CN117743958A - Photovoltaic array fault identification method and device, electronic equipment - Google Patents

Abstract

The invention relates to the technical field of photovoltaic power generation, and discloses a photovoltaic array fault identification method and device, and electronic equipment. The method includes: establishing a topological structure of a grid-connected photovoltaic power generation system based on power grid operating parameters; simulating the preset conditions of the photovoltaic array based on the topological structure. Under abnormal operating conditions, the first output characteristics of the photovoltaic array are obtained; by comparing the first output characteristics with the second output characteristics of the photovoltaic array under normal operating conditions, a three-dimensional fault characteristic quantity group is generated; SVM support is built based on the three-dimensional fault characteristic quantity group Vector machine fault identification model, and optimize the penalty factor and kernel function parameters of the SVM fault identification model to obtain the target SVM fault identification model; based on the target SVM fault identification model, the validity of the target three-dimensional fault characteristic quantity is verified to ensure the photovoltaic array Perform fault identification. The invention can improve the accuracy of photovoltaic array fault identification.

Description

Photovoltaic array fault identification method and device, and electronic equipment

Technical Field

The present invention relates to the technical field of photovoltaic power generation, and in particular to a photovoltaic array fault identification method and device, and electronic equipment.

Background Art

Since the "dual carbon" goal was proposed, suburban distributed photovoltaics have grown rapidly, and photovoltaic-related industries have grown stronger and stronger. However, photovoltaic arrays are prone to various faults due to the complex working environment, which in turn affects the operational reliability and energy conversion efficiency of photovoltaic systems. How to introduce mature and efficient artificial intelligence algorithms into the field of photovoltaic array fault diagnosis has gradually attracted the attention of scholars and researchers.

As an important component of new energy, photovoltaic systems operate in complex and random environments for a long time. When the photovoltaic module battery cells are oxidized inside and run for a long time regardless of their lifespan, resulting in different degrees of abnormal aging, the array output voltage and current parameters decrease to varying degrees. When the photovoltaic array is in a state of partial shadow to varying degrees, the array output characteristic maximum power point voltage and maximum power point current are affected, resulting in distortion of the photovoltaic array output power, affecting output stability and reducing power generation efficiency.

There are a large number of photovoltaic arrays connected to the power grid. Failure of components or arrays will cause the entire photovoltaic system to be unable to operate safely and efficiently. Therefore, there is an urgent need for technical personnel in this field to provide a photovoltaic array fault identification method with a simple process and fast identification speed.

Machine learning methods in photovoltaic array intelligent detection methods are mostly used in the field of array fault identification. The extracted fault feature data are input into the machine learning algorithm for model training to obtain an intelligent fault identification model, thereby realizing fast and accurate intelligent learning classification of photovoltaic array fault modes.

At present, the commonly used machine learning methods are mainly neural networks and various improved algorithms of neural networks. However, it is difficult to determine the number of nodes in the middle layer of the neural network algorithm, which makes the neural network prone to local optimal problems during learning and training, and the calculation process is complicated, resulting in a decrease in the accuracy of photovoltaic array fault identification results.

Summary of the invention

The purpose of the embodiments of the present invention is to provide a photovoltaic array fault identification method and device, and electronic equipment, which can solve the problem of reduced accuracy of identification results in existing photovoltaic array fault identification solutions.

In order to solve the above technical problems, the present invention provides the following technical solutions:

An embodiment of the present invention provides a photovoltaic array fault identification method, comprising:

Based on the grid operation parameters, establish the topology of the grid-connected photovoltaic power generation system;

Simulating the operation status of the photovoltaic array under a preset abnormal condition based on the topological structure of the grid-connected photovoltaic power generation system to obtain a first output characteristic of the photovoltaic array;

Compare the first output characteristic with the second output characteristic of the photovoltaic array under normal operating conditions to generate a three-dimensional fault feature quantity group;

Building an SVM fault recognition model according to the three-dimensional fault feature quantity group, and optimizing the penalty factor and kernel function parameters of the SVM fault recognition model to obtain a target SVM fault recognition model;

The target three-dimensional fault feature quantity is verified for effectiveness based on the target SVM fault identification model to identify faults of the photovoltaic array.

Optionally, the step of simulating the operating condition of the photovoltaic array under a preset abnormal condition based on the topological structure of the grid-connected photovoltaic power generation system to obtain the first output characteristic of the photovoltaic array includes:

Based on the topological structure and the preset simulation platform, a simulation model is established;

The simulation model is subjected to photovoltaic array failure simulation under preset abnormal conditions to obtain the first output characteristics of the photovoltaic array corresponding to each abnormal condition, wherein the preset abnormal conditions include: abnormal aging of the photovoltaic array to varying degrees, and local shadows of varying degrees in the photovoltaic array.

Optionally, the photovoltaic array having different degrees of abnormal aging states includes: the first group of strings in the photovoltaic array having a first degree of aging, and the second group of strings being normal; the first group of strings in the photovoltaic array having a second degree of aging, and the second group of strings being normal; the first group of strings and the second group of strings in the photovoltaic array being both aged;

The states where the photovoltaic array has different degrees of local shadows include: the irradiance of a single string of photovoltaic components in the photovoltaic array is 0, and the irradiance of multiple strings of photovoltaic components in the photovoltaic array is 0.

