Distributed power supply power distribution network fault positioning method and device

Abstract

The invention belongs to the technical field of power distribution network fault positioning, and particularly relates to a distributed power supply power distribution network fault positioning method and device. The method comprises the following steps: obtaining fault data of different fault types in a power distribution network, and carrying out harmonic extraction on the fault data to obtain each harmonic component of the fault data; for a certain harmonic component, calculating the average amplitude change degree of the harmonic component under different fault types; according to the average amplitude change degree corresponding to each harmonic component, selecting an important harmonic component influencing a fault and taking the important harmonic component as a fault feature; and constructing a sample set according to the fault feature data containing the fault positioning tag to obtain a data set. And training a fault positioning model by using the data set, so that fault positioning can be performed by using the fault positioning model. According to the method, a large amount of redundant information is removed, the fault positioning speed is improved, and the fault of the distributed power supply power distribution network is quickly positioned.

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network fault location, and in particular relates to a distributed power distribution network fault location method and device. Background Art

[0002] The access of distributed power sources changes the distribution of current and voltage in the distribution network. Under normal operation, distributed power sources can supplement the power demand of the distribution network and improve the reliability of power supply. However, when a phase-to-phase fault occurs, the distributed power source may cause an increase in short-circuit current, or even cause excessive current to cause equipment damage or malfunction of the protection device. Therefore, it is very important to locate the fault of the distributed power source after the fault occurs.

[0003] At present, there are many studies on short-circuit faults in distributed power distribution networks at home and abroad, including the use of matrix algorithms, ant colony algorithms, binary particle swarm algorithms and neural networks to locate faults in distributed power sources.

[0004] The matrix algorithm is used to consider the access of distributed power sources in the distribution network fault location, and dynamically construct the information matrix, but this algorithm has problems such as large variable dimension, complex calculation, and slow positioning speed. The ant colony algorithm has good fault tolerance, but the processing of information distortion problems depends on the accuracy of the measurement equipment, and it is difficult to handle large-scale information distortion. Combined with the idea of ​​regional division, the binary particle swarm algorithm is used to solve the fault location hierarchical model, but the characteristics of each layer model are not considered, resulting in a slow positioning speed. The use of neural networks for fault location is simple, but the input data dimension of the existing neural network positioning method is high, generally including a large number of fault features, and these large number of fault features may contain a large amount of redundant information; these large number of fault features undoubtedly make the calculation amount of fault location detection large, making the training process of the fault location model take a long time, and in the prediction stage, the fault location model needs to spend a lot of time to calculate to achieve fault location, which undoubtedly makes the fault location efficiency low. Summary of the invention

[0005] The object of the present invention is to provide a method and device for locating a fault in a distributed power distribution network, so as to solve the problem of low efficiency of the existing distributed power distribution network fault locating algorithm.

[0006] In order to solve the above technical problems, the present invention provides a method for locating a fault in a distributed power distribution network, the method comprising:

[0007] Obtaining fault feature data required for fault location using a fault location model in the distribution network, and inputting the trained fault location model to obtain the probability of each distributed power source in the distribution network failing, and locating the fault according to the probability;

[0008] The process of obtaining the data set used to train the fault location model includes:

[0009] Acquire fault data of different fault types in the distribution network, perform harmonic extraction on the fault data to obtain harmonic components of the fault data;

[0010] For a certain harmonic component, calculate the average amplitude variation of the harmonic component under different fault types; select the important harmonic components that affect the fault according to the average amplitude variation corresponding to each harmonic component and use them as fault characteristics;

[0011] A sample set is constructed according to the fault feature data including the fault location label, and a data set is obtained using the sample set.

