WO2020029344A1 - 一种基于随机森林算法的光伏阵列故障诊断方法 - Google Patents

一种基于随机森林算法的光伏阵列故障诊断方法 Download PDF

Info

Publication number
WO2020029344A1
WO2020029344A1 PCT/CN2018/102907 CN2018102907W WO2020029344A1 WO 2020029344 A1 WO2020029344 A1 WO 2020029344A1 CN 2018102907 W CN2018102907 W CN 2018102907W WO 2020029344 A1 WO2020029344 A1 WO 2020029344A1
Authority
WO
WIPO (PCT)
Prior art keywords
photovoltaic array
photovoltaic
state
fault
branch
Prior art date
Application number
PCT/CN2018/102907
Other languages
English (en)
French (fr)
Inventor
周超超
陶洪峰
沈建强
魏强
刘巍
周龙辉
Original Assignee
江南大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江南大学 filed Critical 江南大学
Priority to US16/557,220 priority Critical patent/US11114977B2/en
Publication of WO2020029344A1 publication Critical patent/WO2020029344A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02S—GENERATION OF ELECTRIC POWER BY CONVERSION OF INFRARED RADIATION, VISIBLE LIGHT OR ULTRAVIOLET LIGHT, e.g. USING PHOTOVOLTAIC [PV] MODULES
    • H02S50/00—Monitoring or testing of PV systems, e.g. load balancing or fault identification
    • H02S50/10—Testing of PV devices, e.g. of PV modules or single PV cells
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E10/00—Energy generation through renewable energy sources
    • Y02E10/50—Photovoltaic [PV] energy