Optionally, the step of comparing the first output characteristic with a second output characteristic of the photovoltaic array in a normal operating state to generate a three-dimensional fault feature quantity group includes:

For each degree of abnormal aging state, analyzing the output characteristics corresponding to the abnormal aging state, and determining the maximum power point voltage and maximum power point current of the photovoltaic array;

Comparing the maximum power point voltage and the maximum power point current corresponding to each degree of abnormal aging state with the maximum power point voltage and the maximum power point current under the normal operating state of the photovoltaic array to obtain a second output characteristic under the abnormal aging state;

For each degree of local shadow state, the output characteristics corresponding to the local shadow state are analyzed to determine the number of inflection points of the U-I curve of the photovoltaic array;

The number of inflection points of each U-I curve step is used as the second output characteristic under the local shadow state.

Optionally, the step of building an SVM support vector machine fault recognition model according to the three-dimensional fault feature quantity group, and optimizing the penalty factor and kernel function parameters of the SVM fault recognition model to obtain a target SVM fault recognition model includes:

Performing deviation normalization processing on the three-dimensional fault feature quantity group to obtain a target three-dimensional fault feature quantity group; wherein the target three-dimensional fault feature quantity group includes multiple groups of three-dimensional fault feature quantities composed of the maximum power point voltage and maximum power point current of the photovoltaic array and the number of step inflection points of the U-I curve;

Dividing the fault feature quantities in the target three-dimensional fault feature quantity group into a training set and a test set;

An initial SVM fault recognition model is constructed, and the penalty factor and kernel function parameters of the initial SVM fault recognition model are iteratively optimized according to the training set until an iterative optimization cutoff condition is met, thereby obtaining a target SVM fault recognition model.

The embodiment of the present invention further provides a photovoltaic array fault identification device, comprising:

Establishing a module for establishing a grid-connected photovoltaic power generation system topology based on grid operation parameters;

A simulation module, used to simulate the operation status of the photovoltaic array under a preset abnormal condition based on the topological structure of the grid-connected photovoltaic power generation system, and obtain a first output characteristic of the photovoltaic array;

A generating module, configured to compare the first output characteristic with a second output characteristic of the photovoltaic array under normal operating conditions, and generate a three-dimensional fault feature quantity group;

A model optimization module, used to build an SVM fault recognition model according to the three-dimensional fault feature quantity group, and optimize the penalty factor and kernel function parameters of the SVM fault recognition model to obtain a target SVM fault recognition model;

The identification module is used to verify the validity of the target three-dimensional fault feature quantity based on the target SVM fault identification model to identify the fault of the photovoltaic array.

Optionally, the simulation module includes:

The first submodule is used to establish a simulation model based on the topological structure of the grid-connected photovoltaic power generation system and a preset simulation platform;

The second submodule is used to simulate the photovoltaic array failure under preset abnormal conditions for the simulation model, and obtain the first output characteristics of the photovoltaic array corresponding to each abnormal condition, wherein the preset abnormal conditions include: the photovoltaic array has different degrees of abnormal aging status, and the photovoltaic array has different degrees of local shadow status.

Optionally, the photovoltaic array having different degrees of abnormal aging states includes: the first group of strings in the photovoltaic array having a first degree of aging, and the second group of strings being normal; the first group of strings in the photovoltaic array having a second degree of aging, and the second group of strings being normal; the first group of strings and the second group of strings in the photovoltaic array being both aged;

The states where the photovoltaic array has different degrees of local shadows include: the irradiance of a single string of photovoltaic components in the photovoltaic array is 0, and the irradiance of multiple strings of photovoltaic components in the photovoltaic array is 0.

Optionally, the generating module includes:

The third submodule is used to analyze the output characteristics corresponding to each degree of abnormal aging state, and determine the maximum power point voltage and maximum power point current of the photovoltaic array;

A fourth submodule is used to compare the maximum power point voltage and the maximum power point current corresponding to each degree of abnormal aging state with the maximum power point voltage and the maximum power point current under the normal operating state of the photovoltaic array to obtain a second output characteristic under the abnormal aging state;

A fifth submodule is used to analyze the output characteristics corresponding to each degree of local shadow state, and determine the number of inflection points of the U-I curve of the photovoltaic array;

The sixth submodule is used to use the number of inflection points of each U-I curve step as the second output characteristic under the local shadow state.

Optionally, the model optimization module includes:

The seventh submodule is used to perform deviation normalization processing on the three-dimensional fault characteristic quantity group to obtain a target three-dimensional fault characteristic quantity group; wherein the target three-dimensional fault characteristic quantity group includes multiple groups of three-dimensional fault characteristic quantities composed of the maximum power point voltage and maximum power point current of the photovoltaic array and the number of step inflection points of the U-I curve;

An eighth submodule, configured to divide the fault feature quantities in the target three-dimensional fault feature quantity group into a training set and a test set;

The ninth submodule is used to build an initial SVM fault recognition model, and iteratively optimize the penalty factor and kernel function parameters of the initial SVM fault recognition model according to the training set until the iterative optimization cutoff condition is met to obtain a target SVM fault recognition model.

An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction implements the steps of any one of the above-mentioned photovoltaic array fault identification methods when executed by the processor.

An embodiment of the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of any one of the above photovoltaic array fault identification methods are implemented.

