CN106330095A - Large-scale photovoltaic power station's collection system fault location method - Google Patents

Abstract

The invention belongs to the technical field of new energy grid-connected power generation technology, in particular to a large-scale photovoltaic power station's collection system fault location method. The method, based on the control of the inverter in an actual photovoltaic power plant, through the utilization of the characteristics of a photovoltaic power generation unit of not outputting negative sequence current under an asymmetrical voltage condition and the combination of the node negative sequence voltage equation corresponding to the sparse measurement point after an asymmetric fault occurs in the photovoltaic power station's collection system and the compression perception theory, and through the assistance of an improved Bayesian compression perceptual reconstruction algorithm, solves the node negative sequence injection current vector to locate a fault. Of all node negative sequence injection current amplitude vectors in a data window, the nodes corresponding to the largest element and appear at the most frequency are regarded as problematic ones. According to the invention, all the method requires is the non-synchronized measurement information of sparse measurement spots. And the method is free of impacts of fault types and transitional resistors. Due to its strong ability to combat noise and its high precision, the method can be practically used to meet the application requirement of large scale photovoltaic power stations.

Description

Fault positioning method for collecting system of large photovoltaic power station

Technical Field

The invention belongs to the technical field of new energy grid-connected power generation, and particularly relates to a fault positioning method in a collecting system of a large photovoltaic power station.

Background

The large-scale centralized photovoltaic power station is formed by collecting a large number of photovoltaic power generation units, and overhead wires and buried cables coexist in a collecting system. The collecting system is not provided with an automatic reclosing device generally, and faults need to be manually checked after the faults trip. The photovoltaic array area is large, the degree of the appearance area is small, line patrol is difficult, the recovery power generation time of a disconnected photovoltaic system is long, and the power generation efficiency is lost. Therefore, the method is beneficial to improving the power generation efficiency of the photovoltaic power station by accurately and quickly positioning the fault position in the collecting system.

At present, fault location methods in power systems mainly include an impedance method, a traveling wave method, an S signal injection method, and a wide area communication method. The impedance method is easily influenced by factors such as power supply parameters and load parameters, and when more branches in the network have complex structures, false fault points are easy to appear. The traveling wave method is difficult to realize the accurate synchronous acquisition of the multi-end traveling wave signals under the condition of numerous complicated branches of the network. The injection signal intensity of the S signal injection method is limited by PT capacity, and the positioning accuracy is greatly influenced by factors such as wire distributed capacitance, grounding resistance and the like. The positioning accuracy of the method based on the single-point measurement information is greatly influenced by a complex network structure and the feeding fault current of the photovoltaic system.

The wide area communication method utilizes fault information of a plurality of measuring points in the power distribution network, and determines a fault section according to the difference of fault characteristics of the fault section and a non-fault section. In the existing fault positioning method based on multipoint measurement information, the method based on information interaction among measurement points needs synchronous measurement information and has high communication requirements. Most methods based on voltage sag can not avoid the cycle traversal process of possible fault points of the whole network, and when the network scale is large, the operation process is very long and the positioning accuracy is poor. When the access of the distributed power supply is considered, the distributed power supply is simply equivalent to a constant current source or a controllable source under the system frequency by the conventional method, the distributed and centralized photovoltaic grid-connected power generation systems are obviously different in topology and control, so that the fault current characteristics are greatly different, the conventional equivalent method is difficult to apply, the photovoltaic power supply permeability in the collection system reaches 100%, and the control characteristic of a photovoltaic inverter cannot be ignored. Therefore, it is necessary to research a fault location method suitable for the control characteristics of the inverter in the large photovoltaic power station.

Disclosure of Invention

In order to solve the problems, the invention provides a method for positioning faults in a collecting system of a large photovoltaic power station, which is characterized by comprising the following steps:

step 1, analyzing and obtaining a fault current characteristic rule in a photovoltaic power station collection system by combining a photovoltaic power station controlled by an inverter;

step 2, offline line connection and parameters in a photovoltaic power station collection system form a node negative sequence impedance matrix, selecting a line where a sparse measurement node is located, and taking absolute values of all elements to form a perception matrix;

step 3, after asymmetric faults occur in the photovoltaic power station collection system, combining the characteristic that the fault current of the photovoltaic system in the photovoltaic power station collection system has no negative sequence component, calculating by using the voltage information of the measured nodes after the faults to obtain a measured point negative sequence voltage amplitude vector, forming an underdetermined equation set for solving the sparse node negative sequence injection current amplitude vector by using a sensing matrix, and solving by using an improved Bayesian compression sensing reconstruction algorithm to obtain a sparse solution;

and 4, counting the occurrence times of the node corresponding to the maximum element in the node negative sequence injection current amplitude vector in the data window length, and regarding the maximum node as a fault node.