Definitions

  • the invention relates to the field of photovoltaic technology, in particular to a photovoltaic array fault diagnosis method based on a random forest algorithm.
  • CSP solar thermal power
  • photovoltaic power is similar to thermal power generation, but the thermal energy mainly comes from large-scale mirror collection. After heating the water, it uses steam to promote the operation of traditional generators to achieve power generation. Although this type of power generation avoids the photoelectric conversion of silicon crystals, Light intensity requirements are high and power generation costs are high.
  • Photovoltaic power generation is based on the photovoltaic effect, which directly converts solar energy into electricity. The construction period is short, and no waste residue, waste water and other pollutants will be generated during operation, especially for mountainous areas, islands and remote areas with less developed transportation. More important value.
  • photovoltaic power generation uses photovoltaic cell modules in series and parallel to form a photovoltaic array to operate, and photovoltaic arrays use more photovoltaic cell modules, the probability of failure of the photovoltaic cell module itself is higher, and the photovoltaic array runs for a long time Outdoors, the operating environment is harsh, and it is prone to aging and damage. As a result, the photovoltaic cell module's power generation efficiency is reduced or even stopped. When a photovoltaic cell module in a photovoltaic array fails, it will lead to a decrease in system efficiency and adversely affect the operation and scheduling of the power system. In severe cases, it may even cause property damage and casualties. Therefore, the photovoltaic array is faulty.
  • Diagnosis is of great significance.
  • the currently used fault diagnosis methods mainly include two types: direct method and indirect method. Indirect method is more typical with infrared heat detection method and power generation comparison method. Direct method is more typical with ground capacitance method and time. Domain reflection method, intelligent diagnosis algorithm and electrical characteristic detection method. Among them, fault diagnosis through the combination of intelligent diagnostic algorithms and electrical characteristic detection methods is currently a very promising method. At present, most intelligent diagnosis algorithms use neural networks represented by BP (Back Propagation). Due to the many types of faults that may occur in photovoltaic arrays, a large number of training samples and training time are required. The algorithm has a complex structure, is difficult to implement, and has a long diagnosis time Long and the accuracy and reliability of the diagnosis are not high.
  • BP Back Propagation
  • the inventor proposed a photovoltaic array fault diagnosis method based on a random forest algorithm.
  • This method is based on a data-driven idea and uses a random forest algorithm to construct a fault diagnosis model, which is suitable for the characteristics of an actual photovoltaic array. It overcomes the problems that traditional neural network algorithms require large amounts of data and long training time, and can complete diagnostic tasks simply and quickly.
  • a photovoltaic array fault diagnosis method based on a random forest algorithm includes:
  • the photovoltaic array includes n branches, each branch includes m photovoltaic modules, m and n are positive integers and m ⁇ 4, n ⁇ 4;
  • each circuit parameter group includes k circuit parameters, where k is a positive integer, according to the collected data.
  • a data sample set of the photovoltaic array is constructed, and the data sample set is divided into a training sample set and a test sample set;
  • a photovoltaic array fault diagnosis model is constructed based on a random forest algorithm using training sample sets, and a test sample set is used to test the photovoltaic array fault diagnosis model.
  • the photovoltaic array fault diagnosis model includes s decision trees, s ⁇ 2 and s is a positive integer. ;
  • the PV array fault diagnosis model completed by the test is used to diagnose the PV array to be diagnosed, and the voting results of s decision trees for each typical operating state are obtained;
  • the fault diagnosis result of the photovoltaic array to be diagnosed is obtained according to the voting results of each typical operating state, and the fault diagnosis result is used to indicate the operating state of each branch in the photovoltaic array.
  • Sampling is performed on the training sample set s times, and each sub-training sample set is obtained, and a total of s sub-training sample set is obtained, where s ⁇ 2 and s is a positive integer;
  • each circuit parameter group includes an open circuit voltage, a short circuit current, a maximum power point voltage and a maximum power point current.
  • the first category is the normal operating state, and the first category includes a small category, which is the state when each photovoltaic module in each branch in the photovoltaic array is normally operating;
  • the second category is the short-circuit fault state, and the second category includes three categories, which are the states when a photovoltaic module in a branch in a photovoltaic array is short-circuited, and there are two photovoltaics in a branch in a photovoltaic array. The state when the module is short-circuited, and the state when the two branches in the photovoltaic array have a photovoltaic module short-circuited;
  • the third category is an open-circuit fault state, and the third category includes a small category, which is the state when a branch in the photovoltaic array is open;
  • the fourth category is the shadow fault state, and the fourth category includes three sub-categories, which are the states when a photovoltaic module in a branch in a photovoltaic array has a shadow, and the two branches in the photovoltaic array have both.
  • the state of a photovoltaic module when there is shadow and the state of one photovoltaic module in one branch in the photovoltaic array, the state of two photovoltaic modules in the other branch, and the two branches in the photovoltaic array.
  • the fifth category is a mixed fault state.
  • the fifth category includes four sub-categories, which are the state when one branch in the photovoltaic array is open, the other branch has a photovoltaic module in the shadow state, and one in the photovoltaic array.
  • This application discloses a photovoltaic array fault diagnosis method based on a random forest algorithm.
  • the method is based on a data-driven idea and uses a random forest algorithm to build a photovoltaic array fault diagnosis model. It is applicable to the characteristics of actual photovoltaic arrays and overcomes the traditional nerves.
  • Network algorithms require problems such as large amounts of data and long training time. They can easily and quickly complete diagnostic tasks and quickly implement fault diagnosis for small photovoltaic arrays, especially 3 ⁇ 2 photovoltaic arrays.
  • the invention adopts a random forest fault diagnosis model structure. Through the integration of multiple decision trees, a strong classifier is constructed from many weak classifiers. The diagnosis result is generated by voting. Take the diagnosis results of the second and third votes for the reference of maintenance personnel, improve the maintenance efficiency, and shorten the system failure time.
  • FIG. 1 is a flowchart of a photovoltaic array fault diagnosis method disclosed in the present application.
  • Figure 2-a is the U-I output characteristic curve of a single solar cell module under different light intensities.
  • Figure 2-b is the U-P output characteristic curve of a single solar cell module under different light intensities.
  • Figure 2-c is the U-I output characteristic curve of a single solar cell module under different temperature conditions.
  • Figure 2-d is the U-P output characteristic curve of a single solar cell module at different temperatures.
  • FIG. 3 is an output characteristic curve of a single solar cell module under a light intensity of 600 W / m 2 , 800 W / m 2 , and 1000 W / m 2 .
  • Figure 4-a is the U-I output characteristic curve when a branch in a photovoltaic array fails.
  • Figure 4-b is the U-P output characteristic curve when a branch in the photovoltaic array fails.
  • Figure 4-c is the U-I output characteristic curve when two branches in the photovoltaic array fail.
  • Figure 4-d is the U-P output characteristic curve when two branches in the photovoltaic array fail.
  • Figure 5-a is the U-I output characteristic curve when the photovoltaic array is in the operating state corresponding to the state label F1.
  • Figure 5-b is the U-I output characteristic curve when the photovoltaic array is in the operating state corresponding to the state label F2.
  • Figure 5-c is the U-I output characteristic curve when the photovoltaic array is in the running state corresponding to the state label F5.
  • Figure 5-d is the U-I output characteristic curve when the photovoltaic array is in the running state corresponding to the state label F8.
  • Figure 5-e is the U-I output characteristic curve when the photovoltaic array is in the running state corresponding to the status label F13.
  • This application discloses a photovoltaic array fault diagnosis method based on a random forest algorithm. Please refer to FIG. 1.
  • the entire fault diagnosis process includes the following steps:
  • the first step is to build a photovoltaic array in the form of m * n.
  • the photovoltaic array includes n branches in parallel, each branch includes m photovoltaic modules connected in series, and m and n are positive integers.
  • the photovoltaic modules in the array are usually solar cell modules.
  • Each photovoltaic module has the same specifications to ensure the potential balance of the photovoltaic array, and protection and detection devices such as bypass diodes, isolation diodes, fast fuses, voltage and current sensors are installed correctly. And works fine.
  • the method disclosed in the present application is more suitable for small photovoltaic arrays, so m ⁇ 4 and n ⁇ 4 are usually used.
  • Six SunTech STP 270-24 / Vd solar cell modules of Wuxi Suntech Solar Co., Ltd. are used to build a photovoltaic array. This type of solar cell module
  • the parameter specification table is shown in the following table:
  • Figure 2a shows the UI (voltage-current) output characteristic curve of a single solar cell module under the same temperature conditions and different light intensities
  • Figure 2b shows the UP (voltage-power) output characteristic curve.