The photovoltaic array fault identification scheme provided by the embodiment of the present invention establishes the topological structure of the grid-connected photovoltaic power generation system based on the grid operation parameters; simulates the operation status of the photovoltaic array under the preset abnormal conditions based on the topological structure to obtain the first output characteristic of the photovoltaic array; compares the first output characteristic with the second output characteristic under the normal operation state of the photovoltaic array to generate a three-dimensional fault feature quantity group; builds an SVM fault identification model according to the three-dimensional fault feature quantity group, and optimizes the penalty factor and kernel function parameters of the SVM fault identification model to obtain a target SVM fault identification model; verifies the effectiveness of the target three-dimensional fault feature quantity based on the target SVM fault identification model to identify the fault of the photovoltaic array. In this photovoltaic array fault identification scheme, the SVM estimates the expected risk based on the principle of structural risk minimization, which can overcome the overfitting problem in the model training process, so that the fault identification model based on the SVM has good classification and discrimination capabilities; in addition, the SVM fault identification model is trained by the three-dimensional fault feature quantity group, which can improve the accuracy of the SVM fault identification model in identifying photovoltaic array faults.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

FIG1 is a schematic diagram showing a flow chart of a photovoltaic array fault identification method according to an embodiment of the present application;

FIG2 is a topological diagram of a grid-connected photovoltaic power generation system according to an embodiment of the present application;

FIG3 (a) and (b) show the output characteristics of the photovoltaic array under three abnormal aging faults according to the embodiment of the present application;

FIG4 (a) and (b) show the output characteristics of the photovoltaic array under two local shadow faults according to an embodiment of the present application;

FIG5 is a flowchart showing fault identification of a photovoltaic array based on an improved support vector fault identification model according to an embodiment of the present application;

FIG6 is a schematic diagram of a support vector machine binary classification according to an embodiment of the present application;

7 is a search flow chart showing a cuckoo search algorithm according to an embodiment of the present application;

FIG8 is a fitness evolution curve diagram showing an embodiment of the present application;

FIG9 is a diagram showing the fault recognition accuracy of the improved SVM fault recognition model for the sample training set according to an embodiment of the present application;

FIG10 is a diagram showing the fault recognition accuracy of the improved SVM fault recognition model for the sample test set according to an embodiment of the present application;

FIG11 is a block diagram showing a photovoltaic array fault identification device according to an embodiment of the present application;

FIG. 12 is a block diagram showing the structure of an electronic device according to an embodiment of the present application.

DETAILED DESCRIPTION

The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

The photovoltaic array fault identification solution provided by the embodiment of the present application is described in detail below through specific embodiments and application scenarios in conjunction with the accompanying drawings.

As shown in FIG. 1 , the photovoltaic array fault identification method according to the embodiment of the present application includes the following steps:

Step 101: Based on the grid operation parameters, a grid-connected photovoltaic power generation system topology is established.

In the embodiment of the present application, the topology of the grid-connected photovoltaic power generation system is established by combining the actual working conditions of the photovoltaic array in a complex environment and taking into account the operating parameters of the power grid.

Step 102: Simulate the operation status of the photovoltaic array under a preset abnormal condition based on the topological structure of the grid-connected photovoltaic power generation system to obtain a first output characteristic of the photovoltaic array.

Optionally, based on the topological structure of the grid-connected photovoltaic power generation system, the operation status of the photovoltaic array under a preset abnormal condition is simulated to obtain the first output characteristic of the photovoltaic array as follows:

Firstly, a simulation model is established based on the topological structure of the grid-connected photovoltaic power generation system and the preset simulation platform;

The preset simulation platform may include but is not limited to: Matlab/Simulink simulation platform.

Secondly, a photovoltaic array fault simulation under a preset abnormal condition is performed on the simulation model to obtain a first output characteristic of the photovoltaic array corresponding to each abnormal condition.

The preset abnormal conditions include: the photovoltaic array experiencing abnormal aging to varying degrees, and the photovoltaic array having local shadows to varying degrees.

In the actual implementation process, a simulation model is established to simulate photovoltaic array failures under complex working conditions, and the first output characteristics of the photovoltaic array are obtained when different degrees of abnormal aging and different degrees of local shadow failure occur.

The photovoltaic array undergoes different degrees of abnormal aging conditions, including: the first group of strings in the photovoltaic array undergoes first degree aging, and the second group of strings is normal; the first group of strings in the photovoltaic array undergoes second degree aging, and the second group of strings is normal; both the first group of strings in the photovoltaic array are aged; the photovoltaic array has different degrees of local shadow conditions, including: the irradiance of a single group of photovoltaic components in the photovoltaic array is 0, and the irradiance of multiple groups of photovoltaic components in the photovoltaic array is 0.

When an abnormal aging fault occurs in a component in the array, the internal resistance of the faulty component increases, that is, in the equivalent circuit of the component, the series resistance increases. Therefore, by changing the series resistance value of the strings in the array, the aging fault of the array component can be simulated. Local shadow faults are achieved by setting the irradiance of the photovoltaic module.

It should be noted that the above are only exemplary simulations of the specific states included in the normal aging state and the state of local shadows of different degrees. In the actual implementation process, technical personnel in this field can flexibly set the specific states included in the above two types of states according to actual needs, and no specific restrictions are made here.

Step 103: Compare the first output characteristic with the second output characteristic under normal operation of the photovoltaic array to generate a three-dimensional fault feature quantity group.

An optional method of comparing the first output characteristic with the second output characteristic under normal operation of the photovoltaic array to generate a three-dimensional fault feature quantity group may be as follows:

For each degree of abnormal aging state, the output characteristics corresponding to the abnormal aging state are analyzed to determine the maximum power point voltage and the maximum power point current of the photovoltaic array; the maximum power point voltage and the maximum power point current corresponding to each degree of abnormal aging state are compared with the maximum power point voltage and the maximum power point current of the photovoltaic array under normal operating conditions to obtain the second output characteristics under the abnormal aging state; for each degree of local shadow state, the output characteristics corresponding to the local shadow state are analyzed to determine the number of U-I curve step inflection points of the photovoltaic array; the number of each U-I curve step inflection point is used as the second output characteristic under the local shadow state.