The specific process of the step 1 is as follows: the method is characterized in that the characteristic analysis of fault current in a photovoltaic power station collecting system is combined with the control of a photovoltaic power generation system in the photovoltaic power station collecting system, the suppression of negative sequence component in inverter grid-connected current is taken as a control target, and reference current under a double-synchronous rotating coordinate is expressed as follows:

$\left[\begin{array}{c}{i}_d^{+*}\\ i_q^{+*}\\ i_d^{-*}\\ i_q^{-*}\end{array}\right]=\frac{2}{3\[{\left(e_d^{+}\right)}^2+{\left(e_q^{+}\right)}^2\]}\left[\begin{array}{cc}{e}_d^{+}& e_q^{+}\\ e_q^{+}& -e_d^{+}\\ 0& 0\\ 0& 0\end{array}\right]\left[\begin{array}{c}{P}_0^{*}\\ Q_0^{*}\end{array}\right]$

in the formula,andthe positive sequence components of the d and q axes of the grid-connected point voltage,andis a grid-connected reference current d and q axis positive sequence component,andnegative sequence components of d and q axes of grid-connected reference current are obtained;for a given active power reference value,a given reactive power reference value;

obtaining the positive sequence component amplitude of the grid-connected current of the photovoltaic inverter:wherein,is the amplitude of the positive sequence component of the grid-connected current,for the magnitude of the positive sequence component of the grid-connected point voltage,

according to the apparent power actually provided by the photovoltaic power generation system during the fault, the amplitude of the grid-connected current positive sequence component can be written as follows:wherein gamma is a positive sequence voltage drop coefficient, E_mThe grid-connected point voltage amplitude before the fault; p₀' active Power direct Current component, Q, supplied to photovoltaic Power generating Unit during Fault₀' is a reactive power direct current component provided by the photovoltaic power generation unit during a fault;

obtaining a three-phase fault current expression:

$\left\{\begin{array}{c}{I}_a=\frac{2P_0^{\′}}{3{\mathrm{\γE}}_m}\sin (\mathrm{\ω}t+\mathrm{\π}/2+{\mathrm{\θ}}^{+})\\ I_b=\frac{2P_0^{\′}}{3{\mathrm{\γE}}_m}\sin (\mathrm{\ω}t-\mathrm{\π}/6+{\mathrm{\θ}}^{+})\\ I_c=\frac{2P_0^{\′}}{3{\mathrm{\γE}}_m}\sin (\mathrm{\ω}t+7\mathrm{\π}/6+{\mathrm{\θ}}^{+})\end{array}\right.$

in the formula I_a、I_bAnd I_cThree-phase fault currents respectively; intermediate variables

In the step 2, the perception matrix calculation method comprises the following steps:

node negative sequence impedance array Z formed according to line topology and parameters in photovoltaic power station collection system_N×N，

$Z_{N\×N}={\left[\begin{array}{cccccc}{Z}_{11}^{neg}& Z_{12}^{neg}& \mathrm{...}& Z_{1l}^{neg}& \mathrm{...}& Z_{1N}^{neg}\\ Z_{21}^{neg}& Z_{22}^{neg}& \mathrm{...}& Z_{2l}^{neg}& \mathrm{...}& Z_{2N}^{neg}\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ Z_{k1}^{neg}& Z_{k2}^{neg}& \mathrm{...}& Z_{kl}^{neg}& \mathrm{...}& Z_{kN}^{neg}\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ Z_{N1}^{neg}& Z_{N2}^{neg}& \mathrm{...}& Z_{Nl}^{neg}& \mathrm{...}& Z_{NN}^{neg}\end{array}\right]}_{N\×N}$

In the formula, N is the number of network nodes in the photovoltaic power station collection system,is the negative sequence mutual impedance between node k and node l, when k equals lIs a negative sequence self-impedance.

Selecting a row corresponding to a measuring point from a node negative sequence impedance array according to the distribution position of the measuring points in the photovoltaic power station collection system, and taking absolute values of all elements to obtain a sensing matrix Z^neg，

$Z^{neg}={\left[\begin{array}{cccccc}\left|Z_{i_{1}1}^{neg}\right|& \left|Z_{i_{1}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{1}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{1}N}^{neg}\right|\\ \left|Z_{i_{2}1}^{neg}\right|& \left|Z_{i_{2}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{2}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{2}N}^{neg}\right|\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ \left|Z_{i_{j}1}^{neg}\right|& \left|Z_{i_{j}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{j}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{j}N}^{neg}\right|\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ \left|Z_{i_{M}1}^{neg}\right|& \left|Z_{i_{M}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{M}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{M}N}^{neg}\right|\end{array}\right]}_{M\×N}$

In the formula, M is the installation number of the measuring points; i.e. i_jThe node number j corresponding to the jth measuring point is 1,2, … and M;represents a measurement point i_jAnd the negative-sequence trans-impedance modulus value with node l.