  • the UI output characteristic curve when the temperature drops from 35 ° C to 15 ° C (the figure is 35 ° C, 30 ° C, 25 ° C, 20 ° C, and 15 ° C, respectively) is shown in Figure 2-c
  • the output and UP output characteristics are shown in Figure 2-d. It can be seen that as the light intensity increases and the temperature environment decreases, the output power of the solar cell module increases, and the power generation amount increases.
  • the output characteristic curves for the light intensity of 600 W / m 2 , 800 W / m 2 , and 1000 W / m 2 are separately shown in FIG. 3.
  • the second step is to determine the typical operating status of the photovoltaic array during operation.
  • the typical operating status of the photovoltaic array is used to indicate the operating status of each branch in the photovoltaic array.
  • There are many types of faults that occur in the actual operation of photovoltaic arrays which are mainly divided into four types: short-circuit faults, open-circuit faults, shadow faults, and mixed faults.
  • the operating states of photovoltaic arrays mainly include five categories: Normal operating state, short-circuit fault state, open-circuit fault state, shadow fault state, and mixed fault state.
  • the UI output characteristic curve when a branch fails in a photovoltaic array is shown in Figure 4-a; there is a branch Figure 4-b shows the UP output characteristic curve when there is a fault on the road; Figure 4-c shows the UI output characteristic curve when there are two branches on the fault; The curve is shown in Figure 4-d. It can be seen from the figure that when a photovoltaic module fails and its branch fails, whether it is the faulty branch or the entire photovoltaic array, there will be a significant power loss. The output of the normal branch and normal array The characteristic curve is unimodal, while the fault branch and fault array are multimodal.
  • the five major operating states of photovoltaic arrays can be further divided into multiple sub-categories.
  • several of the most common sub-categories can be selected as typical operating states of photovoltaic arrays.
  • the first category is a normal operating state.
  • the first category includes a small category, which is a state when each photovoltaic module in each branch in a photovoltaic array is normally operating.
  • the second category is the short-circuit fault state.
  • the second category includes three small categories, which are the states when a photovoltaic module in a branch in a photovoltaic array is short-circuited, and a branch in a photovoltaic array. There are states when two photovoltaic modules are short-circuited, and when two branches in a photovoltaic array have a photovoltaic module that is short-circuited.
  • the short-circuit fault can be regarded as a special shadow fault: the photovoltaic module is completely blocked and cannot work and is short-circuited by the bypass diode.
  • the shadow fault there are at most two in a branch. Photovoltaic modules are short-circuited. If all photovoltaic modules are short-circuited, the fuse device will blow and the branch will open.
  • the third category is an open-circuit fault state.
  • the third category includes a small category, which is the state when a branch in the photovoltaic array is open.
  • the fourth major category is the shadow fault state.
  • the fourth major category includes three sub-categories, which are the states when a photovoltaic module in a branch in a photovoltaic array has a shadow, and the two branches in the photovoltaic array. There is a state where a photovoltaic module has shadows on the road, and a state where one photovoltaic module has shadows on one of the branches in the photovoltaic array, a state where two photovoltaic modules have shadows on the other branch, and The state when two photovoltaic modules have shadows in both branches. It should be noted that a shadow fault is different from a short-circuit fault, and there may be a situation in which all photovoltaic modules have a shadow.
  • the fifth category is a mixed fault state.
  • the fifth category includes four small categories, which are the state when one branch of the photovoltaic array is open, the other branch has a photovoltaic module in the shadow, and the photovoltaic When one of the branches in the array has a short-circuited photovoltaic module, the other has a photovoltaic module in the shadowed state, and the two branches in the photovoltaic array have a short-circuited photovoltaic module and a photovoltaic module has a shadow. And the state when the two branches in the photovoltaic array have one photovoltaic module short-circuited and the two photovoltaic modules have shadows.
  • each typical operating state can be represented by a status label.
  • the above thirteen types of typical operating states are represented in the form of a table as follows:
  • the photovoltaic array is in the state label F1.
  • the corresponding UI output characteristic curve in the corresponding operating state is shown in Figure 5-a.
  • the photovoltaic output characteristic curve in the operating state corresponding to the photovoltaic array in the state label F2 is shown in Figure 5-b.
  • the photovoltaic array is in the corresponding state label F5.
  • Figure 5-c shows the UI output characteristic curve during the running state.
  • Figure 5-d shows the UI output characteristic curve when the photovoltaic array is in the running state corresponding to the state label F8.
  • Figure 5-d shows the photovoltaic array in the running state corresponding to the state label F13.
  • the UI output characteristic curve at this time is shown in Figure 5-e.
  • the marks on the curves are superimposed.
  • the overlap of the curves of the branch B and the array will cause a hexagonal icon to be displayed on the figure.
  • those skilled in the art should The meaning of the actual representation is clear in the figure.
  • the present application does not provide examples of UI output characteristic curves when the photovoltaic array is in other typical operating states.
  • the third step is to construct the fault feature vector.
  • the circuit parameter groups corresponding to each branch and main trunk in the photovoltaic array are collected separately.
  • the interval of the light intensity and temperature of the photovoltaic module is approximately [800,900] W / m. 2 and [40,50] ° C.
  • the four parameters I m constitute a circuit parameter group, because these four parameters can well describe the UI output characteristic curve of a photovoltaic module.
  • p K * (n + 1).
  • the data sample set of the photovoltaic array can be summarized and constructed.
  • the method of collecting the data sample set from the fault feature vector is a method known to those skilled in the art, and this application does not perform this. Expand the narrative.
  • the data sample set is divided to obtain a training sample set and a test sample set, which can usually be divided into a training sample set and a test sample set according to a ratio of 2: 1.
  • the fifth step is the construction of a photovoltaic array fault diagnosis model based on the random forest algorithm.
  • the training sample set be X and p be the dimension of the fault feature vector.
  • p 12;
  • s is The number of decision trees to be established in fault diagnosis, s ⁇ 2 and s is a positive integer.
  • each sub-training sample set to train to obtain a decision tree that is, use the sub-training sample set X i to train a decision tree Q i .
  • the calculation method of the Gini coefficient can be Some calculation formulas are not described in detail in this application.
  • the sixth step is the test of the photovoltaic array fault diagnosis model.
  • indicators such as diagnostic accuracy, average training time, and test time can be obtained.
  • the fault diagnosis based on the random forest (RF) algorithm of this application is compared with the fault diagnosis based on neural network (BP) and extreme learning machine (ELM).
  • BP neural network
  • ELM extreme learning machine
  • the PV array fault diagnosis model completed by the test is used to diagnose the PV array to be diagnosed, that is, the s decision trees in the PV array fault diagnosis model are used to vote. Since the diagnosis result of the random forest is generated by voting, Voting results of s decision trees for each typical running state will be obtained. According to each voting result, the fault diagnosis result of the photovoltaic array to be diagnosed can be obtained, that is, the operating status of each branch in the photovoltaic array to be diagnosed is determined. Under normal circumstances, the typical operating state with the most votes is the real-time operating state of the photovoltaic array to be diagnosed, which can also determine the type of failure of the photovoltaic array. Taking 20 decision trees as an example, the partial voting results of 5 groups of measured results are summarized, and the corresponding fault diagnosis results are shown in the following table, which is indicated by status labels:
  • the typical number of votes corresponding to the typical operating status of the status label F2 is 19 votes
  • the diagnosis result is that the photovoltaic array to be diagnosed is in the typical operating status corresponding to the status label F2, that is, diagnosis It is obtained that a photovoltaic module in a branch of the photovoltaic array to be diagnosed has a short circuit, and the diagnosis result is consistent with the true value of the fault.
  • the number of votes corresponding to the typical operating status corresponding to the status label F11 is 10
  • the diagnosis result is that the photovoltaic array to be diagnosed is in the typical operating status corresponding to the status label F10, that is, the diagnosis obtains the PV array to be diagnosed.
  • One branch has an open circuit, and one photovoltaic module has a shadow in the other branch.
  • this diagnosis result is not consistent with the true value of the fault, since this application will get the voting results corresponding to each typical operating state, even if this diagnosis occurs In the case of an error, the type of actual operating status will also appear in the voting statistics (you can see that the typical operating status corresponding to the status label F7 is 5 votes, the second most votes), which is equivalent to having an alternative diagnosis result, so The staff can retrieve the votes of each typical operating state and check the statistics of the votes to find out the types of possible failures. Even if the diagnosis corresponding to the highest ticket has an error, they can still troubleshoot according to the number of votes. That is, this application actually provides a diagnostic route, thereby improving maintenance efficiency and shortening System downtime.