In the actual implementation process, when analyzing the first output characteristics of the photovoltaic array, for different degrees of abnormal aging faults, the array's U _m (i.e., maximum power point voltage) and _Im (i.e., maximum power point current) gradually decrease with the deepening of aging, so Um and Im are selected to identify abnormal aging faults. For the two degrees of minor and severe local shadow faults, the number of step inflection points of the photovoltaic array UI curve is introduced, and the effectiveness of U _m and _Im as feature quantities to distinguish the two degrees of local shadow is verified. It is proposed that the three-dimensional fault feature quantity is composed of the photovoltaic array maximum power point voltage U _m , the maximum power point current _Im , and the number of step inflection points of the UI curve.

Step 104: construct an SVM fault recognition model according to the three-dimensional fault feature quantity group, and optimize the penalty factor and kernel function parameters of the SVM fault recognition model to obtain a target SVM fault recognition model.

Support vector machine (SVM) is a binary classification model. Its basic model is a linear classifier with the largest margin defined in the feature space. The largest margin distinguishes it from the perceptron. SVM also includes kernel techniques, which makes it an essentially nonlinear classifier.

An optional method of building an SVM fault recognition model based on the three-dimensional fault feature quantity group and optimizing the penalty factor and kernel function parameters of the SVM fault recognition model to obtain a target SVM fault recognition model may be as follows:

The three-dimensional fault characteristic quantity group is subjected to deviation normalization processing to obtain a target three-dimensional fault characteristic quantity group; wherein the target three-dimensional fault characteristic quantity group includes multiple groups of three-dimensional fault characteristic quantities composed of maximum power point voltage and maximum power point current of the photovoltaic array and the number of step inflection points of the U-I curve;

The fault feature quantities in the target three-dimensional fault feature quantity group are divided into a training set and a test set;

An initial SVM fault recognition model is built, and the penalty factor and kernel function parameter γ of the initial SVM fault recognition model are iteratively optimized according to the training set until the iterative optimization cutoff condition is met to obtain the target SVM fault recognition model.

Step 105: verifying the validity of the target three-dimensional fault feature quantity based on the target SVM fault identification model to identify faults of the photovoltaic array.

The target three-dimensional fault feature quantity may be one or more three-dimensional fault features selected from the three-dimensional fault feature quantity group, or may be one or more three-dimensional fault features in a test set divided from the target three-dimensional fault feature quantity group.

The photovoltaic array fault identification method provided in the embodiment of the present application establishes the topological structure of the grid-connected photovoltaic power generation system based on the grid operation parameters; simulates the operation status of the photovoltaic array under the preset abnormal conditions based on the topological structure to obtain the first output characteristic of the photovoltaic array; compares the first output characteristic with the second output characteristic under the normal operation state of the photovoltaic array to generate a three-dimensional fault feature quantity group; builds an SVM fault identification model based on the three-dimensional fault feature quantity group, and optimizes the penalty factor and kernel function parameters of the SVM fault identification model to obtain a target SVM fault identification model; verifies the effectiveness of the target three-dimensional fault feature quantity based on the SVM fault identification model to identify the fault of the photovoltaic array. In this photovoltaic array fault identification method, SVM estimates the expected risk based on the principle of structural risk minimization, which can overcome the overfitting problem in the model training process, so that the fault identification model based on the SVM has good classification and discrimination capabilities; in addition, the SVM fault identification model is trained by the three-dimensional fault feature quantity group, which can improve the accuracy of the SVM fault identification model in identifying photovoltaic array faults.

The photovoltaic array fault identification method provided in the embodiment of the present application is described below with a specific example.

In this specific example, the topology of the grid-connected photovoltaic power generation system is established by combining the actual working conditions of the photovoltaic array under complex environments and considering the grid operation parameters. The output characteristics of the photovoltaic array are analyzed. For different degrees of abnormal aging faults, the U _m and _Im of the array gradually decrease with the deepening of the aging degree. Therefore, U _m and _Im are selected to identify abnormal aging faults. For the two degrees of minor and severe local shadow faults, the number of step inflection points of the photovoltaic array UI curve is introduced, and the effectiveness of U _m and _Im as feature quantities to distinguish the two degrees of local shadows is verified. It is proposed that the three-dimensional fault feature quantity is composed of the maximum power point voltage U _m of the photovoltaic array, the maximum power point current _Im , and the number of step inflection points of the UI curve.

Then, based on the three-dimensional fault feature quantity composed of U _m , I _m , and the number of step inflection points of the UI curve, a SVM (support vector machine) fault recognition model is built, and the penalty factor and kernel function parameter γ of the SVM fault recognition model are optimized. Based on the improved SVM fault recognition model, namely the target SVM fault recognition model, the effectiveness of the selected three-dimensional fault feature quantity is verified, thereby achieving accurate identification of photovoltaic array faults.

This specific example includes the following processes S1-S3:

S1: Combined with the actual working conditions of the photovoltaic array in a complex environment, considering the grid operation parameters, the topological structure of the grid-connected photovoltaic power generation system is established using the Matlab/Simulink simulation platform. The topological structure diagram of the constructed grid-connected photovoltaic power generation system is shown in Figure 2. _L1 and _L2 are filter inductors, _Cdc and C are filter capacitors, _T1 ~ _T6 are switch tubes of the photovoltaic grid-connected inverter circuit, and R is a passive damping resistor.

In order to meet the stability requirements of the grid-connected system and reduce the harmonics of the grid-connected current, a passive damping design is used to suppress the resonance problem. The selected filter circuit is determined to be an LCL filter with a passive damping resistor R in series in the filter capacitor C branch.

