly identify the initial fault set that may cause cascading failures, lacking description of the influence of source-side uncertainty and subsequent cascading failures caused by the failure of some components.

[0004] With the large-scale integration of wind power and other new energy sources, the cascading failure characteristics of power systems have changed significantly, and the cascading failure characteristics of wind power systems have become a research hotspot. From the actual cascading failure events closely related to wind power, after a high proportion of wind power is connected to the grid, the power outage accident has obvious cascading failure characteristics, and the coupling between the power generation side and the power transmission side of the power grid is more obvious. From a theoretical point of view, first, compared with synchronous generators, wind power has strong uncertainty; second, wind turbines are connected to the grid through a converter, which decouples frequency and power. The inertia and frequency regulation capability of the current mainstream wind turbine are weaker than that of traditional synchronous machines. Therefore, compared with traditional power systems dominated by synchronous generators, the cascading failure characteristics and weak line distribution of wind power systems will change; however, the current weak line identification methods lack consideration of wind power output uncertainty, wind power output correlation and system cascading failure characteristics, and the accuracy of traditional weak line identification methods needs to be improved in the face of new power systems with high proportion of wind power, and it is necessary to study new weak line identification methods. SUMMARY

[0005] The application aims to provide a weak line evaluation method and system for a wind power system to overcome the low identification accuracy of weak lines in the prior art.

[0006] A weak line evaluation method for a wind power system, comprising the following steps:

[0007] S1, obtaining active power output data records accumulated in the historical wind power operation of the power system, fitting an ultra-short-term wind power output edge distribution function of each wind power node according to the active power output data records using a kernel density function, and fitting the correlation of wind power output using a copula theory;

[0008] S2, establishing a system equivalent frequency response model considering the low inertia and weak frequency response characteristics of the wind power system, and calculating the dynamic process of the system when disturbed through the system equivalent frequency response model;

[0009] S3, establishing a quasi-steady-state cascading failure model considering the uncertainty of new energy output, the quasi-steady-state cascading failure model including island identification, power generation and load rescheduling, deterministic power flow calculation, cascading failure propagation judgment and cascading failure end judgment, and giving the total loss of load and the total propagation time as the final risk indicators of the cascading failure chain based on the quasi-steady-state cascading failure model;

[0010] S4, repeating step S3 to obtain a cascading failure chain data set, the cascading failure chain data set including the propagation process and the final risk indicators of the cascading failure, first constructing a cascading failure network according to the propagation process of the cascading failure chain, and then calculating the cascading failure network value according to the total loss of load and the total propagation time value of the cascading failure chain;

[0011] S5, obtaining the relative weakness value of each line of the system according to the calculation result of the cascading failure network value, constantly repeating steps S3-S4 to update the cascading failure network when the line weakness value does not meet the convergence criterion, and sorting the weakness values of each line in descending order when the line weakness value meets the convergence criterion, so as to obtain the weak lines of the system.

[0012] Preferably, the historical output data of each wind farm is sorted, the kernel density estimation method is used to fit the historical output data of each wind farm, and the edge distribution of the output of each wind farm is obtained.

[0013]

[0014] In the formula, f h (x) is the probability density at x point, K() is a Gaussian kernel function, X i is a sample point, n is the sample capacity, and h is 1.

[0015] Preferably, then the edge distribution is fitted with different copula functions, so as to build a copula function describing the correlation of each random variable, and then compare the square Euclidean distance of each function with the empirical copula function, and the minimum distance is the most appropriate copula function to describe the wind power error distribution.

[0016] Preferably, sampling is performed according to the generated copula function to obtain wind farm time series output sample data with correlation. m*n , m is the dimension of the random variable, which is the number of wind farms to be fitted, and n is the number of data samples. m*n Based on the equal probability conversion principle, the inverse function of the CDF of each wind farm output distribution and the aforementioned U m*n sample are used to obtain the output sample data of the m wind farms that meet the given correlation condition and edge distribution.

[0017] Preferably, the system equivalent frequency response model aims to obtain the frequency response and power response process under system active imbalance, and the input quantity of the system equivalent frequency response model is the system power disturbance and the numerical value of the parameters of each part of the model, and the output is the output of the conventional unit and the load output under the new steady state.

[0018] Preferably, the cascading failure model first sets a random N-2 line fault as the initial fault, samples the time series output of each wind farm based on the joint probability density function of each wind farm, and then performs island identification, power generation and load rescheduling.