[0012] Furthermore, the calculation formula for calculating the average amplitude variation degree of the subharmonic component under different fault types is:

[0013]

[0014] Among them, β ζ is the average amplitude change corresponding to the ζth harmonic component, Y ξζ is the amplitude corresponding to the ζth harmonic component of the ξth fault type, ι is the total number of fault types, κ is the harmonic magnitude, and

[0015] Further, the method of selecting the important harmonic components that affect the fault is as follows: the amplitude fluctuation contribution rate of each harmonic component is calculated according to the average amplitude change degree of each harmonic component, and the cumulative amplitude fluctuation rate of several harmonic components is calculated using the amplitude fluctuation contribution rate of each harmonic component; when the cumulative amplitude fluctuation rate is greater than or equal to the set important harmonic component threshold, the harmonic component corresponding to the cumulative amplitude harmonic fluctuation rate is identified as the important harmonic component that affects the fault;

[0016] The calculation formula of the amplitude fluctuation contribution rate of each harmonic component is:

[0017]

[0018] in, is the amplitude fluctuation contribution rate corresponding to the ζth harmonic component, β ζ is the average amplitude change corresponding to the ζth harmonic component, κ is the harmonic magnitude, and ζ∈[1,κ];

[0019] The calculation formula of the cumulative amplitude volatility is:

[0020]

[0021] in, is the cumulative amplitude fluctuation rate of k harmonic components, κ>k, A L To set the threshold of important harmonic components.

[0022] Furthermore, the method of obtaining the data set by using the sample set is: using a genetic algorithm and / or a RS algorithm to simplify the fault features in the sample set, and using the simplified sample set as the data set.

[0023] Furthermore, the fault location method according to the probability is:

[0024] If P l ≤ε1, the fault location cannot be determined;

[0025] like If a single distributed generation fails and P l The corresponding distributed power supply fails;

[0026] like If two distributed generation sources fail and P l and P sl The corresponding distributed power supply fails;

[0027] Among them, P l The maximum probability of a distributed generation failure, P sl The second largest value of the probability of a distributed generation failure, P ot Indicates the addition of P l and P sl , ε1 represents the first set threshold, ε2 represents the second set threshold, ε3 represents the third set threshold, and ε1<ε3<ε2.

[0028] Furthermore, the fault location model is a BN fault location model.

[0029] Further, the harmonic contents of the harmonic components of the calculated average amplitude variation degrees are all greater than or equal to a set harmonic content threshold.

[0030] To solve the above technical problems, the present invention further provides a distributed power distribution network fault locating device, comprising a processor, wherein the processor executes a computer program to implement the steps of the above distributed power distribution network fault locating method.

[0031] The beneficial effects are as follows: the present invention is an improved invention creation. The present invention no longer uses all fault information to locate the fault of the distributed power distribution network as in the fault location method in the prior art, but obtains fault data of different fault types in the distribution network, performs harmonic extraction on the fault data to obtain each harmonic component of the fault data; for a certain harmonic component, calculates the average amplitude change degree of the harmonic component under different fault types; selects the important harmonic components that affect the fault according to the average amplitude change degree corresponding to each harmonic component and uses it as the fault feature; constructs a sample set according to the fault feature data containing the fault location label and only retaining the important harmonic components, and obtains the data set using the sample set. The fault location model is trained using the data set, and the corresponding data in the fault data of the distributed power distribution network at the time to be tested is input into the trained fault location model to obtain the probability of each distributed power source failure, and then the fault location of the distributed power distribution network is completed according to the probability of failure. In the fault location method of the present invention, when locating the fault at the time to be tested, only the fault characteristics corresponding to the trained fault location model need to be obtained, which removes a large amount of redundant information, improves the speed of fault location, and realizes the rapid location of distributed power distribution network faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a flow chart of fault location of a distributed power distribution network according to an embodiment of the present invention;

[0033] Figure 2 is a GA attribute reduction flow chart of an embodiment of the present invention;

[0034] Figure 3 It is a flow chart of building a BN fault location model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] A method and device for locating faults in a distributed power distribution network according to an embodiment of the present invention obtains fault data of different fault types in the distribution network, performs harmonic extraction on the fault data to obtain each harmonic component of the fault data; for a certain harmonic component, calculates the average amplitude change degree of the harmonic component under different fault types; selects important harmonic components that affect the fault according to the average amplitude change degree corresponding to each harmonic component and uses them as fault features; constructs a sample set based on fault feature data that contains fault location labels and only retains important harmonic components, and obtains a data set using the sample set. The data set is used to train and test the fault location model, and the corresponding data in the distributed power distribution network fault data at the time to be tested is input into the trained fault location model to complete the fault location of the distributed power distribution network. The problem of low positioning efficiency of the existing distributed power distribution network fault location algorithm is solved.