The underdetermined equation set in the step 3 is as follows:

${\left[\begin{array}{c}{V}_{i_1}^{neg}\\ V_{i_2}^{neg}\\ .\\ .\\ .\\ V_{i_j}^{neg}\\ .\\ .\\ .\\ V_{i_M}^{neg}\end{array}\right]}_{M\×1}={\left[\begin{array}{cccccc}\left|Z_{i_{1}1}^{neg}\right|& \left|Z_{i_{1}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{1}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{1}N}^{neg}\right|\\ \left|Z_{i_{2}1}^{neg}\right|& \left|Z_{i_{2}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{2}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{2}N}^{neg}\right|\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ \left|Z_{i_{j}1}^{neg}\right|& \left|Z_{i_{j}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{j}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{j}N}^{neg}\right|\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ \left|Z_{i_{M}1}^{neg}\right|& \left|Z_{i_{M}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{M}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{M}N}^{neg}\right|\end{array}\right]}_{M\×N}{\left[\begin{array}{c}0\\ 0\\ .\\ .\\ .\\ \left|I_j^{neg}\right|\\ .\\ .\\ .\\ 0\end{array}\right]}_{N\×1}$

the above formula is abbreviated as V^neg＝Z^neg·I^neg；V^negIs M × 1 dimension measure point negative sequence voltage amplitude vector, wherein the elementRepresents a measurement point i_jThe node is a negative sequence voltage module value; z^negIs an M × N-dimensional perception matrix, I^negInjecting a current amplitude vector for a node negative sequence to be solved in N × 1 dimensions;representing the negative sequence current injected into the negative sequence network by the fault point.

The improved Bayes compressed sensing reconstruction algorithm in the step 3 is based on a compressed sensing theory and a sparse Bayes learning SBL algorithm;

the compressed sensing theoretical model y is phi theta + e, y is observation data with M × 1 dimensions, phi is a sensing matrix with M × N dimensions, theta is a sparse vector to be reconstructed with N × 1 dimensions, and e is N × 1 dimensions and obeys N (0, sigma)²) White Gaussian noise of (1), wherein M<<N; the compressed sensing optimization reconstruction algorithm utilizes the sparsity of theta, and solves the underdetermined equation set to recover the sparse theta with high probability and high precision through a small amount of measuring point data; i is^negCorresponding to the sparse vector theta, Z to be reconstructed^negCorresponding to the sensing matrix phi, V^negCorresponds to the observed data y.

The SBL algorithm assumes that each element of θ obeys a gaussian distribution:α_iis an unknown hyper-parameterAnd the number is the inverse variance of the Gaussian distribution, and a Gaussian likelihood model of compressed sensing is obtained:in the SBL algorithm, a Relevance Vector Machine (RVM) layering prior is used as a sparse prior of theta, and is conjugated with a Gaussian likelihood model, so that the calculation is convenient. Defining the independent of each variable of theta in the sparse vector, and following the Gaussian distribution of zero mean:further, assume that the hyperparameter α obeys the gamma prior distribution:and obtaining a prior distribution form of theta through marginalizing the hyper-parameter α:when theta is_iAs a result of the observation data,is a likelihood function, gamma distribution (α)_i| a, b) is the inverse variance α from the Gaussian distribution_iIs conjugated. The theta prior distribution now satisfies the t-distribution. When the parameters a and b are reasonably selected, the t-distribution at theta can be obtained_iAlso, a gamma prior p (β | c, d) — (β | a, b) is introduced, where β ═ σ ═ a σ a^-2RVM is a mechanism for solving hyper-parameters α and noise inverse variance sigma of sparse vector from learning iteration process^-2All of a, b, c, d are generally taken to be 0, in which case the hyperparameters α and σ are^-2From the prior distribution and the observed data y, the posterior distribution of p (theta, α, sigma) can be obtained by Bayesian formula²|y)＝p(θ|y,α,σ²)p(α,σ²Y), which involves a large number of matrix inversionsAnd (6) operation. In order to avoid huge calculation amount, a rapid RVM algorithm provided by taping and Faul is adopted, point estimation of a sparse vector theta is obtained through posterior estimation, and theta sparsity can be automatically ensured in an optimization solving process without parameter control.