[0019] Preferably, the cascading failure network structure construction includes forming nodes, forming directed edges, connecting nodes and directed edges according to rules, the network node formation corresponds to a fault set, which refers to a certain fault set of the system at this stage; the directed edge formation corresponds to a fault, which refers to the fault that will occur in the next stage under the fault set corresponding to the edge.

[0020] Preferably, the cascading failure network value calculation includes fault set value calculation (denoted as F vertex ) and fault value calculation (denoted as F edge ). The fault value of the end fault F edge (F k j ,0) is equal to the fault risk value corresponding to the cascading failure chain, and the fault value of the non-end fault is equal to the fault set value corresponding to the fault; the fault set value is equal to the product of the fault value corresponding to the next stage and the probability of occurrence of each fault.

[0021] Preferably, the cascading failure network is updated constantly, the relative weakness values of each line are calculated according to the cascading failure network value calculation results, the cascading failure network is considered to be relatively complete when the weakness values converge, and the weak lines can be identified according to the descending order of the weakness values of each line at this time.

[0022] A weak line evaluation system for a wind power system, comprising a preprocessing module, a fault network calculation module and an evaluation output module.

[0023] The preprocessing module is used to obtain active power output data records accumulated in historical wind power operation of the power system, fit an ultra-short-term wind power output edge distribution function of each wind power node by using a kernel density function according to the active power output data records, and fit the correlation of the wind power output by using a copula theory; an equivalent frequency response model of the power system considering low inertia and weak frequency response characteristics of the wind power system is established, and a dynamic process of the system when disturbed is calculated through the equivalent frequency response model; a quasi-steady-state cascading failure model considering uncertainty of new energy output is established, the quasi-steady-state cascading failure model includes island identification, power generation and load rescheduling, deterministic power flow calculation, cascading failure propagation judgment and cascading failure end judgment, and the total loss of load and the total propagation time are given as final risk indicators of the cascading failure chain based on the quasi-steady-state cascading failure model;

[0024] The fault network calculation module is used to calculate the cascading failure network value according to the cascading failure chain data set, the cascading failure chain data set including a propagation process of the cascading failure and the final risk indicators.

[0025] The evaluation output module is used to calculate the relative weakness values of each line of the system according to the cascading failure network value calculation results, constantly update the cascading failure network when the weakness values of the lines do not satisfy the convergence criterion, and sort the weakness values of each line in descending order when the weakness values of the lines satisfy the convergence criterion, so as to obtain the weak lines of the system.

[0026] Compared with the prior art, the present application has the following beneficial technical effects:

[0027] The application discloses a weak line evaluation method for a wind power system, considers a system aggregated frequency response model of wind power access, further proposes a quasi-steady-state cascading failure model considering wind power super-short-term output correlation, uncertainty and low inertia characteristics, overcomes defects of existing quasi-steady-state cascading failure models, such as insufficient consideration of wind power output uncertainty and incapability of considering system dynamic frequency characteristics when disturbed, thereby improving accuracy of cascading failure simulation of the wind power system.

[0028] The application simulates cascading failures based on a quasi-steady-state cascading failure model. The model considers system steady-state power flow distribution characteristics and power-frequency response characteristics, performs generation-load rescheduling based on a frequency response model when the system suffers from a large disturbance, and performs generation-load rescheduling based on a dynamic power flow method when power imbalance caused by wind power output uncertainty is generated. Subsequently, according to reliability theory, a generated cascading failure chain data set is used to take loss-of-load degree and total power flow overrun degree as indexes, a cascading failure network method is used to perform structure construction and value calculation. Cascading failure sets are continuously written, and when the cascading failure network value calculation converges, key lines are identified according to risk value sorting of each line. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 For the wind power system aggregated frequency response model in the embodiment of the application, Figure 1 a is a simplified model selected for a large-scale system; Figure 1 b is a model schematic diagram for a small-scale system;

[0030] Figure 2 For the main process of quasi-steady-state cascading failure simulation in the embodiment of the application,

[0031] Figure 3 For the cascading failure network construction process in the embodiment of the application,

[0032] Figure 4 For the complete process of weak line evaluation in the embodiment of the application. DETAILED DESCRIPTION

[0033] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the accompanying drawings of the embodiments of the present application, so that those skilled in the art can better understand the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present application.

[0034] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to the process, method, product or device.