Landscapes

  • Photovoltaic Devices (AREA)

Abstract

本发明公开了一种基于随机森林算法的光伏阵列故障诊断方法，涉及光伏技术领域，该方法包括：采集光伏阵列处于各个典型运行状态时，各个支路和总干路对应的电路参数组，根据采集到的电路参数组构建得到故障特征向量，从而构建数据样本集，利用数据样本集基于随机森林算法构建得到光伏阵列故障诊断模型，利用该模型对待诊断光伏阵列进行诊断，得到各个典型运行状态对应的投票结果，根据各个投票结果得到待诊断光伏阵列的故障诊断结果，该方法基于数据驱动的思想，利用随机森林算法构建光伏阵列故障诊断模型，适用于实际光伏阵列的特点，克服了传统神经网络算法需要数据量大、训练时间长等问题，能够简单快速地完成诊断任务。

Description

一种基于随机森林算法的光伏阵列故障诊断方法 技术领域
本发明涉及光伏技术领域，尤其是一种基于随机森林算法的光伏阵列故障诊断方法。
背景技术
随着能源需求的增长、化石能源的日益枯竭与成本上涨，以及全球气候变暖等诸多因素的影响，可再生能源技术得到了飞速发展，其中，太阳能具有易获取、无噪声、清洁、无穷无尽等优点，成为了可再生能源的重要部分。目前利用太阳能发电主要有两种形式：光热发电和光伏发电。光热发电与火力发电类似，但热能主要来自大规模的镜面收集，将水加热后，利用蒸汽推动传统发电机工作，从而实现发电，这种发电方式虽然规避了硅晶的光电转换，但对光照强度的要求高且发电成本偏高。而光伏发电基于光生伏特效应，直接将太阳能转化为电能，建设周期短，在运行时不会产生废渣、废水和其他污染物，尤其是对于交通不发达的山区、海岛和偏远地区，光伏发电具有更重要的价值。
一般情况下，光伏发电都是利用光伏电池组件串、并联组成光伏阵列进行运行的，而光伏阵列使用的光伏电池组件较多，光伏电池组件本身发生故障的概率就较高，并且光伏阵列长期运行在户外，运行环境恶劣，易出现老化和损坏等情况，从而造成光伏电池组件发电效率的降低甚至停止工作。当光伏阵列中某个光伏电池组件发生故障后，会导致系统效率的下降，并对电力系统的运行调度造成不利的影响，严重时，甚至造成财产损失和人员伤亡，因此，对光伏阵列进行故障诊断具有重要意义，目前所采用的故障诊断方法主要包括直接法和间接法两类，间接法比较典型的有红外热量检测法和发电功率对比法，直接法比较典型的有对地电容法、时域反射法、智能诊断算法和电特性检测法。其中，通过将智能诊断算法和电特性检测法进行结合进行故障诊断，是目前非常具有潜力的方法。目前的智能诊断算法多采用以BP(Back Propagation)为代表的神经网络，由于光伏阵列可能出现的故障类型较多，因此需要大量训练样本和训练时间，算法结构复杂、实现难度大、诊断时间过长且诊断的精度和可靠 性也不高。
发明内容
本发明人针对上述问题及技术需求，提出了一种基于随机森林算法的光伏阵列故障诊断方法，该方法基于数据驱动的思想，利用随机森林算法构建故障诊断模型，适用于实际光伏阵列的特点，克服了传统神经网络算法需要数据量大、训练时间长等问题，能够简单快速地完成诊断任务。
本发明的技术方案如下：
一种基于随机森林算法的光伏阵列故障诊断方法，该方法包括：
确定光伏阵列在运行过程中的典型运行状态，光伏阵列包括n个支路，每个支路包括m个光伏组件，m和n均为正整数且m≤4、n≤4；
在光伏阵列处于各个典型运行状态时，分别采集光伏阵列中的各个支路和总干路对应的电路参数组，每个电路参数组中包括k个电路参数，k为正整数，根据采集到的n+1个电路参数组构建得到一个p维的故障特征向量，其中，p＝k*(n+1)；
根据获取到的故障特征向量构建光伏阵列的数据样本集，将数据样本集划分为训练样本集和测试样本集；
基于随机森林算法利用训练样本集构建得到光伏阵列故障诊断模型，并利用测试样本集对光伏阵列故障诊断模型进行测试，光伏阵列故障诊断模型中包括s个决策树，s≥2且s为正整数；
利用测试完成的光伏阵列故障诊断模型对待诊断光伏阵列进行诊断，得到s个决策树对各个典型运行状态的投票结果；
根据各个典型运行状态的投票结果得到待诊断光伏阵列的故障诊断结果，故障诊断结果用于指示光伏阵列中的各个支路的运行状态。
其进一步的技术方案为，基于随机森林算法利用训练样本集构建得到光伏阵列故障诊断模型，包括：
对训练样本集进行s次有放回抽样，每次抽样得到一个子训练样本集，共得到s个子训练样本集，s≥2且s为正整数；
利用每个子训练样本集训练得到一个决策树；
确定训练得到的s个决策树汇总得到的集合即为光伏阵列故障诊断模型。
其进一步的技术方案为，在利用子训练样本集训练决策树的过程中，在进 行节点分裂时，从p个故障特征中随机选择q个故障特征，q为正整数且q≤p，计算q个故障特征中的各个故障特征对应的基尼系数，将与最小的基尼系数对应的故障特征作为最佳分裂特征来进行节点分裂。
其进一步的技术方案为，k＝4，每个电路参数组中包括开路电压、短路电流、最大功率点电压和最大功率点电流。
其进一步的技术方案为，n＝2、m＝3，光伏阵列的典型运行状态包括五大类共十二小类，分别为：
第一大类为正常运行状态，第一大类包括一小类，为光伏阵列中的各个支路中的各个光伏组件均正常运行时的状态；
第二大类为短路故障状态，第二大类包括三小类，分别为光伏阵列中的一个支路中有一个光伏组件短路时的状态，以及光伏阵列中的一个支路中有两个光伏组件短路时的状态，以及光伏阵列中的两个支路均有一个光伏组件短路时的状态；
第三大类为开路故障状态，第三大类包括一小类，为光伏阵列中有一个支路开路时的状态；
第四大类为阴影故障状态，第四大类包括三小类，分别为光伏阵列中的一个支路中有一个光伏组件存在阴影时的状态，以及光伏阵列中的两个支路中均有一个光伏组件存在阴影时的状态，以及光伏阵列中的一个支路中有一个光伏组件存在阴影、另一个支路中有两个光伏组件存在阴影时的状态，以及光伏阵列中的两个支路中均有两个光伏组件存在阴影时的状态；
第五大类为混合故障状态，第五大类包括四小类，分别为光伏阵列中的一个支路开路、另一个支路中有一个光伏组件存在阴影时的状态，以及光伏阵列中的一个支路中有一个光伏组件短路、另一个支路中有一个光伏组件存在阴影时的状态，以及光伏阵列中的两个支路均有一个光伏组件短路、一个光伏组件存在阴影时的状态，以及光伏阵列中的两个支路均有一个光伏组件短路、两个光伏组件存在阴影时的状态。
本发明的有益技术效果是：
1、本申请公开了一种基于随机森林算法的光伏阵列故障诊断方法，该方法基于数据驱动的思想，利用随机森林算法构建光伏阵列故障诊断模型，适用于实际光伏阵列的特点，克服了传统神经网络算法需要数据量大、训练时间长等问题，能够简单快速地完成诊断任务，快速实现小型光伏阵列，尤其是3×2光 伏阵列的故障诊断。