The specific process of the step 4 is

1) Firstly, a node negative sequence impedance matrix is generated in an off-line manner according to structural parameters in a photovoltaic power station collection system, and Z is formed according to measuring point distribution^neg；

2) Reading voltage data of a measuring point after a fault, and calculating V of a continuous window length^neg；

3) Will Z^negV of starting time in window length^negSending the parameters into a Bayesian compressed sensing reconstruction algorithm for calculation to obtain I^neg；

4) Continuously reading V at the next moment^negRepeating the step 3) until the V of the whole window length is read^neg；

5) And counting the occurrence times of the node corresponding to the element with the maximum median of the reconstructed vectors at each moment, wherein the node with the maximum occurrence times is regarded as a fault node.

In the step 4, the number of non-zero elements in the reconstructed vector theoretically corresponds to the number of fault nodes. However, due to the influence of the actual performance of the algorithm, measurement, transmission noise and the like, it is difficult to ensure that the number of non-zero elements in the reconstructed vector corresponds to the number of fault nodes. In order to reduce the possibility of misjudgment, the method observes the reconstruction result of a section of continuous window length, and counts the occurrence frequency of the node corresponding to the element with the maximum median of the reconstruction vector at each moment, wherein the node with the maximum occurrence frequency is regarded as a fault node.

Advantageous effects

(1) The influence of photovoltaic output conditions is avoided;

(2) measuring points are sparse, measuring information does not need to be synchronous, and data used for positioning is short;

(3) the device is not affected by fault types and transition resistance, and has strong noise resistance.

Drawings

FIG. 1 is a flow chart of a method for locating faults in a collection system of a large photovoltaic power station according to the present invention

FIG. 2 is a topological schematic diagram of a collection system of a large photovoltaic power station

FIG. 3 is a schematic view of a negative sequence network of the convergence system

FIG. 4 is a flowchart of a fault location algorithm of the present invention

I at node 15 of FIG. 5 when an ABG-10 Ω fault occurs^negReconstructing the results

I at BC-25 Ω failure at node 40 of FIG. 6^negReconstructing the results

I at node 8 of FIG. 7 in the event of a CA-10 Ω failure^negReconstructing the results

Detailed Description

The present invention will be described in detail below with reference to the accompanying drawings. The invention provides a method for positioning faults in a collecting system of a large photovoltaic power station, and fig. 1 is a flow chart of the method for positioning the faults in the collecting system of the large photovoltaic power station, which comprises the following steps:

step 1, analyzing and obtaining a fault current characteristic rule in a collecting system by combining control adopted by an inverter in a photovoltaic power station;

step 2, collecting line connection and parameter offline in the photovoltaic power station to form a node negative sequence impedance matrix, selecting a line where a sparse measurement node is located, and taking absolute values of all elements to form a sensing matrix;

step 3, after asymmetric faults occur in the collection system, combining the characteristic that fault current of a photovoltaic system in the station has no negative sequence component, calculating by using voltage information of measured nodes after the faults to obtain a voltage amplitude vector of the negative sequence of the measured points, forming an underdetermined equation set for solving sparse node negative sequence injection current amplitude vectors by using a sensing matrix, and solving by using an improved Bayesian compressed sensing reconstruction algorithm to obtain a sparse solution;

and 4, counting the occurrence times of the node corresponding to the maximum element in the node negative sequence injection current amplitude vector in the data window length, and regarding the maximum node as a fault node.

Fig. 2 is a topological structure diagram of access of a single collection station in a collection system of a large photovoltaic power station, wherein a three-winding transformer in the diagram represents a photovoltaic power generation unit with a rated capacity of 1MW, a plurality of photovoltaic power generation units are accessed to a collection cable, and a plurality of collection cables are accessed to the collection station in a photovoltaic array area and then accessed to a booster station through an overhead line for centralized grid connection.