[0035] As shown in Figure 4 , a weak line evaluation method for a power system containing wind power, comprising the following steps:

[0036] S1, obtaining active power output data records accumulated by historical wind power operation of the power system, fitting an ultra-short-term wind power output edge distribution function of each wind power node using a kernel density function according to the active power output data records, and fitting the correlation of wind power output using copula theory;

[0037] S2, establishing a system equivalent frequency response model considering the low inertia and weak frequency response characteristics of the power system containing wind power, the system equivalent frequency response model being used to calculate the dynamic process of the system under disturbance;

[0038] S3, as shown in Figure 2 , a quasi-steady-state cascading failure model considering new energy output uncertainty is established, the quasi-steady-state cascading failure model including island identification, power generation and load rescheduling, deterministic power flow calculation, cascading failure propagation judgment and cascading failure end judgment, and the total loss of load and the total propagation time are given as the final risk indicators of the cascading failure chain based on the quasi-steady-state cascading failure model;

[0039] S4, repeating step S3 to obtain a cascading failure chain data set (the cascading failure chain data set includes a cascading failure propagation process and a final risk indicator), first constructing a cascading failure network according to the cascading failure propagation process, and then performing cascading failure network value calculation according to the total load loss and total propagation time values of the cascading failure chain;

[0040] S5, obtaining a relative weakness value of each line of the system according to the cascading failure network value calculation result, constantly repeating steps S3-S4 to update the cascading failure network when the line weakness value does not satisfy a convergence criterion, and considering that the cascading failure network has been relatively complete when the line weakness value satisfies the convergence criterion, and sorting the line weakness values in descending order at this time, so as to obtain a weak line of the system.

[0041] In step S1, different wind farms with adjacent geographical positions are often in the same wind belt, and the wind speeds thereof have strong correlation, so that the outputs of the wind farms have obvious regional correlation. Therefore, the space-time correlation of the wind farm outputs needs to be considered when sampling the wind farm outputs. First, the historical data of the wind farm outputs are sorted, and the edge distribution of the wind farm outputs is obtained by fitting the historical wind farm output data of each wind farm using a kernel density estimation method. The kernel density function used is:

[0042]

[0043] In the formula, f h (x) is the probability density at the point x, K() is a Gaussian kernel function, X i is a sample point, n is the sample capacity, and h is 1.

[0044] Then, different copula functions (including gaussian copula, gumbel copula, clayton copula, and frank copula) are used to fit the edge distribution, so as to construct a copula function for describing the correlation of the random variables. Then, the square Euclidean distances of the functions and the empirical copula function are compared, and the copula function with the smallest distance is the most suitable copula function for describing the wind power error distribution.

[0045] The empirical copula function is defined as follows:

[0046]

[0047] In the formula, u, v ∈ [0, 1]; F n (), G n are empirical distribution functions of random variables, and I [] is a 0-1 function, which is 1 when F n (X i ≤ u), and 0 otherwise.

[0048] According to the generated copula function, sampling is performed to obtain wind farm time series output sample data with correlation m*n (m is the dimension of random variables, here is the number of wind farms to be fitted, n is the number of data samples), and finally based on the equal probability conversion principle, according to the inverse function of the CDF of the output distribution of each wind farm and the aforementioned generated U m*n samples, that is, the output sample data of m wind farms satisfying the given correlation condition and edge distribution (sample data is m*n dimensional).

[0049] The edge distribution of each wind farm output is obtained

[0050] Suppose there are m wind farms. First, arrange the historical output data of each wind farm by time period, and use kernel density estimation method to fit the historical output data of each wind farm, and obtain the edge cumulative distribution function and inverse function of each wind farm output.

[0051] Model the correlation of each wind farm output:

[0052] Different theoretical copula functions (including gaussian copula, gumbel copula, clayton copula, frank copula, etc.) are used to fit the edge distribution of m wind farms. The parameters of the theoretical copula function are obtained by maximum likelihood method, so as to construct the theoretical copula function describing the correlation of each random variable. Then compare the square Euclidean distance of each theoretical copula function and empirical copula function, and the one with the smallest distance is the most appropriate copula function to describe the correlation of each wind farm output.

[0053] In S2, the system equivalent frequency response model aims to obtain the frequency response and power response process under system active imbalance. The input of the system equivalent frequency response model is the system power disturbance and the numerical value of each part of the model parameters, and the output is the output of the conventional unit and the load output under the new steady state.