2、本发明采用随机森林的故障诊断模型结构，通过多个决策树的集成，由许多弱分类器构建了一个强分类器，诊断结果由投票产生，即使最多票数的诊断结果错误，还能调取第二、三票数的诊断结果，供维修人员参考，提高维修效率，从而缩短系统的故障时间。
附图说明
图1是本申请公开的光伏阵列故障诊断方法的流程图。
图2-a是单个太阳能电池组件在不同光照强度下的U-I输出特性曲线。
图2-b是单个太阳能电池组件在不同光照强度下的U-P输出特性曲线。
图2-c是单个太阳能电池组件在不同温度情况下的U-I输出特性曲线。
图2-d是单个太阳能电池组件在不同温度情况下的U-P输出特性曲线。
图3是单个太阳能电池组件在光照强度600W/m 2、800W/m 2、1000W/m 2情况下的输出特性曲线。
图4-a是光伏阵列中有一个支路发生故障时的U-I输出特性曲线。
图4-b是光伏阵列中有一个支路发生故障时的U-P输出特性曲线。
图4-c是光伏阵列中有两个支路发生故障时的U-I输出特性曲线。
图4-d是光伏阵列中有两个支路发生故障时的U-P输出特性曲线。
图5-a是光伏阵列处于状态标签F1对应的运行状态时的U-I输出特性曲线。
图5-b是光伏阵列处于状态标签F2对应的运行状态时的U-I输出特性曲线。
图5-c是光伏阵列处于状态标签F5对应的运行状态时的U-I输出特性曲线。
图5-d是光伏阵列处于状态标签F8对应的运行状态时的U-I输出特性曲线。
图5-e是光伏阵列处于状态标签F13对应的运行状态时的U-I输出特性曲线。
具体实施方式
下面结合附图对本发明的具体实施方式做进一步说明。
本申请公开了一种基于随机森林算法的光伏阵列故障诊断方法，请参考图1，整个故障诊断流程包括如下步骤：
第一步，搭建m*n形式的光伏阵列，该光伏阵列包括并联的n个支路，每个支路包括串联的m个光伏组件，m和n均为正整数。阵列中的光伏组件通常是太阳能电池组件，每个光伏组件规格相同，以保证光伏阵列的电位平衡，并 且旁路二极管、隔离二极管、快速熔断器、电压和电流传感器等保护、检测装置均安装正确并正常工作。在实际实现时，本申请公开的方法更适用于小型光伏阵列，因此通常有m≤4、n≤4。
本申请以更为典型的n＝2、m＝3的情况为例进行说明，采用6块无锡尚德太阳能有限公司SunTech STP 270-24/Vd型太阳能电池组件搭建光伏阵列，该型号的太阳能电池组件的参数规格表如下表所示：
参数类型 参数值
开路电压(V oc) 44.5V
最大功率点电压(V m) 35.0V
短路电流(I sc) 8.20A
最大功率点电流(I m) 7.71A
最大功率(P max) 270W
运行温度 -40℃～+85℃
系统最大承受电压 1000V DC
串联后保险丝的额定电流 20A
功率误差 ±3％
单个太阳能电池组件在温度情况相同的条件下，在不同光照强度下的U-I(电压-电流)输出特性曲线如图2-a所示、U-P(电压-功率)输出特性曲线如图2-b所示；在光照强度相同的情况下，温度从35℃下降至15℃(图中分别为35℃、30℃、25℃、20℃和15℃)时的U-I输出特性曲线如图2-c所示、U-P输出特性曲线如图2-d所示。可以看出，随着光照强度的增加和温度环境的降低，太阳能电池组件的输出功率增加，发电量增大。单独给出了光照强度600W/m 2、800W/m 2、1000W/m 2情况下的输出特性曲线如图3所示。
第二步，确定光伏阵列在运行过程中的典型运行状态，光伏阵列的典型运行状态用于指示光伏阵列中的各个支路的运行状态。光伏阵列在实际运行时会出现较多类型的故障，主要划分为四种：短路故障、开路故障、阴影故障以及混合故障，再加上正常运行的情况，光伏阵列的运行状态主要包括五大类：正常运行状态、短路故障状态、开路故障状态、阴影故障状态以及混合故障状态。
当光伏阵列中的一个支路或多个支路发生故障时，故障的支路以及整个光伏阵列的输出特性曲线都会发生变化，如图4-a至4-d分别示出了不同情况下的故障支路、故障阵列、正常支路和正常阵列的输出特性曲线，其中正常支路是指正常运行的支路，正常阵列是指各个支路均正常运行时的光伏阵列，故障支路是指发生故障时的支路，故障阵列是指有支路发生故障时的光伏阵列，具体的：光伏阵列中有一个支路发生故障时的U-I输出特性曲线如图4-a所示；有一个支路发生故障时的U-P输出特性曲线如图4-b所示；有两个支路发生故 障时的U-I输出特性曲线如图4-c所示；有两个支路发生故障时的U-P输出特性曲线如图4-d所示。由图中可以看出，当光伏组件发生故障导致其所在支路发生故障时，无论是该发生故障的支路还是整个光伏阵列，都会出现一个明显的功率损失，正常支路和正常阵列的输出特性曲线呈单峰特性，而故障支路和故障阵列呈多峰特性。
光伏阵列的五大类运行状态可以进一步分为多个小类，实际可以在其中选择若干最为常见的小类作为光伏阵列的典型运行状态，其具体选择可以根据实际经验进行。比如在本申请n＝2、m＝3的例子中，可以在这些运行状态大类中选择十三小类最为常见的运行状态作为典型运行状态，分别为：
(1)、第一大类为正常运行状态，第一大类包括一小类，为光伏阵列中的各个支路中的各个光伏组件均正常运行时的状态。
(2)、第二大类为短路故障状态，第二大类包括三小类，分别为光伏阵列中的一个支路中有一个光伏组件短路时的状态，以及光伏阵列中的一个支路中有两个光伏组件短路时的状态，以及光伏阵列中的两个支路均有一个光伏组件短路时的状态。
需要说明的是，短路故障可以看作是一种特殊的阴影故障：光伏组件被完全遮挡而无法工作而被旁路二极管全程短路，但与阴影故障不同的是，一条支路中最多存在两个光伏组件短路，若全部光伏组件短路，则熔断装置熔断，该支路会开路。
(3)、第三大类为开路故障状态，第三大类包括一小类，为光伏阵列中有一个支路开路时的状态。
(4)、第四大类为阴影故障状态，第四大类包括三小类，分别为光伏阵列中的一个支路中有一个光伏组件存在阴影时的状态，以及光伏阵列中的两个支路中均有一个光伏组件存在阴影时的状态，以及光伏阵列中的一个支路中有一个光伏组件存在阴影、另一个支路中有两个光伏组件存在阴影时的状态，以及光伏阵列中的两个支路中均有两个光伏组件存在阴影时的状态。需要说明的是，阴影故障与短路故障不同，可以存在所有光伏组件均存在阴影的情况。
(5)、第五大类为混合故障状态，第五大类包括四小类，分别为光伏阵列中的一个支路开路、另一个支路中有一个光伏组件存在阴影时的状态，以及光伏阵列中的一个支路中有一个光伏组件短路、另一个支路中有一个光伏组件存在阴影时的状态，以及光伏阵列中的两个支路均有一个光伏组件短路、一个光 伏组件存在阴影时的状态，以及光伏阵列中的两个支路均有一个光伏组件短路、两个光伏组件存在阴影时的状态。