Based on control adopted by an inverter in an actual photovoltaic power station, theoretical analysis is carried out on the characteristics of fault current in a collecting system, and a photovoltaic power generation unit for inhibiting negative sequence current control does not output negative sequence current under the condition of asymmetric voltage. When an asymmetric fault occurs at a node in the trunking system, only one negative-sequence power supply is available at the faulty node in the entire network, as shown in fig. 3. The fault point negative sequence power supply can be regarded as a negative sequence current source, and the negative sequence voltage of each node can be obtained by the product of the node negative sequence impedance array and a sparse node negative sequence injection current vector (containing only one non-zero element):

${\left[\begin{array}{c}{V}_1^{neg}\\ V_2^{neg}\\ .\\ .\\ .\\ V_k^{neg}\\ .\\ .\\ .\\ V_N^{neg}\end{array}\right]}_{N\&times;1}={\left[\begin{array}{cccc}{Z}_{11}^{neg}& Z_{12}^{neg}& \mathrm{...}& Z_{1N}^{neg}\\ Z_{21}^{neg}& Z_{22}^{neg}& \mathrm{...}& Z_{2N}^{neg}\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ Z_{k1}^{neg}& Z_{k2}^{neg}& \mathrm{...}& Z_{kN}^{neg}\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ Z_{N1}^{neg}& Z_{N2}^{neg}& \mathrm{...}& Z_{NN}^{neg}\end{array}\right]}_{N\&times;N}{\left[\begin{array}{c}0\\ 0\\ .\\ .\\ .\\ I_k^{neg}\\ .\\ .\\ .\\ 0\end{array}\right]}_{N\&times;1}$

in the formula,and the negative sequence current injected into the negative sequence network by the fault point is shown, and N is the number of nodes of the network of the in-station collection system.

And (3) considering that voltage measuring devices are arranged on M (M < N) nodes in the collection system, extracting corresponding M rows in the formula to obtain a node negative sequence voltage equation consisting of rows with known node negative sequence voltages. In order to avoid phase calculation and synchronization requirements on measurement information, both sides of the equation are in the form of amplitude values:

${\left[\begin{array}{c}\left|V_{i_1}^{neg}\right|\\ \left|V_{i_2}^{neg}\right|\\ .\\ .\\ .\\ \left|V_{i_M}^{neg}\right|\end{array}\right]}_{M\&times;1}={\left[\begin{array}{cccc}\left|Z_{i_{1}1}^{neg}\right|& \left|Z_{i_{1}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{1}N}^{neg}\right|\\ \left|Z_{i_{2}1}^{neg}\right|& \left|Z_{i_{2}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{2}N}^{neg}\right|\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ \left|Z_{i_{M}1}^{neg}\right|& \left|Z_{i_{M}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{M}N}^{neg}\right|\end{array}\right]}_{M\&times;N}{\left[\begin{array}{c}0\\ 0\\ .\\ .\\ .\\ \left|I_k^{neg}\right|\\ .\\ .\\ .\\ 0\end{array}\right]}_{N\&times;1}$

wherein M is the number of measuring point installations. i.e. i_kThe node number corresponding to the kth measurement point (k ═ 1,2, …, M) is shown. The above formula is abbreviated as V^neg＝Z^neg·I^neg，V^negIs a M × 1 dimension measure point negative sequence voltage amplitude vector, Z^negIs an M × N-dimensional perception matrix, I^negThe current magnitude vector is injected for the N × 1 dimension node negative sequence to be solved.Representing the negative sequence current injected into the negative sequence network by the fault point.

The above formula is an underdetermined equation set, infinite solutions exist, sparse solutions cannot be obtained by adopting the traditional least square method, and misjudgment of fault points or difficulty in judgment are easily caused. And the compressed sensing theory can effectively avoid the problems. Compressed sensing utilizes a small amount of measuring point data, and an original signal which is sparse enough can be recovered by solving an underdetermined equation set through an optimized reconstruction algorithm. In view of strong correlation between the sensing matrix array and the array obtained by the node impedance matrix and original signal I to be reconstructed^negThe method has the characteristic of extremely high sparsity, and the Bayes compressed sensing reconstruction algorithm based on the sparse Bayes learning algorithm is adopted, so that the algorithm is more suitable for processing the condition of strong correlation between columns in a sensing matrix than other optimization reconstruction algorithms, and the most sparse solution and stronger anti-noise capability are more easily obtained.

Fig. 4 is a flowchart of the fault location provided by the present invention, and the location steps after a fault occurs in the convergence system are as follows:

1) firstly, a node negative sequence impedance matrix is generated in an off-line mode according to the structural parameters of a collection system in a photovoltaic power station, and Z is formed according to the distribution of measuring points^neg；

2) Reading voltage data of a measuring point after a fault, and calculating V of a continuous window length^neg；

3) Will Z^negV of starting time in window length^negSending the parameters into a Bayesian compressed sensing reconstruction algorithm for calculation to obtain I^neg；

4) Continuously reading V at the next moment^negRepeating the step 3) until the V of the whole window length is read^neg；

5) Theoretically, the number of non-zero elements in the reconstructed vector corresponds to the number of fault nodes. However, due to the influence of the actual performance and measurement of the algorithm, transmission noise and the like, it is difficult to ensure that only one non-zero element (considering a single fault) is contained in the reconstructed vector. Therefore, in order to reduce the possibility of erroneous judgment, the reconstruction result of a section of continuous window length is observed, the frequency of occurrence of the node corresponding to the element with the largest mean value of the reconstruction vector at each moment is counted, and the node with the largest frequency of occurrence is regarded as a fault node.