[0054] The specific structure of the system equivalent frequency response model takes into account the primary frequency regulation, secondary frequency regulation, and rotor equivalent swing equation, and adds the prime mover output limit, frequency regulation dead zone nonlinear link (when the system unbalanced power is large or the proportion of wind power installed is large, the power of conventional units participating in primary frequency regulation may be limited by the prime mover output, so it is necessary to add the prime mover frequency regulation dead zone and output limit link to the aggregated frequency response model), load reduction / generator link (considering the action of the safety automatic device when the system is subject to large disturbances) and wind power equivalent link (that is, based on the proportion of wind power access, the original rotor swing equation and frequency regulation link are modified). According to the scale of the example system, large-scale systems choose to use a simplified model, such as Figure 1 As shown in a; the small-scale system uses the conventional model structure to consider primary frequency regulation, secondary frequency regulation, high-frequency cutting and low-frequency load reduction, as shown in Figure 1 b. The conventional model only requires calculation of the system's equivalent inertia constant. The simplified model parameter calculation process involves first fitting the individual parameters of all units to calculate the weights of each unit and the overall equivalent frequency regulation coefficient of the system. The overall model parameters (i.e., the equivalent turbine and governor transfer function time constants) are then weighted based on the weights of each unit.

[0055] The model parameter calculation is as follows. The high-frequency cutting parameter is that when the frequency exceeds 51.5HZ, the conventional units are cut off in descending order of output after a delay of 10 seconds; the low-frequency load shedding parameters are shown in Table 1. According to the different wind power access ratios, the rotor motion equation parameters, equivalent turbine transfer function parameters, and system equivalent frequency regulation coefficient of the WSFR model will change accordingly. The unit frequency regulation dead zone parameter is set to ±0.001pu, and the output limit parameter is set to min{0.15pu, min(P geni -P gen imax )}. In the system equivalent swing equation, (1-η M )*M*S represents the system equivalent inertia time constant M(M=ΣM i ×S i / ΣS i , S i is the output of unit i, M i is the inertia constant of unit i) decreases as the proportion of wind power access increases, η M is the wind power access ratio. The frequency response coefficient D of the system load is considered as a constant. At the same time, it is assumed that the system electromagnetic power P e Equal to the load, the initial mechanical power P of the system prime mover m0 Equal to the initial load P e0 The turbine and speed regulation link are only for conventional units, and the fan is set to have no frequency regulation capability.

[0056] Table 1 Low frequency load reduction parameters

[0057]

[0058] The conventional model only aggregates the equivalent rotor swing equation of the system, and does not need to aggregate each turbine and governor. The simplified WSFR model also needs to obtain the equivalent aggregated parameters of the turbine and governor. The equivalent frequency regulation coefficient R can be directly derived from the definition: R i is the frequency regulation value of unit i. Other parameters to be solved (i.e., the parameters of the equivalent prime mover and the governor part of the system) are denoted as X, X = {F H ,T R ,T C ,T G}. Before calculating X, the weight λ i of each unit needs to be calculated first, as shown in equation (5).

[0059]

[0060] where S i is the rated power of the unit. For the generator unit i, the larger λ i is, the larger the rated power of the unit i is and the faster the frequency response speed is, so the calculation of the parameter group X is as shown in equation (6)

[0061]

[0062] In S3, the objective of the cascading failure model is to obtain the fault propagation process and total load loss of the system after the initial fault. The cascading failure model first sets a random N-2 line outage as the initial fault, and samples the time series output samples of each wind farm based on the joint probability density function of each wind farm obtained in step S1. Then, island identification, power generation and load rescheduling (for large disturbances such as system splitting, the frequency response model obtained in step S2 is used for calculation. For small disturbances such as wind power output fluctuations, an active power balance strategy considering the ramping constraint and static frequency response characteristics is used for calculation), deterministic power flow calculation (based on the dynamic power flow method, first calculate the active power of each node after the system reaches a new steady state after each fault according to the calculation result of the frequency response model in step S2, and then perform deterministic power flow calculation), cascading failure propagation judgment (using the inverse time line overloading protection of the line as the cascading failure propagation criterion). There are two criteria for determining whether the cascading failure model is ended: the criterion calculation shows that the cascading failure will not continue to propagate, or the cascading failure propagation process has reached the limit (the limit refers to comparing the results of the proposed cascading failure model with the results of the electromechanical transient simulation of the system based on the time domain simulation software, and obtaining the average limit of the failure of the proposed model). The flow chart of step S3 is shown in Figure 2 .

[0063] Initial fault setting: Since the actual power system satisfies the N-1 safety criterion, two random line tripping is considered as the initial event. At the same time, based on the wind power output distribution fitted in the first step, the minute-level wind power output curve is obtained to take into account the uncertainty of wind power.