An exemplary photovoltaic array is composed of 9×8 photovoltaic modules connected in series-parallel (SP) mode. The key parameters of the established photovoltaic system simulation model are shown in Table 1:

Table 1 Key parameter settings of photovoltaic system

S2: Based on the simulation model, simulate the faults of photovoltaic arrays under different working conditions, obtain the output characteristics of the photovoltaic arrays under abnormal aging, partial shadow and other operating conditions, i.e., the first output characteristics, analyze the output characteristics of the photovoltaic arrays, and extract multi-dimensional fault features for abnormal aging and partial shadow faults of different degrees. The fault types of photovoltaic arrays simulated in this specific example are shown in Table 2.

Table 2 Working status description

1) An exemplary abnormal aging fault feature extraction method is as follows:

When abnormal aging failure occurs in the components of the photovoltaic array, the internal resistance of the faulty component increases, that is, in the equivalent circuit of the component, the series resistance increases. Therefore, by changing the series resistance _value Rs of the strings in the array, the aging failure of the array components can be simulated.

The simulation simulates the abnormal aging of array strings, and designs three different fault conditions: slight aging: string 1 is aged, string 2 is normal; general aging: string 1 is aged more, string 2 is normal; severe aging: string 1 and string 2 are aged to different degrees. The simulated aging resistance value settings of the photovoltaic array are shown in Table 3:

Table 3 Abnormal aging fault array string series resistance setting

The output characteristics of the photovoltaic array under three abnormal aging faults obtained by simulation are shown in Figure 3 (a) and (b). When the strings in the photovoltaic array have different degrees of abnormal aging faults, the U _oc and I _sc of the array are almost unchanged, but as the aging degree deepens, the U _m and I _m of the array gradually decrease, and the power of the entire system also decreases with the deepening of the aging degree.

2) An exemplary local shadow fault feature extraction method is as follows:

The local shadow fault is realized by setting the irradiance of the photovoltaic module. In this specific example, two different degrees of local shadow faults are considered, namely slight shading and severe shading, as shown in Table 4:

Table 4 Local shadow fault degree

The simulation results, i.e., the output characteristics of the photovoltaic array under local shadow faults, are shown in Figure 4 (a) and (b), which compare and analyze the output characteristics of the photovoltaic array under local shadow faults of different degrees and under normal conditions. When the string is normal, the U-I curve of the photovoltaic array output is relatively smooth. When two local shadow faults occur, the U-I curve of the photovoltaic array becomes a step-shaped curve with an inflection point, and the larger the shadow area, the steeper the step.

When a local shadow fault occurs in the photovoltaic array, a step inflection point appears in the UI curve, but not in normal conditions. Therefore, the number of step inflection points of the photovoltaic array UI curve is selected as the characteristic quantity to identify local shadow faults and distinguish other states.

Considering that the number of inflection points of the UI curve cannot effectively distinguish different degrees of local shadow faults, based on the analysis of the changes in voltage and current parameters under different shadow shading areas, it can be seen that the changes in U _m and I _m of different degrees of local shadow faults of the photovoltaic array are large. Therefore, U _m and I _m can be used as characteristic quantities to effectively distinguish the two types of local shadow faults.

S3: Based on the three-dimensional fault feature quantity composed of U _m , Im , _and the number of inflection points of the U - I curve steps, the SVM fault recognition model is improved, and the fault recognition of the photovoltaic array is performed based on the improved SVM fault recognition model.

Among them, the fault identification process of the photovoltaic array based on the improved SVM fault identification model is shown in Figure 5.

1) Based on the constructed 9×8 photovoltaic array simulation model, the irradiance variation range is set to 200~1200W/m ^2. The photovoltaic array is in six types of working conditions, including normal, three different degrees of abnormal aging, and two different degrees of local shadow. 90 sets of sample data are obtained for each of them, totaling 540 sets. 360 sets of sample data in each operating state are used as training samples, and 180 sets of sample data are used as test samples. The sample data is the first output characteristic.

2) Preprocess the sample data to eliminate sample data distortion, noise and distortion caused in the acquisition stage. The data preprocessing adopts the deviation normalization method.

Since the units of the collected sample data are not uniform and the values of the parameters are quite different, if the unprocessed sample data is directly input into the SVM fault recognition model, it will affect the convergence and accuracy of the classification, thereby reducing the efficiency of fault recognition and being unfavorable for classification. Therefore, the collected fault sample data must be normalized in terms of units and values so that the three parameters are consistent in order of magnitude and units.

Deviation normalization uses the difference between the maximum and minimum values of a parameter in the sample data set as the deviation value, and takes the ratio of the difference between the value of each parameter and the minimum value in the sample to be processed to the deviation value as the normalized result, forming a data set and dividing it into two parts: training set and test set. Deviation normalization can unify the order of magnitude of feature parameters, and the conversion formula is as follows:

(1)

Where Xn _is the value of a parameter of the _sample to be processed, Xmin and Xmax are the maximum and minimum values of the corresponding parameter in the sample data set, Ymin and Ymax are the lower _limit and upper _{limit of} the range respectively, _and Yn is the normalized result.

3) Build the initial SVM fault recognition model, determine the selection of kernel function and the value range of algorithm parameters including kernel function parameters and penalty factor. As shown in the support vector machine binary classification diagram in Figure 6, the optimal classification hyperplane of the support vector machine can be represented as an m -dimensional hyperplane:

(2)

Where ω is the weight vector of the hyperplane, b is the bias, ^Rm represents the m-dimensional set of real numbers, and R represents the set of real numbers.