Taking the collection system given in FIG. 3 as an example model, the number of points measured is 12. Asymmetric faults of different fault types and transition resistances are applied to all nodes in the system, the situation that measured data contain high noise is considered, and the positioning result is classified according to the distance between the calculation result of the method and a real fault node, as shown in table 1. Table 1 shows the positioning results of traversing all the nodes by the method, and the results in table 1 show that when a fault occurs in the convergence system, the positions of most of the fault nodes can be found more accurately by using the sparse measurement point asynchronous amplitude information. Due to the problem of the system structure, part of branches can not circulate negative sequence current, so that errors exist in positioning, and faults are positioned at adjacent nodes. However, such errors do not have much impact on fault clearance considering the actual wiring distance within the station.

TABLE 1

FIG. 5 shows I when an AB two-phase short circuit occurs at node 15 and a 10 Ω transition resistor ground fault occurs^negReconstructing the result, wherein the x axis in the graph is the node number, the y axis is the time window length, and the z axis represents I^negThe magnitude of the middle element. Reconstruction result I at any time in window length^negThe 15 th element in the data is the largest, and the rest elements are all near 0, so that the fault node can be visually judged through a reconstruction result and is consistent with the position of the real fault node.

FIG. 6 shows the BC two-phase 25 Ω transition resistance short circuit fault at node 40, and the voltage measurement signal with 3% noise^negAnd reconstructing a result. At this time, the reconstruction result I at each time point in the window length^negIn which a plurality of non-zero elements with distinct amplitudes are present, but at each instant I^negThe 40 th element is much larger than the rest elements, so that the fault node can still be visually judged through the reconstruction result and is consistent with the real fault node.

FIG. 7 shows the voltage I when the two phases of CA are short-circuited via 10 Ω transition resistor at node 8^negThe maximum element corresponding node at different moments in the reconstruction result is 7-11, and the final positioning result is a node 7 which is a real fault node adjacent node.

In summary, the method utilizes short asynchronous measurement information after the fault of the sparse measurement point, and can accurately find most fault nodes under the conditions of different fault types, transition resistance and high measurement noise. Due to the structural problem of the system, negative sequence current cannot flow in part of the branches, when faults occur at the last measuring point and the downstream node on the collection branch, the node corresponding to the maximum element in the reconstruction result is distributed between the measuring point and the tail end node of the branch, the final positioning result cannot exceed the range, even in a smaller range, the error cannot bring great influence to actual fault clearing, and the actual application requirements can be met.

In view of the analysis basis and conditions of the method, the method is not influenced by the operation conditions (such as output conditions, power factors and the like) of the photovoltaic system, only short asynchronous measurement information after the fault of the sparse measurement point is needed, and good fault positioning precision can be ensured under the conditions of different fault types, transition resistance and high measurement noise.

It should be noted that the above-mentioned embodiments are only preferred embodiments of the present invention, and should not be construed as limiting the scope of the present invention, and any minor changes and modifications to the present invention are within the scope of the present invention without departing from the spirit of the present invention.

Claims (6)

1. A method for positioning faults in a collection system of a large photovoltaic power station is characterized by comprising the following steps:

step 1, analyzing to obtain a fault current characteristic rule in a collection system of the photovoltaic power station by combining an inverter control mode actually adopted in the photovoltaic power station;

step 2, offline line connection and parameters in a photovoltaic power station collection system form a node negative sequence impedance matrix, selecting a line where a sparse measurement node is located, and taking absolute values of all elements to form a perception matrix;

step 3, after asymmetric faults occur in the photovoltaic power station collection system, combining the characteristic that the fault current of the photovoltaic system in the photovoltaic power station collection system has no negative sequence component, calculating by using the voltage information of the measured nodes after the faults to obtain a measured point negative sequence voltage amplitude vector, forming an underdetermined equation set for solving the sparse node negative sequence injection current amplitude vector by using a sensing matrix, and solving by using an improved Bayesian compression sensing reconstruction algorithm to obtain a sparse solution;

and 4, counting the occurrence times of the node corresponding to the maximum element in the node negative sequence injection current amplitude vector in the data window length, and regarding the maximum node as a fault node.