[0064] Island identification: During the process of cascading failure propagation, the system may be split into islands or isolated nodes due to the successive tripping of lines. In this paper, the treatment of isolated nodes is full load or generator cut; for islands, power-frequency response calculation is performed first according to their state, and after power balance, DC power flow calculation is performed. After all islands are calculated, the fault propagation time of each island (i.e. line overload protection action delay and frequency regulation time) is synchronized.

[0065] Protection and tripping mechanism: Considering the inverse time line overload protection, the line tripping probability is determined according to the line overload degree, and the tripping time is determined. If there are multiple lines that may trip, a line is randomly selected to trip.

[0066] When the line l flow is within the normal range, the tripping probability P l of the line is:

[0067] P l = P1,0≤ L l ≤ L N lMAX (7)

[0068] When the line l flow is between the normal value and the limit value, the tripping probability P l of the line is linearly fitted: N lMAX where L lMAX is the maximum normal flow of the line, and L lMAX may be taken as the thermal stability limit of the line.

[0069]

[0070] When the line l flow is greater than the limit value L lMAX , the tripping probability P l = P2.

[0071] If the line l trips, its time delay is calculated as shown in (9) and (10)

[0072]

[0073]

[0074] where T k represents the interval time between each level of failure, and Δt lThe trip delay time, T0, of the line 1 is set to 300 seconds.

[0075] Generation and load rescheduling:

[0076] In the process of cascading failure propagation, the active power imbalance of the system is the fundamental cause of the re-distribution of power flow. For actual power systems, due to the system having reserve capacity, the uncertainty of wind power output will not cause frequency instability and other transient instability problems; but when the system is split, etc. Large disturbance may cause frequency instability. When considering the active power imbalance caused by the uncertainty of wind power output, due to the small uncertainty of wind power ultra-short-term output and the long time scale of this process, the system has more resources to balance this type of active power imbalance, so the active power balancing strategy considering the ramping constraint and the static frequency response characteristics of the system is used to calculate the rescheduling process of the system; and for the active power imbalance caused by the possible splitting and other large disturbances in the process of cascading failure propagation, due to the small time scale of the interval between the large disturbance and the corresponding response of the system, the results calculated by the active power balancing strategy considering the ramping constraint and the static frequency response characteristics have large errors with the actual situation. The dynamic frequency response characteristics of the system need to be considered, and the frequency aggregation model proposed in 2) is used to calculate the dynamic process of the system after the active power imbalance and the new steady state of the system after balancing.

[0077] End criterion:

[0078] There are two points for the end criterion: the criterion calculation finds that the cascading failure will not continue to propagate, or the process of cascading failure propagation has reached the limit. The limit refers to the fact that the proposed cascading failure model considers the line power flow transfer as the criterion for cascading failure propagation, without considering the system voltage, power angle and other characteristics and more detailed transient process, which underestimates the impact of component failure on the system. Therefore, to ensure simulation accuracy, only the initial stage of cascading failure is simulated, and if the cascading failure develops to this limit, the simulation is terminated. Specifically, based on the time domain simulation software, the electromechanical transient simulation of the system is performed, and the electromechanical transient simulation results are compared with the proposed cascading failure model to obtain the average limit when the proposed model fails (for example, for the IEEE39 node system, when the number of broken lines exceeds 4, the system will experience voltage instability, power angle instability and other transient instability, and the line power flow limit is no longer the main cause of cascading failure, so the proposed cascading failure model is no longer accurate at this stage, and the limit is 4, and the simulation is ended.

[0079] In step 4), a cascading failure network needs to be established, as shown in Figure 3 which includes structure construction and value calculation.

[0080] The cascading failure network structure construction includes forming nodes, forming directed edges, and connecting nodes and directed edges according to rules. The network node formation corresponds to a failure set, which refers to a certain failure set of the system at this stage; the directed edge formation corresponds to a failure, which refers to a failure that will occur at the next stage under the failure set corresponding to the edge.

[0081] The network is divided into different stages according to the number of line failures, and all failure chains are sorted according to the following rules, that is, the formation of the cascading failure network: the starting point of the directed edge is the failure set of the previous stage corresponding to the failure, the edge is the failure, and the ending point is the failure set of the next stage after the failure occurs.

[0082] A two-state model is used to describe the line, where 0 represents a normal line and 1 represents a line outage. Assuming that the system has n lines, the failure stage k is defined as the total number of line failures at this time, and each failure set is represented as a 1*n-dimensional 01 variable vector, such as the failure set

[10001] , which represents a system composed of 5 lines. The failure set is at stage 2, line 1 and line 5 have failed, and lines 2, 3, and 4 are in normal state. As the failure chain is gradually written, the cascading failure network will tend to be complete.