为了清楚简洁地表示各个典型运行状态，可以用状态标签表示各个典型运行状态，上述十三类典型运行状态以表格的形式表示如下：
Figure PCTCN2018102907-appb-000001
当光伏阵列处于不同的运行状态时，光伏阵列中的两个支路以及整个光伏阵列的U-I输出特性曲线和U-P输出特性曲线均不相同，以U-I输出特性曲线为例：光伏阵列处于状态标签F1对应的运行状态时的U-I输出特性曲线如图5-a所示，光伏阵列处于状态标签F2对应的运行状态时的U-I输出特性曲线如图5-b所示，光伏阵列处于状态标签F5对应的运行状态时的U-I输出特性曲线如图5-c所示，光伏阵列处于状态标签F8对应的运行状态时的U-I输出特性曲线如图5-d所示，光伏阵列处于状态标签F13对应的运行状态时的U-I输出特性曲线如图5-e所示。需要说明的是，上述各个图中由于曲线重合，使得曲线上的标记有所叠加，比如支路B和阵列的曲线重合会导致图上显示六角形图示，但本领域技术人员应该可以从附图中明白其实际表示的含义，另外本申请对光伏阵列处于其他各个典型运行状态时的U-I输出特性曲线不一一示例。
第三步，构建故障特征向量。主要在光伏阵列处于各个典型运行状态时，分别采集光伏阵列中的各个支路和总干路对应的电路参数组，光伏组件所处的光照强度和温度的区间大致分别在[800,900]W/m 2和[40,50]℃。每个电路参数组中包括k个电路参数，k为正整数，在本申请中，k＝4，本申请采集开路电压V oc、短路电流I sc、最大功率点电压V m和最大功率点电流I m这四个参数构成电路参数组，这是由于这四个参数能够很好的刻画光伏组件的U-I输出特性曲线。 在光伏阵列中，共采集得到n+1个电路参数组，每个电路参数组中包括k个电路参数，因此可以根据采集到的电路参数组构建得到一个p维的故障特征向量，其中，p＝k*(n+1)。在本申请举例的3*2光伏阵列中，主要有两个支路和一个总干路，因此需要采集两个支路和总干路的V oc、I sc、V m和I m，因此会得到一个12维的故障特征向量。
第四步，得到故障特征向量后，就能汇总并构建成光伏阵列的数据样本集，由故障特征向量汇总得到数据样本集的方法是本领域技术人员都知道的方法，本申请对此不进行展开叙述。对数据样本集进行划分得到训练样本集和测试样本集，通常可以按照2:1的比例划分成为训练样本集和测试样本集。
第五步，基于随机森林算法的光伏阵列故障诊断模型的构建，设训练样本集为X，p是故障特征向量的维度，在本申请举例的3*2光伏阵列中，p＝12；s是故障诊断中需要建立的决策树的数目，s≥2且s为正整数，则基于随机森林算法利用训练样本集构建得到光伏阵列故障诊断模型的步骤可以简述为：
1、对训练样本集X进行s次有放回抽样，每次抽样得到一个子训练样本集，共得到s个子训练样本集，记第i个子训练样本集为X i，i为参数，i＝1,2……,s。
2、利用每个子训练样本集训练得到一个决策树，也即利用子训练样本集X i训练决策树Q i，在进行节点分裂时，从p个故障特征中随机选择q个故障特征，q为正整数且q≤p，计算q个故障特征中的各个故障特征对应的基尼系数，将与最小的基尼系数对应的故障特征作为最佳分裂特征来进行节点分裂，基尼系数的计算方法可以采用现有的计算公式，本申请对此不作详细介绍。
3、将s个决策树Q i(i＝1,2……,s)进行汇总，汇总得到的集合即为光伏阵列故障诊断模型。
第六步，光伏阵列故障诊断模型的测试。利用测试样本集对光伏阵列故障诊断模型进行测试，可以最终得到诊断精度、平均训练时间和测试时间等指标。将本申请基于随机森林(RF)算法的故障诊断与基于神经网络(BP)和极限学习机(ELM)的故障诊断进行对比，实际实验结果如下表所示。从表中可以看出，基于随机森林(RF)的故障诊断方法的诊断精度较高，同时训练和测试时间的表现也较好。
诊断方法 训练时间/s 测试时间/s 诊断精度/％
BP 2.6664 0.10852 96.92
ELM 0.3449 0.00663 97.69
RF 0.3718 0.01052 98.46
第七步，利用测试完成的光伏阵列故障诊断模型对待诊断光伏阵列进行诊断，也即采用光伏阵列故障诊断模型中的s个决策树进行投票，由于随机森林的诊断结果是由投票产生的，因此会得到s个决策树对各个典型运行状态的投票结果。根据各个投票结果即可以得到待诊断光伏阵列的故障诊断结果，也即确定待诊断光伏阵列中各个支路的运行状态。通常情况下，票数最多的典型运行状态即为待诊断光伏阵列的实时运行状态，也即可以确定光伏阵列出现的故障类型。以20棵决策树为例，5组实测结果的部分投票结果汇总，以及对应的故障诊断结果如下表，表中以状态标签进行表示：
序号 投票结果 故障诊断结果 故障真值
1 F2：19票、F11:1票 F2 F2
2 F5：17票、F10：3票 F5 F5
3 F11：10票、F7：5票、F1：2票、F4：1票、F6：1票 F11 F7
4 F9：11票、F8：9票 F9 F8
5 F11：15票、F2：4票、F4：1票 F11 F11
如上表可以看出，对于第一组数据，状态标签F2对应的典型运行状态对应的票数为19票为最多，则诊断结果为待诊断光伏阵列处于状态标签F2对应的典型运行状态，也即诊断得到待诊断光伏阵列的一个支路中有一个光伏组件短路，并且这一诊断结果与故障真值是相符的。
对于第三组数据，状态标签F11对应的典型运行状态对应的票数为10票为最多，则诊断结果为待诊断光伏阵列处于状态标签F10对应的典型运行状态，也即诊断得到待诊断光伏阵列的一个支路开路、另一个支路中有一个光伏组件存在阴影，虽然这一诊断结果与故障真值不符，但由于本申请会得到各个典型运行状态对应的投票结果，因此即使发生了这种诊断错误的情况，实际运行状态的类型也会出现在投票统计中(可以看到状态标签F7对应的典型运行状态对应的票数为5票，票数第二多)，相当于有备选诊断结果，所以工作人员能对各个典型运行状态的票数进行调取并查看票数统计情况，从而找出可能发生故障的类型，即使是最高票对应的诊断发生了错误，仍可以按票数高低依次进行故障排除，也即本申请实际是提供了一个诊断路线，从而提高维修效率，缩短系统的故障时间。
以上所述的仅是本申请的优选实施方式，本发明不限于以上实施例。可以理解，本领域技术人员在不脱离本发明的精神和构思的前提下直接导出或联想到的其他改进和变化，均应认为包含在本发明的保护范围之内。