Finding the optimal classification hyperplane of the support vector machine, that is, the maximum classification interval, can be achieved by transforming it into solving the following constraint problem:

(3)

From formula (3), we can see that solving the objective function of the support vector machine is a quadratic programming problem, so this problem is transformed into another dual problem:

(4)

Let’s introduce a Lagrangian function to solve the corresponding dual problem:

(5)

Among them, α ≥ 0 is the Lagrange multiplier.

Deriving the Lagrangian function L ( ω, b, α ) with respect to ω and b respectively and taking the derivative as 0, we can obtain:

(6)

(7)

From equations (6) and (7), we can see that when the subset of the training sample set is expanded, the normal vector of the optimal classification hyperplane is formed, where the Lagrange multiplier of the expanded sample is not 0, that is, the support vector of the optimal classification hyperplane.

The classification discriminant function of the support vector machine is expressed as:

(8)

However, in practical problems, linear inseparability may occur, which may lead to failure of hard margin classification. Therefore, the mapping transformation method can be used to transform the training samples in the original space into a higher-dimensional linearly separable space through nonlinear mapping φ ( x ). In applications, since the problem of finding nonlinear mapping φ ( x ) and _performing inner product operations is more complicated, the kernel function K ( x , xi ) is introduced based on the basic concept of functional analysis in mathematics.

Common kernel functions in support vector machines include radial basis function, RBF function, polynomial kernel function, Gaussian kernel function and multilayer perceptron function. RBF function can reduce model complexity when implementing mapping and has fewer parameters, and has better performance. Therefore, the present invention selects RBF kernel function as one of the SVM classification model parameters, which can be expressed as the following function form:

(9)

Among them, γ is the kernel function parameter.

In order to obtain the optimal solution of the support vector machine parameter penalty factor and kernel function parameters, the cuckoo search algorithm, namely the CS algorithm, is introduced to improve the support vector machine to realize the search of various parameters. The search process of the cuckoo search algorithm is shown in Figure 7.

The CS algorithm is based on the cuckoo's habit of searching for the best nest to lay eggs. By setting the input parameter group as the nest, the input parameters are optimized according to the Levy flight path principle, and the solution with poor output is replaced to form a solution with better results. The main steps of the algorithm are:

(I) Randomly generate m bird ^nests , set as Xi = [ _X10 , _X20 , ..., _Xm0 ], _and calculate _the initial optimal bird nest position ^X0 by searching ^;

(II) Calculate the fitness value of each position and update the nest position with better fitness through Levy flight:

(10)

As shown in formula (10), each generation is compared with the previous generation position. After continuous replacement iteration, the optimal nest position is finally selected as X _t = [X ₁ ^t , X ₂ ^t , …, X _m ^t ];

(III) When the obtained optimal nest position reaches the accuracy requirement or the final number of iterations, the parameter optimization search is completed and the obtained result is assigned.

The number of bird nests is set to 25, the upper limit of the number of iterations is set to 250, the lower limit of the fitness function is set to 0.05, and the upper limit is set to 0.1. The fitness evolution curve obtained after the cycle end condition is met is shown in Figure 8, and the optimal solution of the parameters is output at the same time.

By searching the parameters of the support vector machine using the CS algorithm, it can be found that when the RBF kernel function is used, the penalty factor and the kernel function parameter γ are respectively around penalty factor = 10.0 and kernel function parameter γ = 0.01, the model classification accuracy is the highest.

4) The parameters obtained by the optimization algorithm are used to construct an improved SVM fault recognition model, namely the target SVM fault recognition model, and the preprocessed fault sample data is input into the constructed improved SVM fault recognition model for training and learning. After model verification, the fault diagnosis accuracy of the sample training set is 99.4444%, and the fault recognition accuracy of the improved SVM fault recognition model for the sample training set is shown in Figure 9. The fault diagnosis accuracy of the test set is 98.8889%, and the fault recognition accuracy of the improved SVM fault recognition model for the sample test set is shown in Figure 10. Therefore, the three-dimensional fault feature quantity composed of U _m , I _m , and the number of inflection points of the UI curve can effectively characterize different working conditions such as abnormal aging and local shadow of the photovoltaic array, and complete fault identification.

The photovoltaic array fault identification method provided in this specific example, support vector machine has a very broad application field in fault diagnosis. The support vector machine algorithm is a small sample algorithm with a deep theoretical foundation. It can quickly predict the sample data during the sample data training process and avoid the steps from induction to deduction. Adding and removing non-support vectors in the sample data has no effect on the model, so it has good robustness. The support vector machine estimates the expected risk based on the principle of structural risk minimization, which can overcome the overfitting problem in the model training process and has a strong generalization ability in classification and discrimination. The introduction of kernel function enables the support vector machine to perform dimensionality reduction operations on the sample space, greatly reducing the dimension of the sample data, reducing the complexity of classification, and can solve the problem of nonlinear inseparability.

Not only as follows, in this specific example, a three-dimensional fault feature group composed of U _m , I _m , and the number of inflection points of the UI curve step is proposed to optimize the penalty factor and kernel function parameters of the SVM fault recognition model, and the selected three-dimensional fault feature is input into the improved SVM fault recognition model to verify its effectiveness, which can accurately, conveniently and efficiently identify different degrees of abnormal aging and different degrees of local shadow faults of photovoltaic arrays.

FIG. 11 is a structural block diagram of a photovoltaic array fault identification device implementing an embodiment of the present application.