[0083] The network is divided into different stages according to the total number of line failures, and the connection rule means that all failure chains are connected according to the following rules: assuming that there are j failure sets at stage k, each failure set at stage k (i.e., the node of the network) can be represented as F k j ; assuming that the failure of the next stage of each failure set at stage k is f k , then for the node F k j , the corresponding subsequent failure (i.e., the directed edge) is (F k j , f k ), the starting point of the directed edge is the failure set of the previous stage corresponding to the failure, the edge is the failure, and the ending point is the failure set of the next stage after the failure occurs. In addition, the directed edge (F k j , 0) is the terminal failure of the failure chain (i.e., f k = 0, which is the last failure of the failure chain, and the cascading failure stops after the failure occurs). After connecting the nodes and directed edges according to the rules, the cascading failure network structure is formed.

[0084] The cascading failure network value calculation includes failure set value calculation (denoted as F vertex ) and failure value calculation (denoted as F edge ). The terminal failure F edge (F k jThe failure value of the end fault is equal to the failure risk value corresponding to the chain of the failure, and the failure value of the non-end fault is equal to the failure set value pointed by the failure; the failure set value is equal to the sum of the failure value corresponding to the failure set in the next stage and the product of the occurrence probability of each failure. The value calculation process needs to start from the end fault, which shows the iterative transmission property from the end to the previous stage.

[0085] Chain failure network value calculation:

[0086] In the chain failure propagation process, each failure occurs successively in time, which has the Markov property, that is, the failure occurring in the next stage is only related to the failure set in the current stage and is irrelevant to the previous failure set. The chain failure network described in this paper should also have this property: the failure in the next stage is only related to the failure set in the current stage, and this property will be embodied in the value calculation process.

[0087] In order to identify the weak lines, the weight of each line corresponding to the overall chain failure risk of the system is required. For this purpose, the value calculation of the nodes and directed edges of the chain failure network is required. Value calculation includes failure set value calculation (denoted as F vertex ) and failure value calculation (denoted as F edge ).

[0088] Let k represent the current stage of the chain failure (i.e. the current common k line failure), and the failure set F k j All possible failure sets in the next stage are denoted as All failure sets pointing to the failure set F k j are Define the last fault of each chain of the chain failure as the end fault, and the fault other than the end fault as the non-end fault, in order to facilitate the algebraic description of the chain failure network.

[0089] First, calculate the failure value (i.e. the value of the directed edge of the network) due to the Markov property of the chain failure, the failure value of the non-end fault is equal to the failure set value pointed by the failure, that is,

[0090] F edge (F k j ,f)=F vertex (F k+1 i ),

[0091]

[0092] The end fault F edge (F k jThe failure value of each failure f in the set F (0) is equal to the failure risk value corresponding to the failure chain of f in the set F (0). In this paper, the final failure risk value of a failure chain is described by the total loss of load and the total propagation time of the failure chain.

[0093] F edge (F k j loss (F (0) ) = loss (12)

[0094] Then the failure set value (i.e. the network node value) is calculated. According to the reliability theory, the risk value of an accident is equal to the product of the loss of the accident and the probability of the accident, so the failure set value should be equal to the sum of the product of the failure value of the failure set in the next stage and the probability of each failure: the product of the failure value and the probability of each failure can represent the risk after the failure occurs, and the failure value represents the expected final risk of the whole system after the failure occurs, so the failure set value represents the expected final risk when the system is in the state described by the failure set, as shown in (13).

[0095]

[0096] In (13), n (F (0) ) represents the total number of times that the failure set F (0) occurs in the set of cascading failure chains, n (F (0), f) represents the total number of times that the failure f occurs in the next stage under the condition that the failure set is F (0) in the set of cascading failure chains. k i) indicates that in the set of cascading failure chains, the failure set F (0) is in the state of F (0), and the failure f occurs in the next stage. k i k i i k i i Figure 3 The process of calculating the value of the cascading failure network is shown in (14) : when the value calculation starts, only the failure value of the final failure is known, and the failure values of other failures and the failure set values are unknown, i.e. the value calculation process starts from the end failure, showing the iterative transmission property from the end to the previous stage.

[0097] Step 5) needs to identify the weak line according to the value calculation result of the cascading failure network, which includes the following steps:

[0098] The cascading failure network is constantly updated, and the relative weakness value of each line is obtained according to the value calculation result of the cascading failure network. When the weakness value converges, it is considered that the cascading failure network has been relatively complete, and the weak line can be identified according to the ranking result of the weakness value of each line at this time.