[0036] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0037] Method Example:

[0038] A method for locating faults in a distributed power distribution network according to an embodiment of the present invention targets the characteristics of redundant fault nodes and characteristic attributes in a diagnostic model, performs harmonic extraction on the acquired fault data, simplifies the attributes of each harmonic component, and constructs a sample set based on fault characteristic data that contains fault location labels and retains only important harmonic components; discretizes the data of the sample set using the RS algorithm (rough set), and then uses a genetic algorithm (GA) to simplify the redundant characteristic attribute set to obtain a data set; uses the data set to train and test the constructed fault location model, inputs the corresponding fault characteristics in the fault information at the time to be tested into the trained fault location model, and completes the fault location of the distributed power supply. The process of the distributed power distribution network fault location algorithm of the present invention is as follows: Figure 1 As shown, the specific steps include:

[0039] Step 1: Obtain various fault data in the distributed generation distribution network, and perform feature extraction, i.e. simplification, on the acquired fault data to obtain a sample set S of fault data.

[0040] Among them, a variety of short-circuit fault symptom modes can be set according to the simulation model to obtain a variety of fault data in the distributed power distribution network.

[0041] (1) According to the acquired fault data of various fault types (specifically voltage and current), the harmonic components of different fault data are extracted respectively.

[0042] Short-circuit faults are usually accompanied by current distortion. Therefore, the corresponding short-circuit current is selected as the fault variable, and the short-circuit data is processed in the form of Fourier transform, and the amplitude fluctuation method is used for feature extraction. According to different types of fault harmonics (such as unidirectional grounding, two-phase short circuit, direct current directly connected to the distribution network and other short-circuit fault types), α is extracted respectively. i =1 harmonic component. The extraction method is:

[0043]

[0044] in, Indicates the i-th harmonic content in the current amplitude α i The content rate (i.e. setting the harmonic content threshold). ci When it is greater than or equal to 20%, α iThe value is 1, otherwise it is 0. Since high-dimensional harmonics will have a great impact on the accurate classification of fault components, in order to obtain accurate information, it is necessary to simplify the fault components, so only α is extracted. i =1. ci The value of is 20%, and can be set according to actual needs as other implementation methods.

[0045] (2) Based on the harmonic components extracted in step (1), for a certain harmonic component, calculate the average amplitude change degree of the harmonic component under different fault types; select the important harmonic components that affect the fault from the average amplitude change degree corresponding to each harmonic component and use them as fault features. Specifically:

[0046] ① Calculate the average amplitude variation of the same harmonics of different fault types β ζ :

[0047]

[0048] Among them, β ζ is the average amplitude change corresponding to the ζth harmonic component, Y ξζ is the amplitude corresponding to the ζth harmonic component of the ξth fault type, ι is the total number of fault types studied, κ is the harmonic magnitude, and

[0049] ② The β calculated in step ① ζ Arrange in order from large to small: β1>β2>β3>…β ζ >0, in order to facilitate subsequent calculations and avoid missing data. Calculate β ζ The cumulative amplitude volatility is calculated as follows:

[0050]

[0051] in, is the amplitude fluctuation contribution rate corresponding to the ζth harmonic, represents the cumulative amplitude volatility, A L In order to meet the requirements of setting the important harmonic component threshold, and κ>k, A in this embodiment L =75%, as other implementation methods, can be set according to actual needs.

[0052] According to the set important harmonic component threshold, the harmonic component that meets the requirements is selected as the fault feature, that is, the harmonic component that satisfies formula (4) is used as the fault feature, and then the fault feature data containing the fault location label is used to construct the sample set S.

[0053] Step 2: Reduce the sample set S obtained in step 1 to obtain the data set S*.

[0054] The RS algorithm is used for preliminary simplification to transform the sample set S into a discrete data table, and then GA (Genetic Algorithm) is used for attribute simplification to obtain the data set S*. Figure 2 As shown, the specific simplification process is as follows:

[0055] Step S1: Initialize the population P(T). Use a random method to perform binary encoding on individuals;

[0056] Step S2: Evaluate the population. Use the fitness function to determine the fitness value Φ(x) of each chromosome;

[0057] Step S3: Evaluate the quality of the chromosome based on the calculated Φ(x). If yes, terminate the algorithm and output the result directly. If no, continue the genetic operation.