2. The method for locating faults in a collection system of large photovoltaic power stations as claimed in claim 1, wherein the specific process of step 1 is as follows: the method is characterized in that the characteristic analysis of fault current in a photovoltaic power station collecting system is combined with the control of a photovoltaic power generation system in the photovoltaic power station collecting system, the suppression of negative sequence component in inverter grid-connected current is taken as a control target, and reference current under a double-synchronous rotating coordinate is expressed as follows:

$\left[\begin{array}{c}{i}_d^{+*}\\ i_q^{+*}\\ i_d^{-*}\\ i_q^{-*}\end{array}\right]=\frac{2}{3\&lsqb;{\left(e_d^{+}\right)}^2+{\left(e_q^{+}\right)}^2\&rsqb;}\left[\begin{array}{cc}{e}_d^{+}& e_q^{+}\\ e_q^{+}& -e_d^{+}\\ 0& 0\\ 0& 0\end{array}\right]\left[\begin{array}{c}{P}_0^{*}\\ Q_0^{*}\end{array}\right]$

in the formula,andthe positive sequence components of the d and q axes of the grid-connected point voltage,andis a grid-connected reference current d and q axis positive sequence component,andnegative sequence components of d and q axes of grid-connected reference current are obtained;for a given active power reference value, Q₀ ^*A given reactive power reference value;

obtaining the positive sequence component amplitude of the grid-connected current of the photovoltaic inverter:wherein,is the amplitude of the positive sequence component of the grid-connected current, for the magnitude of the positive sequence component of the grid-connected point voltage,

according to the apparent power actually provided by the photovoltaic power generation system during the fault, the amplitude of the positive sequence component of the grid-connected currentWherein gamma is a positive sequence voltage drop coefficient, E_mThe grid-connected point voltage amplitude before the fault; p₀' active Power direct Current component, Q, supplied to photovoltaic Power generating Unit during Fault₀' is a reactive power direct current component provided by the photovoltaic power generation unit during a fault;

obtaining a three-phase fault current expression:

$\left\{\begin{array}{c}{I}_a=\frac{2P_0^{\&prime;}}{3{\mathrm{\&gamma;E}}_m}\sin (\mathrm{\&omega;}-\mathrm{\&pi;}/2+{\mathrm{\&theta;}}^{+})\\ I_b=\frac{2P_0^{\&prime;}}{3{\mathrm{\&gamma;E}}_m}\sin (\mathrm{\&omega;}-\mathrm{\&pi;}/6+{\mathrm{\&theta;}}^{+})\\ I_c=\frac{2P_0^{\&prime;}}{3{\mathrm{\&gamma;E}}_m}\sin (\mathrm{\&omega;}t+7\mathrm{\&pi;}/6+{\mathrm{\&theta;}}^{+})\end{array}\right.$

in the formula I_a、I_bAnd I_cThree-phase fault currents respectively; intermediate variable theta⁺＝arctan(i_q ⁺/i_d ⁺)。

3. The method for locating faults in the collection system of the large photovoltaic power station as claimed in claim 1, wherein in the step 2, the sensing matrix calculation method comprises the following steps:

node negative sequence impedance array Z formed according to line topology and parameters in photovoltaic power station collection system_N×N，

$Z_{N\&times;N}={\left[\begin{array}{cccccc}{Z}_{11}^{neg}& Z_{12}^{neg}& \mathrm{...}& Z_{1l}^{neg}& \mathrm{...}& Z_{1N}^{neg}\\ Z_{21}^{neg}& Z_{22}^{neg}& \mathrm{...}& Z_{2l}^{neg}& \mathrm{...}& Z_{2N}^{neg}\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ Z_{k1}^{neg}& Z_{k2}^{neg}& \mathrm{...}& Z_{kl}^{neg}& \mathrm{...}& Z_{kN}^{neg}\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ Z_{N1}^{neg}& Z_{N2}^{neg}& \mathrm{...}& Z_{Nl}^{neg}& \mathrm{...}& Z_{NN}^{neg}\end{array}\right]}_{N\&times;N}$

In the formula, N is the number of network nodes in the photovoltaic power station collection system,is the negative sequence mutual impedance between node k and node l, when k equals lIs a negative sequence self-impedance.