[0099] The calculation process of the line weakness value is as follows: the weak line is divided into two types: the line that is prone to failure itself in the cascading failure and the line that is prone to triggering subsequent large-scale power flow transfer, so two cascading failure risk indicators CFRI l ​​​​​, respectively, are denoted as RI risk l (Indicates the contribution of line l fault to the total load loss of the system) and RI powerflow l (Indicates the contribution of line l fault to the total flow limit of the system), and then the risk value of line l at each stage is summed up as shown in (4) (When calculating the risk value of the line, k = 3 is started, because k = 1, 2 is the initial triggering fault, the occurrence mechanism between the initial triggering stage fault and the subsequent propagation stage fault is different, and the present invention only pays attention to the influence of the propagation stage fault on the system). The risk index RI risk l of line l is large, indicating that the line is not resistant to power fluctuation and splitting disturbance, and the opening of the line during the process of cascading failure propagation is easy to cause system load loss; the risk index RI powerflow l of line l is large, indicating that the opening of the line during the process of cascading failure propagation is easy to cause flow transfer, thereby aggravating the cascading failure propagation.

[0100]

[0101]

[0102]

[0103] The weak value CFRI l of all lines is sorted in descending order, that is, the weak line of the system is obtained. Since it is impossible to obtain all fault chains of the system, when the CFRI l index of each line approaches convergence, it is considered that enough fault chains have been read in, and the recognition result tends to be reliable.

[0104] Since it is impossible to obtain all fault chains of the system, when the CFRI l index of each line approaches convergence, it is considered that enough fault chains have been read in, and the recognition result tends to be reliable. The convergence index is shown in (5), and ε is 0.001. Wherein k max is the limit of the model failure obtained in the third section.

[0105] The present application proposes a system aggregated frequency response model considering wind power integration, and further proposes a quasi-steady cascading failure model considering wind power ultra-short-term output correlation, uncertainty and low inertia characteristics, which overcomes the defects of existing quasi-steady cascading failure models that cannot consider wind power output uncertainty and cannot consider system dynamic frequency characteristics when disturbed, thereby improving the accuracy of cascading failure simulation of wind power integrated power systems.

[0106] The present application first considers wind power output uncertainty and low inertia and weak frequency response characteristics, and simulates cascading failures based on a quasi-steady cascading failure model. The model considers system steady-state power flow distribution characteristics and power-frequency response characteristics, and performs generation-load rescheduling based on a frequency response model when the system is subjected to a large disturbance, and performs generation-load rescheduling based on a dynamic power flow method when power imbalance caused by wind power output uncertainty occurs. Subsequently, based on the generated cascading failure chain data set, the loss of load degree and the total power flow overrun degree are used as indicators to construct the structure and calculate the value using the cascading failure network method. The cascading failure set is written continuously, and when the cascading failure network value calculation converges, the key lines are identified according to the risk value sorting of each line.

Claims

1. A method for evaluating weak lines in a wind power system, characterized in that: The following steps are involved: S1, obtain the active power output data records accumulated from the historical wind power operation of the power system, use the kernel density function to fit the ultra-short-term wind power output marginal distribution function of each wind power node based on the active power output data records, and then use the copula theory to fit the correlation of wind power output; S2, establish a system equivalent frequency response model that takes into account the low inertia and weak frequency response characteristics of the wind power system, and calculate the dynamic process of the system when it is disturbed through the system equivalent frequency response model; S3: Establish a quasi-steady-state cascading failure model that takes into account the uncertainty of renewable energy output. The quasi-steady-state cascading failure model includes island identification, power generation and load rescheduling, deterministic power flow calculation, cascading failure propagation judgment, and cascading failure termination judgment. Based on the quasi-steady-state cascading failure model, the total load loss and total propagation time are given as the final risk indicators of the cascading failure chain. S4, repeating step S3 to obtain a cascading failure chain data set, which includes the propagation process of the cascading failure and the final risk index. First, a cascading failure network is constructed based on the propagation process of the cascading failure chain, and then the cascading failure network value is calculated based on the total load loss and total propagation time of the cascading failure chain. The construction of the cascading failure network structure includes forming nodes, forming directed edges, and connecting nodes and directed edges according to rules. The network nodes form a corresponding fault set, which refers to a certain fault set of the system in the network node formation stage; the directed edges form a corresponding fault, which refers to the fault that will occur in the next stage of the fault set corresponding to the edge. Cascading failure network value calculation includes fault set value calculation F vertex Calculate the fault value F edge , terminal fault F edge (F k j ,0) is equal to the fault risk value corresponding to the chain fault chain, and the fault value of the non-terminal fault is equal to the fault set value pointed to by the fault; The fault set value is equal to the sum of the corresponding fault value of the fault set in the next stage and the product of the corresponding probability of each fault; S5. Calculate the relative weakness value of each line in the system based on the calculation results of the cascading failure network value. When the line weakness value does not meet the convergence criterion, repeat steps S3-S4 to update the cascading failure network. When the line weakness value meets the convergence criterion, sort the line weakness values ​​at this time in descending order to obtain the weak line of the system.