The photovoltaic array fault identification device provided in the embodiment of the present application includes the following functional modules:

Establishing module 801, for establishing a topological structure of a grid-connected photovoltaic power generation system based on grid operation parameters;

A simulation module 802 is used to simulate the operation status of the photovoltaic array under a preset abnormal condition based on the topological structure of the grid-connected photovoltaic power generation system to obtain a first output characteristic of the photovoltaic array;

A generating module 803, configured to compare the first output characteristic with a second output characteristic of the photovoltaic array in a normal operating state, and generate a three-dimensional fault feature quantity group;

A model optimization module 804 is used to build an SVM fault recognition model according to the three-dimensional fault feature quantity group, and optimize the penalty factor and kernel function parameters of the SVM fault recognition model to obtain a target SVM fault recognition model;

The identification module 805 is used to verify the validity of the target three-dimensional fault feature quantity based on the target SVM fault identification model to identify the fault of the photovoltaic array.

Optionally, the simulation module includes:

The first submodule is used to establish a simulation model based on the topological structure of the grid-connected photovoltaic power generation system and a preset simulation platform;

The second submodule is used to simulate the photovoltaic array failure under preset abnormal conditions for the simulation model, and obtain the first output characteristics of the photovoltaic array corresponding to each abnormal condition, wherein the preset abnormal conditions include: the photovoltaic array has different degrees of abnormal aging status, and the photovoltaic array has different degrees of local shadow status.

Optionally, the photovoltaic array having different degrees of abnormal aging states includes: the first group of strings in the photovoltaic array having a first degree of aging, and the second group of strings being normal; the first group of strings in the photovoltaic array having a second degree of aging, and the second group of strings being normal; the first group of strings and the second group of strings in the photovoltaic array being both aged;

The states where the photovoltaic array has different degrees of local shadows include: the irradiance of a single string of photovoltaic components in the photovoltaic array is 0, and the irradiance of multiple strings of photovoltaic components in the photovoltaic array is 0.

Optionally, the generating module includes:

The third submodule is used to analyze the output characteristics corresponding to each degree of abnormal aging state, and determine the maximum power point voltage and maximum power point current of the photovoltaic array;

A fourth submodule is used to compare the maximum power point voltage and the maximum power point current corresponding to each degree of abnormal aging state with the maximum power point voltage and the maximum power point current under the normal operating state of the photovoltaic array to obtain a second output characteristic under the abnormal aging state;

A fifth submodule is used to analyze the output characteristics corresponding to each degree of local shadow state, and determine the number of inflection points of the U-I curve of the photovoltaic array;

The sixth submodule is used to use the number of inflection points of each U-I curve step as the second output characteristic under the local shadow state.

Optionally, the model optimization module includes:

The seventh submodule is used to perform deviation normalization processing on the three-dimensional fault characteristic quantity group to obtain a target three-dimensional fault characteristic quantity group; wherein the target three-dimensional fault characteristic quantity group includes multiple groups of three-dimensional fault characteristic quantities composed of the maximum power point voltage and maximum power point current of the photovoltaic array and the number of step inflection points of the U-I curve;

An eighth submodule, configured to divide the fault feature quantities in the target three-dimensional fault feature quantity group into a training set and a test set;

The ninth submodule is used to build an initial SVM fault recognition model, and iteratively optimize the penalty factor and kernel function parameters of the initial SVM fault recognition model according to the training set until the iterative optimization cutoff condition is met to obtain a target SVM fault recognition model.

In the photovoltaic array fault identification device provided in the application embodiment, SVM estimates the expected risk based on the principle of structural risk minimization, which can overcome the overfitting problem in the model training process, so that the SVM-based fault identification model has good classification and discrimination capabilities; in addition, the SVM fault identification model is trained by a three-dimensional fault feature quantity group, which can improve the accuracy of the SVM fault identification model in identifying photovoltaic array faults.

The photovoltaic array fault identification device shown in FIG. 11 in the embodiment of the present application can be set in a mobile device or in a server. The mobile device or server provided with the device can be a device with an operating system. The operating system can be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

The photovoltaic array fault identification device shown in FIG. 11 provided in the embodiment of the present application can implement each process implemented in the method embodiment of FIG. 1 , and will not be described again here to avoid repetition.

Optionally, referring to Figure 12, it is shown that an embodiment of the present application also provides an electronic device 900, including a processor 901, a memory 902, and a program or instruction stored in the memory and executable on the processor. When the program or instruction is executed by the processor, the various processes performed by the above-mentioned photovoltaic array fault identification device are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

It should be noted that the electronic device in the embodiment of the present application includes the server described above.

The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims (10)

1. A method for identifying faults in a photovoltaic array, comprising:

establishing a topological structure of the grid-connected photovoltaic power generation system based on the power grid operation parameters;

simulating the operation condition of the photovoltaic array under a preset abnormal condition based on the topological structure of the grid-connected photovoltaic power generation system to obtain a first output characteristic of the photovoltaic array;

comparing the first output characteristic with a second output characteristic of the photovoltaic array in a normal operation state to generate a three-dimensional fault characteristic quantity set;

building an SVM fault recognition model according to the three-dimensional fault characteristic quantity set, and optimizing penalty factors and kernel function parameters of the SVM fault recognition model to obtain a target SVM fault recognition model;

and verifying the validity of the target three-dimensional fault characteristic quantity based on the target SVM fault identification model so as to identify faults of the photovoltaic array.

2. The method according to claim 1, wherein the step of simulating the operation condition of the photovoltaic array under the preset abnormal condition based on the topology of the grid-connected photovoltaic power generation system to obtain the first output characteristic of the photovoltaic array comprises:

Establishing a simulation model based on the topological structure of the grid-connected photovoltaic power generation system and a preset simulation platform;

performing photovoltaic array fault simulation under preset abnormal conditions on the simulation model, and acquiring first output characteristics of a corresponding photovoltaic array under each abnormal condition, wherein the preset abnormal conditions comprise: the photovoltaic array is in an abnormal aging state with different degrees and in a state with different degrees of local shadows.