[0058] Step S4: Use the roulette wheel selection method to perform the selection operation of the genetic algorithm, then implement the crossover operation of the genetic algorithm according to the crossover probability, and finally implement the crossover operation of the genetic algorithm according to the mutation probability P m Implement the mutation operation of the genetic algorithm, and start the next round of evolution after all operations are completed;

[0059] Step S5: Decode the optimal individual. The decoded optimal individual is the output optimal solution.

[0060] Since there are many types of short-circuit faults in distributed power distribution networks, it will cause variable redundancy and complex problems. Therefore, a heuristic approach can be used to reduce the search space. The basic idea of ​​GA is to iteratively select, cross, and mutate individuals in the encoding space based on the fitness evaluation function, and finally obtain the global optimal solution.

[0061] Step 3: Build a fault location model.

[0062] The fault location model of this embodiment is a BN fault location model, which specifically includes: BN structure, BN parameters and BN reasoning.

[0063] BN fault diagnosis model:

[0064] BN (Batch Normalization) is an uncertain causal association model based on probabilistic reasoning. It consists of network structure and parameters and can represent complex relationships in diagnostic models. Figure 3As shown in Figure 1, the BN fault location model consists of three parts: BN structure, BN parameters and BN reasoning. Among them, the network structure can be obtained through expert knowledge learning or data learning. For the fault location model, it can be known from expert knowledge that the typical network structure consists of a fault state layer and a fault feature layer.

[0065] Parameter learning and inference algorithm:

[0066] The BN algorithm mainly consists of three parts: model structure learning, parameter learning and fault reasoning algorithm. Specifically:

[0067] ① Structural learning represents the relationship between nodes, that is, whether the mutual dependence is strong or weak. According to different Bayesian network toolboxes, it can be divided into three methods: K2 algorithm, MCMC algorithm and MATLAB implementation based on BDAGL toolbox.

[0068] ② Parameter learning refers to determining the conditional probability set of each node and analyzing the quantitative table of dependencies. This embodiment adopts the parameter learning method of Maximum Likelihood Estimation (MLE), which determines the state of the node by determining the parameter θ (probability value of the node) that maximizes the likelihood function L(θ, X).

[0069] ③BN inference algorithm uses BN structure and CPT to calculate the posterior probability of certain nodes under given evidence. i , X i There are j values, respectively (x i1 , x i2 ,...,x ij ), assuming that the given i Evidence information of all nodes except Calculate X i The formula for the posterior probability is:

[0070]

[0071] Among them, the larger the posterior probability of a node state is, the higher the possibility of the state occurring is.

[0072] Step 4: Divide the data set S* obtained after simplification in step 2 into a training set and a test set, and train and test the BN fault location model.

[0073] Step 5: Input the fault feature data required for fault location using the fault location model in the distribution network into the trained BN fault location model to obtain the probability of failure of each distributed power source in the distribution network, and locate the fault according to the probability of failure of each distributed power source. The specific method of fault location based on probability is:

[0074] If P l ≤ε1, the fault location cannot be determined;

[0075] like It is considered a single point failure, that is, a single distributed power generation site fails, specifically P l The corresponding distributed power generation site fails;

[0076] like It is judged as a double-point fault, that is, two distributed power generation sites fail, specifically P l and P sl The corresponding distributed power generation site fails.

[0077] Among them, P l Represents the maximum failure probability value, P sl Represents the second largest failure probability value, P ot Indicates the addition of P l and P sl ε1 is the first set threshold, ε2 is the second set threshold, and ε3 is the third set threshold; and ε1<ε3<ε2, P l >P sl It should be noted that, in general, as long as one or two distributed power sources fail in the entire distribution network, the circuit breaker will trip or the backup power source will be started, so in this step, it is only necessary to judge whether it is a single-point failure or a double-point failure; moreover, since it is only necessary to judge whether it is a single-point failure or a double-point failure, only the above-mentioned situations are defined, and other value ranges are not within the scope of discussion of the scenarios faced by the present invention.