2. A method for evaluating weak lines in a wind power system according to claim 1, characterized in that: The historical output data of each wind farm were sorted out, and the kernel density estimation method was used to fit the historical wind power output data of each wind farm to obtain the marginal distribution of the output of each wind farm. The specific kernel density function used is: Where f h (x) is the probability density at point x, K() uses the Gaussian kernel function, X i is the sample point, n is the sample size, and h is 1.

3. The method for evaluating weak lines in a wind power system according to claim 1, wherein: Then, different copula functions are used to fit the marginal distribution to construct a copula function that describes the correlation of each random variable. The squared Euclidean distance between each function and the empirical copula function is compared. The copula function with the smallest distance is the most suitable copula function to describe the wind power error distribution.

4. The method for evaluating weak lines in a wind power system according to claim 1, wherein: Sampling is performed based on the generated copula function to obtain the correlated wind farm time series output sample data. After obtaining the appropriate theoretical copula function, simple random sampling is used to generate the m*n dimensional uniformly distributed random variable U in the interval [0,1]. m*n , m is the dimension of random variables, here is the number of wind farms to be fitted, n is the number of data samples, and finally based on the principle of equal probability conversion, according to the inverse function of the CDF of the output distribution of each wind farm and the U generated above m*n Samples, that is, the output sample data of m wind farms that meet the given correlation conditions and marginal distribution are obtained.

5. The method for evaluating weak lines in a wind power system according to claim 1, wherein: The system equivalent frequency response model aims to obtain the frequency response and power response process under the system active power imbalance. The input of the system equivalent frequency response model is the system power disturbance and the numerical value of each part of the model parameter. The output is the output of the conventional unit and the load output under the new steady state.

6. The method for evaluating weak lines in a wind power system according to claim 1, wherein: The cascading failure model first sets a random N-2 disconnection fault as the initial fault, and obtains the time-series output samples of each wind farm based on the joint probability density function of each wind farm, and then performs island identification, power generation and load rescheduling.

7. The method for evaluating weak lines in a wind power system according to claim 1, wherein: The cascading failure network is continuously updated, and the relative weakness value of each line is obtained based on the calculation results of the cascading failure network value. When the weakness value converges, it is considered that the cascading failure network is relatively complete. The weak line can be identified based on the ranking results of the weakness values ​​of each line at this time.

8. A weak line assessment system for a wind power system based on the weak line assessment method for a wind power system according to claim 1, characterized in that: The preprocessing module is used to obtain the active power output data records accumulated from the historical wind power operation of the power system, use the kernel density function to fit the ultra-short-term wind power output marginal distribution function of each wind power node based on the active power output data records, and then use the copula theory to fit the correlation of wind power output; establish a system equivalent frequency response model that takes into account the low inertia and weak frequency response characteristics of the power system containing wind power, and calculate the dynamic process of the system when it is disturbed through the system equivalent frequency response model; establish a quasi-steady-state cascading failure model that takes into account the uncertainty of renewable energy output. The quasi-steady-state cascading failure model includes island identification, power generation and load re-dispatching, deterministic power flow calculation, cascading failure propagation judgment and cascading failure end judgment, and based on the quasi-steady-state cascading failure model, the total load loss and total propagation time are given as the final risk indicators of the cascading failure chain; The fault network calculation module obtains the cascading fault chain data set, which includes the propagation process of the cascading fault and the final risk index. The cascading fault network is first constructed based on the propagation process of the cascading fault chain, and then the cascading fault network value is calculated based on the total load loss and total propagation time of the cascading fault chain. The evaluation output module calculates the relative weakness of each line in the system based on the calculation results of the cascading failure network value. When the line weakness value does not meet the convergence criterion, the cascading failure network is repeatedly updated. When the line weakness value meets the convergence criterion, the weakness values ​​of each line at this time are sorted in descending order to obtain the weak line of the system.

Images