Claims (5)

  1. 一种基于随机森林算法的光伏阵列故障诊断方法，其特征在于，所述方法包括：
    确定光伏阵列在运行过程中的典型运行状态，所述光伏阵列包括n个支路，每个支路包括m个光伏组件，m和n均为正整数且m≤4、n≤4；
    在所述光伏阵列处于各个所述典型运行状态时，分别采集所述光伏阵列中的各个支路和总干路对应的电路参数组，每个电路参数组中包括k个电路参数，k为正整数，根据采集到的n+1个电路参数组构建得到一个p维的故障特征向量，其中，p＝k*(n+1)；
    根据获取到的所述故障特征向量构建所述光伏阵列的数据样本集，将所述数据样本集划分为训练样本集和测试样本集；
    基于随机森林算法利用所述训练样本集构建得到光伏阵列故障诊断模型，并利用所述测试样本集对所述光伏阵列故障诊断模型进行测试，所述光伏阵列故障诊断模型中包括s个决策树，s≥2且s为正整数；
    利用测试完成的所述光伏阵列故障诊断模型对待诊断光伏阵列进行诊断，得到所述s个决策树对各个所述典型运行状态的投票结果；
    根据各个所述典型运行状态的投票结果得到所述待诊断光伏阵列的故障诊断结果，所述故障诊断结果用于指示所述光伏阵列中的各个支路的运行状态。
  2. 根据权利要求1所述的方法，其特征在于，所述基于随机森林算法利用所述训练样本集构建得到光伏阵列故障诊断模型，包括：
    对所述训练样本集进行s次有放回抽样，每次抽样得到一个子训练样本集，共得到s个子训练样本集；
    利用每个所述子训练样本集训练得到一个决策树；
    确定训练得到的s个决策树汇总得到的集合即为所述光伏阵列故障诊断模型。
  3. 根据权利要求2所述的方法，其特征在于，
    在利用子训练样本集训练决策树的过程中，在进行节点分裂时，从p个故障特征中随机选择q个故障特征，q为正整数且q≤p，计算所述q个故障特征中的各个故障特征对应的基尼系数，将与最小的基尼系数对应的故障特征作为最 佳分裂特征来进行节点分裂。
  4. 根据权利要求1所述的方法，其特征在于，k＝4，每个所述电路参数组中包括开路电压、短路电流、最大功率点电压和最大功率点电流。
  5. 根据权利要求1所述的方法，其特征在于，n＝2、m＝3，所述光伏阵列的典型运行状态包括五大类共十二小类，分别为：
    第一大类为正常运行状态，所述第一大类包括一小类，为所述光伏阵列中的各个支路中的各个光伏组件均正常运行时的状态；
    第二大类为短路故障状态，所述第二大类包括三小类，分别为所述光伏阵列中的一个支路中有一个光伏组件短路时的状态，以及所述光伏阵列中的一个支路中有两个光伏组件短路时的状态，以及所述光伏阵列中的两个支路均有一个光伏组件短路时的状态；
    第三大类为开路故障状态，所述第三大类包括一小类，为所述光伏阵列中有一个支路开路时的状态；
    第四大类为阴影故障状态，所述第四大类包括三小类，分别为所述光伏阵列中的一个支路中有一个光伏组件存在阴影时的状态，以及所述光伏阵列中的两个支路中均有一个光伏组件存在阴影时的状态，以及所述光伏阵列中的一个支路中有一个光伏组件存在阴影、另一个支路中有两个光伏组件存在阴影时的状态，以及所述光伏阵列中的两个支路中均有两个光伏组件存在阴影时的状态；
    第五大类为混合故障状态，所述第五大类包括四小类，分别为所述光伏阵列中的一个支路开路、另一个支路中有一个光伏组件存在阴影时的状态，以及所述光伏阵列中的一个支路中有一个光伏组件短路、另一个支路中有一个光伏组件存在阴影时的状态，以及所述光伏阵列中的两个支路均有一个光伏组件短路、一个光伏组件存在阴影时的状态，以及所述光伏阵列中的两个支路均有一个光伏组件短路、两个光伏组件存在阴影时的状态。
PCT/CN2018/102907 2018-08-10 2018-08-29 一种基于随机森林算法的光伏阵列故障诊断方法 WO2020029344A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/557,220 US11114977B2 (en) 2018-08-10 2019-08-30 Photovoltaic array fault diagnosis method based on random forest algorithm

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810907589.1 2018-08-10
CN201810907589.1A CN109150104A (zh) 2018-08-10 2018-08-10 一种基于随机森林算法的光伏阵列故障诊断方法

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/557,220 Continuation US11114977B2 (en) 2018-08-10 2019-08-30 Photovoltaic array fault diagnosis method based on random forest algorithm

Publications (1)

Publication Number Publication Date
WO2020029344A1 true WO2020029344A1 (zh) 2020-02-13

Family

ID=64792591

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/102907 WO2020029344A1 (zh) 2018-08-10 2018-08-29 一种基于随机森林算法的光伏阵列故障诊断方法

Country Status (3)

Country Link
US (1) US11114977B2 (zh)
CN (1) CN109150104A (zh)
WO (1) WO2020029344A1 (zh)