[0078] The distributed power distribution network fault location method of the present invention has a simple and convenient structure, effectively reduces the workload of data calculation, reduces the redundant information of fault characteristics and the data feature dimension, so that a concise and accurate fault diagnosis decision rule can be directly obtained; and the continuous attributes of massive fault data can be discretized, and conditional attributes with high importance can be extracted for attribute simplification, so as to realize efficient fault location.

[0079] Device Example:

[0080] A distributed power distribution network fault location device of the present invention includes a memory, a processor and an internal bus, and the processor and the memory communicate and exchange data with each other through the internal bus. The memory includes at least one software function module stored in the memory, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing a distributed power distribution network fault location method introduced in the method embodiment of the present invention. The principle, implementation process, and achievable effects of the method have been fully introduced in the method embodiment and will not be repeated here.

Claims

1. A method for locating faults in a distributed power distribution network, characterized in that: The method includes: Obtaining fault feature data required for fault location using a fault location model in the distribution network, and inputting the trained fault location model to obtain the probability of each distributed power source in the distribution network failing, and locating the fault according to the probability; The process of obtaining the data set used to train the fault location model includes: Acquire fault data of different fault types in the distribution network, perform harmonic extraction on the fault data to obtain harmonic components of the fault data; For a certain harmonic component, calculate the average amplitude variation of the harmonic component under different fault types; select the important harmonic components that affect the fault according to the average amplitude variation corresponding to each harmonic component and use them as fault characteristics; A sample set is constructed according to the fault feature data including the fault location label, and a data set is obtained using the sample set.

2. The method for locating a fault in a distributed power distribution network according to claim 1, characterized in that: The calculation formula for calculating the average amplitude variation degree of the subharmonic component under different fault types is: Among them, β ζ is the average amplitude change corresponding to the ζth harmonic component, Y ξζ is the amplitude corresponding to the ζth harmonic component of the ξth fault type, ι is the total number of fault types, κ is the harmonic magnitude, and 3. The method for locating a fault in a distributed power distribution network according to claim 1, characterized in that: The way to select the important harmonic components that affect the fault is: The amplitude fluctuation contribution rate of each harmonic component is calculated according to the average amplitude variation degree of each harmonic component, and the cumulative amplitude fluctuation rate of several harmonic components is calculated using the amplitude fluctuation contribution rate of each harmonic component; When the cumulative amplitude fluctuation rate is greater than or equal to the set important harmonic component threshold, the harmonic component corresponding to the cumulative amplitude harmonic fluctuation rate is identified as an important harmonic component affecting the fault; The calculation formula of the amplitude fluctuation contribution rate of each harmonic component is: in, is the amplitude fluctuation contribution rate corresponding to the ζth harmonic component, β ζ is the average amplitude change corresponding to the ζth harmonic component, κ is the harmonic magnitude, and ζ∈[1,κ]; The calculation formula of the cumulative amplitude volatility is: in, is the cumulative amplitude fluctuation rate of k harmonic components, κ>k, A L To set the threshold of important harmonic components.

4. The method for locating a fault in a distributed power distribution network according to claim 1, characterized in that: The method of obtaining the data set by using the sample set is: using a genetic algorithm and / or a RS algorithm to simplify the fault features in the sample set, and using the simplified sample set as the data set.

5. The method for locating a fault in a distributed power distribution network according to claim 1, characterized in that: The fault location method based on the probability is: If P l ≤ε1, the fault location cannot be determined; like If a single distributed generation fails and P l The corresponding distributed power supply fails; like If two distributed generation sources fail and P l and P sl The corresponding distributed power supply fails; Among them, P l The maximum probability of a distributed generation failure, P sl The second largest value of the probability of a distributed generation failure, P ot Indicates the addition of P l and P sl , ε1 represents the first set threshold, ε2 represents the second set threshold, ε3 represents the third set threshold, and ε1<ε3<ε2.

6. The method for locating a fault in a distributed power distribution network according to claim 1, characterized in that: The fault location model is a BN fault location model.

7. The method for locating a fault in a distributed power distribution network according to claim 1, characterized in that: The harmonic contents of the harmonic components of the calculated average amplitude variation degree are all greater than or equal to the set harmonic content threshold.

8. A distributed power distribution network fault location device, comprising a processor, characterized in that: The processor executes a computer program to implement the steps of the distributed power distribution network fault location method according to any one of claims 1 to 7.

Images