Selecting a row corresponding to a measuring point from a node negative sequence impedance array according to the distribution position of the measuring points in the photovoltaic power station collection system, and taking absolute values of all elements to obtain a sensing matrix Z^neg，

$Z^{neg}={\left[\begin{array}{cccccc}\left|Z_{i_{1}1}^{neg}\right|& \left|Z_{i_{1}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{1}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{1}N}^{neg}\right|\\ \left|Z_{i_{2}1}^{neg}\right|& \left|Z_{i_{2}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{2}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{2}N}^{neg}\right|\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ \left|Z_{i_{j}1}^{neg}\right|& \left|Z_{i_{j}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{j}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{j}N}^{neg}\right|\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ \left|Z_{i_{M}1}^{neg}\right|& \left|Z_{i_{M}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{M}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{M}N}^{neg}\right|\end{array}\right]}_{M\&times;N}$

In the formula, M is the installation number of the measuring points; i.e. i_jThe node number j corresponding to the jth measuring point is 1,2, … and M;represents a measurement point i_jAnd the negative-sequence trans-impedance modulus value with node l.

4. The method of locating faults within a collection system of large photovoltaic power plants as claimed in claim 1 wherein the underdetermined system of equations in step 3 is:

${\left[\begin{array}{c}\left|V_{i_1}^{neg}\right|\\ \left|V_{i_2}^{neg}\right|\\ .\\ .\\ .\\ \left|V_{i_j}^{neg}\right|\\ .\\ .\\ .\\ \left|V_{i_M}^{neg}\right|\end{array}\right]}_{M\&times;1}={\left[\begin{array}{cccccc}\left|Z_{i_{1}1}^{neg}\right|& \left|Z_{i_{1}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{1}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{1}N}^{neg}\right|\\ \left|Z_{i_{2}1}^{neg}\right|& \left|Z_{i_{2}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{2}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{2}N}^{neg}\right|\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ \left|Z_{i_{j}1}^{neg}\right|& \left|Z_{i_{j}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{j}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{j}N}^{neg}\right|\\ .& .& & .& & .\\ .& .& & .& & .\\ .& .& & .& & .\\ \left|Z_{i_{M}1}^{neg}\right|& \left|Z_{i_{M}2}^{neg}\right|& \mathrm{...}& \left|Z_{i_{M}l}^{neg}\right|& \mathrm{...}& \left|Z_{i_{M}N}^{neg}\right|\end{array}\right]}_{M\&times;N}{\left[\begin{array}{c}0\\ 0\\ .\\ .\\ .\\ \left|I_j^{neg}\right|\\ .\\ .\\ .\\ 0\end{array}\right]}_{N\&times;1}$

the above formula is abbreviated as V^neg＝Z^neg·I^neg；V^negIs M × 1 dimension measure point negative sequence voltage amplitude vector, wherein the elementRepresents a measurement point i_jThe node is a negative sequence voltage module value; z^negIs an M × N-dimensional perception matrix, I^negInjecting a current amplitude vector for a node negative sequence to be solved in N × 1 dimensions;representing the negative sequence current injected into the negative sequence network by the fault point.

5. The method for locating faults in a collection system of large photovoltaic power stations as claimed in claim 1, wherein the improved Bayesian compressive sensing reconstruction algorithm in the step 3 is based on a compressive sensing theory and a sparse Bayesian learning SBL algorithm;

the compressed sensing theoretical model y is phi theta + e, y is observation data with M × 1 dimensions, phi is a sensing matrix with M × N dimensions, theta is a sparse vector to be reconstructed with N × 1 dimensions, and e is N × 1 dimensions and obeys N (0, sigma)²) White Gaussian noise of (1), wherein M<<N; the compressed sensing optimization reconstruction algorithm utilizes the sparsity of theta, and solves the underdetermined equation set to recover the sparse theta with high probability and high precision through a small amount of measuring point data; i is^negCorresponding to the sparse vector theta, Z to be reconstructed^negCorresponding to the sensing matrix phi, V^negCorresponds to the observed data y.

6. The method for locating faults in a collection system of large photovoltaic power stations as claimed in claim 1, wherein the specific process of step 4 is

1) Firstly, a node negative sequence impedance matrix is generated in an off-line manner according to structural parameters in a photovoltaic power station collection system, and Z is formed according to measuring point distribution^neg；

2) Reading voltage data of a measuring point after a fault, and calculating V of a continuous window length^neg；

3) Will Z^negV of starting time in window length^negSending the parameters into a Bayesian compressed sensing reconstruction algorithm for calculation to obtain I^neg；

4) Continuously reading V at the next moment^negRepeating the step 3) until the V of the whole window length is read^neg；

5) And counting the occurrence times of the node corresponding to the element with the maximum median of the reconstructed vectors at each moment, wherein the node with the maximum occurrence times is regarded as a fault node.