WO2023169098A1 - 基于孤立森林的模块化多电平换流器开路故障诊断方法 - Google Patents

基于孤立森林的模块化多电平换流器开路故障诊断方法 Download PDF

Info

Publication number
WO2023169098A1
WO2023169098A1 PCT/CN2023/073937 CN2023073937W WO2023169098A1 WO 2023169098 A1 WO2023169098 A1 WO 2023169098A1 CN 2023073937 W CN2023073937 W CN 2023073937W WO 2023169098 A1 WO2023169098 A1 WO 2023169098A1
Authority
WO
WIPO (PCT)
Prior art keywords
sub
module
isolated
power switch
circuit fault
Prior art date
Application number
PCT/CN2023/073937
Other languages
English (en)
French (fr)
Inventor
邓富金
陈宇飞
刘诚恺
Original Assignee
东南大学
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 东南大学 filed Critical 东南大学
Publication of WO2023169098A1 publication Critical patent/WO2023169098A1/zh

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
    • G01R31/54—Testing for continuity
    • 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
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
    • Y04S10/52—Outage or fault management, e.g. fault detection or location

Definitions

  • the present invention relates to the technical field of multi-level power electronic converters, specifically a modular multi-level converter open-circuit fault diagnosis method based on an isolated forest.
  • Modular Multilevel Converter With the continuous development and progress of modern science and technology, Modular Multilevel Converter (MMC) is known for its high efficiency, small output AC voltage harmonics, low switching frequency, high redundancy, and high degree of modularity. , flexible operation and other characteristics, it is more suitable for high voltage and high power applications, and has broad application prospects in fields such as flexible DC transmission and high-power motor drive.
  • MMC Modular Multilevel Converter
  • MMC is composed of a large number of cascaded submodules (SMs). By dividing large-capacity DC capacitors into smaller capacitors, series-connected SMs can be used in MMC. This unique structure determines that MMC does not have large-capacity DC capacitors, has lower switching synchronization requirements and has a higher degree of modularity. However, a large number of series-connected sub-modules pose a huge threat to the reliability of the MMC, because each SM is a potential failure point of the MMC. Once a failure occurs, it will seriously affect the stable operation of the MMC system and equipment safety.
  • SMs cascaded submodules
  • Submodule fault is one of the main sources of MMC faults, and there are two main types: submodule open circuit (OC) fault and submodule short circuit (SC) fault.
  • OC submodule open circuit
  • SC submodule short circuit
  • Short-circuit faults are highly destructive, so sub-module drive circuits are generally equipped with short-circuit protection modules. When a short-circuit fault occurs, the sub-module is locally blocked to ensure that the system can still operate normally.
  • the harm of open circuit fault is relatively small, so it is not easy to be detected immediately, causing voltage and current waveform distortion and other consequences, threatening the normal operation of the system.
  • Submodule open circuit (OC) fault diagnosis usually requires fault detection and fault location.
  • an error alarm is generated to notify the submodule of the occurrence of an open circuit (OC) fault.
  • OC open circuit
  • fault location will be used to identify the faulty SM to ensure correct configuration of the topology.
  • the purpose of the present invention is to provide a modular multi-level converter open-circuit fault diagnosis method based on an isolated forest, which solves the above technical problems and simultaneously realizes sub-module open-circuit fault detection and location without additional hardware resources.
  • the open-circuit fault diagnosis method of modular multi-level converter based on isolated forest includes the following steps:
  • Step 1 Sampling the infinite flow signal of the capacitor voltage and time of the modular multi-level converter, the sampling frequency is f s and the sampling interval is T s ;
  • Step 2 Based on the capacitor voltage data sampled at each moment, construct an isolated tree for the capacitor voltage data of one bridge arm;
  • Step 3 Based on the constructed isolated tree, calculate the depth D(i) of each sub-module SMi in the isolated tree;
  • Step 4 Every T s is a sampling moment, m isolated trees constructed based on m consecutive m sampling moments form an isolated forest, and calculate the average depth AD(i) of each sub-module SMi in the isolated forest;
  • Step 5 Take the sequence number i of the submodule with the smallest average depth AD(i) in the isolated forest as the output of the current isolated forest, record it as IFO, and store it in an output buffer that can accommodate k IFOs in an orderly manner;
  • Step 6 Output the sub-module fault location flag Flag according to the output buffer, and determine whether a fault occurs.
  • the isolated tree is a non-linear data structure with a certain number of layers, used to classify sub-modules according to the relationship between the capacitance and voltage of the sub-modules.
  • the construction method of the isolated tree is as follows: the root node N 0 of the 0th layer contains n sub-modules of a bridge arm and corresponding n capacitor voltage values. Starting from the root node, a voltage division value u 0 is randomly selected, and The sub-modules with capacitance voltage less than or equal to u 0 are divided into node N 1_1 , and the sub-modules with capacitance voltage greater than u 0 are divided into node N 1_2 ; repeat the above process for N 1_1 , N 1_2 and each subsequent node N until all nodes cannot be subdivided, then an isolated tree is constructed. At this time, the isolated tree contains n external nodes that cannot be subdivided. Each external node only contains one sub-module, expressed as TN(SMi), where SMi corresponds to the bridge arm. The i-th submodule in and 1 ⁇ i ⁇ n.
  • the depth D(i) in step 3 is defined as: the number of layers in the isolated tree of the external node where the sub-module SMi is located, and the calculation formula is:
  • the average depth AD(i) of each sub-module SMi in m isolated trees in step 4 is calculated as:
  • D(i,j) represents the depth of the external node where the sub-module SMi is located in the j-th (1 ⁇ j ⁇ m) isolated tree of the isolated forest.
  • the working mode of the output buffer in step 5 is: update the data in the output buffer every mT s , delete the earliest generated IFO, and add the latest generated IFO at the same time. into IFO.
  • the sub-module is a half-bridge structure, including two power switches Su and Sl , two diodes Du and Dl and a DC capacitor C, wherein the power switch Su and the diode Du form the upper tube , the power switch S l and the diode D l form the lower tube; the cathode of the diode Du is connected to the collector of the power switch Su , the anode of the diode Du is connected to the emitter of the power switch Su, and the cathode of the diode D l is connected to the power switch S The collector of l and the anode of diode D l are connected to the emitter of power switch S l .