3. The method according to claim 2, characterized in that:

the photovoltaic array is subjected to different degrees of abnormal aging states, including: a first group of strings in the photovoltaic array are aged to a first degree, and a second group of strings are normal; a first set of strings in the photovoltaic array are subject to a second degree of aging, the second set of strings being normal; the first set of strings and the second set of strings in the photovoltaic array are both aged;

the state in which the photovoltaic array has different degrees of local shadows includes: irradiance of a single group of string photovoltaic modules in the photovoltaic array is 0, and irradiance of a plurality of groups of string photovoltaic modules in the photovoltaic array is 0.

4. The method of claim 2, wherein the step of generating a three-dimensional set of fault signatures by comparing the first output characteristic to a second output characteristic of the photovoltaic array during normal operation comprises:

Aiming at the abnormal aging state of each degree, analyzing the output characteristics corresponding to the abnormal aging state, and determining the maximum power point voltage and the maximum power point current of the photovoltaic array;

comparing the maximum power point voltage and the maximum power point current corresponding to the abnormal aging state of each degree with the maximum power point voltage and the maximum power point current in the normal operation state of the photovoltaic array to obtain a second output characteristic in the abnormal aging state;

aiming at the state of the partial shadow of each degree, analyzing the output characteristic corresponding to the state of the partial shadow, and determining the number of step turns of the U-I curve of the photovoltaic array;

and taking the number of step turns of each U-I curve as a second output characteristic in a partial shadow state.

5. The method of claim 4, wherein the step of constructing an SVM support vector machine fault recognition model according to the three-dimensional fault feature set, and optimizing a penalty factor and a kernel function parameter of the SVM fault recognition model to obtain a target SVM fault recognition model comprises:

performing dispersion normalization processing on the three-dimensional fault characteristic quantity set to obtain a target three-dimensional fault characteristic quantity set; the target three-dimensional fault characteristic quantity group comprises a plurality of groups of three-dimensional fault characteristic quantities formed by the number of U-I curve step turning points based on the maximum power point voltage and the maximum power point current of the photovoltaic array;

Dividing fault characteristic quantities in the target three-dimensional fault characteristic quantity group into a training set and a testing set;

and constructing an initial SVM fault recognition model, and carrying out iterative optimization on the penalty factors and the kernel function parameters of the initial SVM fault recognition model according to the training set until the iterative optimization cut-off condition is met, so as to obtain a target SVM fault recognition model.

6. A photovoltaic array fault identification device, comprising:

the building module is used for building a topological structure of the grid-connected photovoltaic power generation system based on the power grid operation parameters;

the simulation module is used for simulating the operation condition of the photovoltaic array under the preset abnormal condition based on the topological structure of the grid-connected photovoltaic power generation system to obtain a first output characteristic of the photovoltaic array;

the generating module is used for comparing the first output characteristic with the second output characteristic of the photovoltaic array in a normal operation state to generate a three-dimensional fault characteristic quantity set;

the model optimization module is used for building an SVM fault recognition model according to the three-dimensional fault characteristic quantity set, and optimizing penalty factors and kernel function parameters of the SVM fault recognition model to obtain a target SVM fault recognition model;

and the identification module is used for carrying out validity verification on the target three-dimensional fault characteristic quantity based on the target SVM fault identification model so as to carry out fault identification on the photovoltaic array.

7. The apparatus of claim 6, wherein the simulation module comprises:

the first sub-module is used for establishing a simulation model based on the topological structure of the grid-connected photovoltaic power generation system and a preset simulation platform;

the second sub-module is used for carrying out photovoltaic array fault simulation under preset abnormal conditions on the simulation model, and obtaining first output characteristics of the corresponding photovoltaic array under each abnormal condition, wherein the preset abnormal conditions comprise: the photovoltaic array is in an abnormal aging state with different degrees and in a state with different degrees of local shadows.

8. The apparatus according to claim 7, wherein:

the photovoltaic array is subjected to different degrees of abnormal aging states, including: a first group of strings in the photovoltaic array are aged to a first degree, and a second group of strings are normal; a first set of strings in the photovoltaic array are subject to a second degree of aging, the second set of strings being normal; the first set of strings and the second set of strings in the photovoltaic array are both aged;

the state in which the photovoltaic array has different degrees of local shadows includes: irradiance of a single group of string photovoltaic modules in the photovoltaic array is 0, and irradiance of a plurality of groups of string photovoltaic modules in the photovoltaic array is 0.

9. The apparatus of claim 7, wherein the generating module comprises:

the third sub-module is used for analyzing the output characteristics corresponding to the abnormal aging state according to the abnormal aging state of each degree and determining the maximum power point voltage and the maximum power point current of the photovoltaic array;

the fourth sub-module is used for comparing the maximum power point voltage and the maximum power point current corresponding to the abnormal aging state of each degree with the maximum power point voltage and the maximum power point current in the normal operation state of the photovoltaic array to obtain a second output characteristic in the abnormal aging state;

a fifth sub-module, configured to analyze, for each degree of local shadow state, an output characteristic corresponding to the local shadow state, and determine a number of step turns of a U-I curve of the photovoltaic array;

and a sixth sub-module, configured to take the number of step turns of each U-I curve as a second output characteristic in a partial shadow state.

10. An electronic device comprising a processor, a memory and a program or instruction stored on the memory and executable on the processor, the program or instruction being executable by the processor to perform the steps of the method for identifying a photovoltaic array fault of any of claims 1-5.