Families Citing this family (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109861644B (zh) * 2018-12-24 2020-07-07 阳光电源股份有限公司 光伏组件故障诊断方法、边缘计算处理装置和逆变器
CN109766952B (zh) * 2019-01-21 2022-08-12 福州大学 基于偏最小二乘法和极限学习机的光伏阵列故障检测方法
CN109977998B (zh) * 2019-02-14 2022-05-03 网易（杭州）网络有限公司 信息处理方法及装置、存储介质和电子装置
CN110705727B (zh) * 2019-09-30 2022-03-01 山东建筑大学 基于随机森林算法的光伏电站阴影遮挡诊断方法及系统
CN111209955A (zh) * 2020-01-03 2020-05-29 山东超越数控电子股份有限公司 基于深度神经网络和随机森林的飞机电源系统故障识别方法
CN111245365B (zh) * 2020-01-09 2020-12-15 南京工业大学 基于动态加权深度森林的光伏组件故障诊断方法
CN111539550B (zh) * 2020-03-13 2023-08-01 远景智能国际私人投资有限公司 光伏阵列工作状态的确定方法、装置、设备及存储介质
TWI741727B (zh) * 2020-08-07 2021-10-01 國立臺灣科技大學 基於電性時序波形的光伏陣列故障診斷方法
KR102448187B1 (ko) * 2020-10-15 2022-09-29 상명대학교산학협력단 I-v 곡선의 단위 벡터해석법을 이용한 pv 패널의 고장 감지 방법
CN112787591B (zh) * 2020-10-30 2022-05-13 福州大学 基于微调密集连接卷积神经网络的光伏阵列故障诊断方法
CN112508889B (zh) * 2020-11-26 2022-09-13 郑州大学 一种染色体核型分析系统
CN112710956B (zh) * 2020-12-17 2023-08-04 四川虹微技术有限公司 一种基于专家系统的电池管理系统故障检测系统及方法
CN112668806B (zh) * 2021-01-17 2022-09-06 中国南方电网有限责任公司 一种基于改进随机森林的光伏功率超短期预测方法
CN113340625B (zh) * 2021-04-21 2023-01-06 北京交通大学 转向架故障诊断方法
CN113867307B (zh) * 2021-08-20 2022-10-21 天津大学 基于深度神经网络的航天器智能故障诊断方法
CN114336742B (zh) * 2021-12-22 2023-10-27 浙江大学 一种基于自适应神经网络的太阳能光伏控制器
CN114614766B (zh) * 2022-01-31 2022-11-29 扬州晶华新能源科技有限公司 一种光伏光热一体化组件异常检测方法及测试系统
CN116310859B (zh) * 2023-01-16 2023-09-12 自然资源部国土卫星遥感应用中心 一种基于多源遥感数据的光伏阵列故障智能检测方法
CN116029703B (zh) * 2023-03-28 2023-06-16 山东科技大学 一种分布式光伏清洁周期确定方法、系统、介质及设备
CN116429339B (zh) * 2023-06-07 2023-09-01 海斯坦普汽车组件(北京)有限公司 一种新能源电池盒泄露检测方法及系统
CN117216481B (zh) * 2023-09-28 2024-10-11 浙江康巴赫科技股份有限公司 一种用于电器的远程监控方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105141255A (zh) * 2015-08-12 2015-12-09 河海大学常州校区 一种光伏阵列故障诊断方法
CN106961249A (zh) * 2017-03-17 2017-07-18 广西大学 一种光伏阵列故障诊断和预警方法
CN107579707A (zh) * 2017-10-13 2018-01-12 江苏大学 一种基于参数辨识的光伏阵列故障诊断方法
CN108062571A (zh) * 2017-12-27 2018-05-22 福州大学 基于差分进化随机森林分类器的光伏阵列故障诊断方法

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104155596B (zh) * 2014-08-12 2017-01-18 北京航空航天大学 一种基于随机森林的模拟电路故障诊断系统
US20180038994A1 (en) * 2016-08-02 2018-02-08 International Business Machines Corporation Techniques to Improve Global Weather Forecasting Using Model Blending and Historical GPS-RO Dataset
CN107274105B (zh) * 2017-06-28 2020-10-13 山东大学 基于线性判别分析的多属性决策树电网稳定裕度评估方法
CN108303632A (zh) * 2017-12-14 2018-07-20 佛山科学技术学院 基于随机森林算法的电路故障诊断方法
US10998853B2 (en) * 2018-04-09 2021-05-04 International Business Machines Corporation Internet of things-enabled solar photovoltaic health monitoring and advising related thereto

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105141255A (zh) * 2015-08-12 2015-12-09 河海大学常州校区 一种光伏阵列故障诊断方法
CN106961249A (zh) * 2017-03-17 2017-07-18 广西大学 一种光伏阵列故障诊断和预警方法
CN107579707A (zh) * 2017-10-13 2018-01-12 江苏大学 一种基于参数辨识的光伏阵列故障诊断方法
CN108062571A (zh) * 2017-12-27 2018-05-22 福州大学 基于差分进化随机森林分类器的光伏阵列故障诊断方法

Also Published As

Publication number Publication date
CN109150104A (zh) 2019-01-04
US20190386611A1 (en) 2019-12-19
US11114977B2 (en) 2021-09-07

Similar Documents

Publication Publication Date Title
WO2020029344A1 (zh) 一种基于随机森林算法的光伏阵列故障诊断方法
WO2020029327A1 (zh) 一种基于改进随机森林算法的光伏阵列故障诊断方法
Natarajan et al. Fault detection of solar PV system using SVM and thermal image processing
CN104391189A (zh) 基于三级诊断的大型光伏阵列故障诊断定位方法
Dhimish et al. Novel open-circuit photovoltaic bypass diode fault detection algorithm
CN113037214B (zh) 一种基于s-v曲线的光伏组件阴影遮挡故障诊断方法
Djalab et al. Robust method for diagnosis and detection of faults in photovoltaic systems using artificial neural networks
Reddy et al. A MATLAB based PV module models analysis under conditions of nonuniform irradiance
Chine et al. ANN-based fault diagnosis technique for photovoltaic stings
Ayesh et al. Design of wireless sensor network for monitoring the performance of photovoltaic panel
Aallouche et al. Online fault detection and identification for an isolated PV system using ANN
Al-Kashashnehand et al. Wireless Sensor Network Based Real-Time Monitoring and Fault Detection for Photovoltaic Systems
Guerriero et al. Innovative algorithm for true maximum detection based on individual PV panel sensor network
Nilsson Fault detection in photovoltaic systems
Alajmi et al. Fault detection and localization in solar photovoltaic arrays framework: hybrid methods of data-analysis and a network of voltage-current sensors
Gong et al. Fault diagnosis method of photovoltaic array based on random forest algorithm
Compaore et al. Analysis of the impact of faults in a photovoltaic generator
Fezzani et al. Modeling and analysis of the photovoltaic array faults
Hachana et al. Fault diagnosis of building integrated PV generator: a metaheuristic technique
Sapountzoglou et al. A grid connected pv system fault diagnosis method
CN113872526B (zh) 基于最小不匹配度故障电流预测的光伏阵列故障诊断方法
Bendale et al. Deep Learning for Solar Panel Maintenance: Detecting Faults and Improving Performance
Ghazali et al. A multi-scale dual-stage model for PV array fault detection, classification, and monitoring technique
Tian et al. Design of photovoltaic array fault online evaluation system
Dhimish et al. Fault detection algorithm for multiple GCPV array configurations

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18929235

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18929235

Country of ref document: EP

Kind code of ref document: A1