  • the emitter of power switch Su and the collector of power switch S l are respectively connected to the current inflow side of the submodule bridge arm.
  • the power switch Su The gate and the gate of the power switch S l are respectively connected to the control circuit that controls the turning on and off of the power switch.
  • the emitter of the power switch S l is connected to the current outflow side of the sub-module bridge arm.
  • the collector of the power switch S u is connected to the DC current.
  • the capacitor is connected to the current outflow side of the submodule bridge arm.
  • step 6 the specific method for determining the fault in step 6 is:
  • Flag 1
  • the converter open-circuit fault diagnosis method proposed by the present invention can realize fault detection and positioning at the same time, and has high practical value; it gets rid of the separation of open-circuit fault detection and positioning in the traditional sub-module open-circuit fault diagnosis algorithm, making fault diagnosis Problems that complicate the process and prolong fault diagnosis time;
  • the open-circuit fault diagnosis method of the present invention does not involve system parameters and does not require the construction of a system mathematical model and the artificial setting of empirical thresholds. Therefore, it is not affected by the uncertainty of system parameters and has high robustness;
  • the open circuit fault diagnosis method of the present invention does not require any changes to the hardware circuit, does not increase additional hardware costs, and is easy to implement;
  • the open circuit fault diagnosis method of the present invention does not need to introduce circulating current in the system and does not change the output characteristics of the system. sex;
  • the open-circuit fault diagnosis method of the present invention is based on unsupervised learning and uses the sparseness and difference of abnormal data to locate faults. It does not require a large amount of data analysis and sample training, and has linear time complexity, small data volume, simple calculation process, and high efficiency. Low cost and other advantages.
  • Figure 1 is a topological structure diagram of the three-phase MMC and sub-modules in the present invention
  • Figure 2 is a specific implementation flow chart of the fault diagnosis method of the present invention.
  • Figure 3 is an example diagram of an isolated tree (IT) constructed in the present invention.
  • the present invention proposes a sub-module fault diagnosis method suitable for MMC.
  • the MMC topology consists of six bridge arms. As shown in Figure 1, each bridge arm contains n identical Submodule (SM) and a bridge arm inductor L s .
  • the submodule adopts a half-bridge structure.
  • Each submodule consists of two power switches Su and S l , two diodes Du and D l and a DC capacitor C. composition;
  • the capacitor voltage balancing method is: according to the comparison between the bridge arm reference voltage and the carrier, the number of sub-modules that need to be invested in a bridge arm is p. When the bridge arm current is greater than 0, the p sub-modules with the lowest capacitance voltage are put in. When the bridge arm current is less than 0, put in the p sub-module with the highest capacitance voltage.
  • an isolation forest-based modular multi-level converter open-circuit fault diagnosis method includes: sampling sub-module capacitor voltages, and constructing an isolated tree based on the capacitor voltage data sampled at each time to Describe the data characteristics of the submodule capacitance voltage and calculate the value of each submodule in the isolated tree. depth.
  • An isolated forest is formed based on the constructed isolated tree. The average depth of each sub-module in the isolated forest is calculated, and the sub-module with the smallest average depth is regarded as the current isolated forest output.
  • the faulty sub-module is located through the output buffer, which can be used in a short time. The faulty sub-module can be accurately located within a certain period of time. Specifically, it includes the following steps:
  • the sampling frequency is f s
  • Isolation Tree (IT) is constructed for a bridge arm.
  • the Isolation Tree is a nonlinear data structure with a certain number of layers, which is used to The submodules are classified according to the relationship between the capacitance and voltage of the submodules.
  • the construction method of the isolated tree is as follows: the root node N 0 of layer 0 contains n sub-modules of a bridge arm and corresponding n capacitor voltage values. Starting from the root node, a voltage division value u 0 (between the current time (between the maximum and minimum capacitance voltage of the submodule), the submodules with capacitance voltage less than or equal to u 0 are divided into node N 1_1 , and the submodules with capacitance voltage greater than u 0 are divided into node N 1_2 . Repeat the above process for N 1_1 , N 1_2 and each subsequent node N until all nodes cannot be further divided, then an isolated tree is constructed.
  • the isolated tree contains n external nodes (Terminal Node, TN), each external node contains only one sub-module, denoted as TN(SMi), where SMi corresponds to the i-th sub-module in the bridge arm and 1 ⁇ i ⁇ n.
  • Isolation Forest, IF isolation forest
  • AD(i) The average depth AD(i) in .
  • the present invention is particularly suitable for MMC systems with a large number of sub-modules. Compared with traditional sub-module fault diagnosis methods, it can significantly reduce the calculation amount of the diagnostic algorithm.
  • the proposed method analyzes the capacitor voltage and utilizes the sparsity and difference of abnormal capacitor voltage data for fault diagnosis. When a submodule has an open circuit (OC) fault, the capacitor voltage change of the faulty submodule will be different from that of the normal submodule. Therefore, the voltage change of the capacitor is monitored in the proposed method. Since only capacitor voltage is involved, the proposed method does not require additional hardware resources and does not increase additional hardware costs.
  • this method Since there are no system parameters involved, no system mathematical model needs to be constructed, and no empirical thresholds need to be set artificially, this method is not affected by the uncertainty of system parameters and has high robustness. Compared with other methods based on artificial intelligence, this method is based on unsupervised learning and does not require a large amount of data analysis and sample training. It has the advantages of linear time complexity, small amount of data, simple calculation process, and low calculation cost.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Tests Of Electronic Circuits (AREA)
  • Inverter Devices (AREA)
  • Testing Electric Properties And Detecting Electric Faults (AREA)

Abstract

一种基于孤立森林的模块化多电平换流器开路故障诊断方法，包括以下步骤：采样子模块电容电压，根据电容电压数据构建孤立树，计算每个子模块在孤立树中的深度；基于孤立树组成孤立森林，计算子模块在孤立森林中的平均深度，并将平均深度最小的子模块作为孤立森林输出，最后输出缓冲区定位故障子模块，实现故障子模块的准确定位。本方法仅涉及子模块电容电压，不需要额外的硬件资源；不涉及系统参数，不需要构建系统数学模型和人为设置经验阈值，具有很高的鲁棒性；本方法基于无监督学习，利用异常数据的稀疏性和差异性定位故障，无需大量数据分析和样本训练，具有线性时间复杂度、数据量小、计算过程简单、计算成本低等优点。

Description

基于孤立森林的模块化多电平换流器开路故障诊断方法 技术领域
本发明涉及多电平电力电子变换器技术领域，具体是基于孤立森林的模块化多电平换流器开路故障诊断方法。
背景技术
随着现代科学技术的不断发展与进步，模块化多电平换流器(Modular Multilevel Converter，MMC)以其效率高、输出交流电压谐波小、开关频率低、冗余度高、高度模块化、操作灵活等特点，更适用于高电压、大功率的应用场合，在柔性直流输电、高功率电机驱动等领域具有广阔的应用前景。
MMC由大量子模块(Submodule,SM)级联而成，通过将大容量直流电容分成较小的电容器，串联的SM可在MMC中应用。这种独特的结构决定了MMC没有大容量的直流电容，较低的开关同步需求和较高的模块化程度。但是，大量串联子模块对MMC可靠性构成巨大的威胁，因为每一个SM都是MMC的潜在故障点，一旦发生故障将严重影响MMC系统的稳定运行和设备安全。
MMC的可靠运行是该技术研究的关键之一。子模块故障是MMC故障的主要来源之一，主要有两种类型：子模块开路(OC)故障和子模块短路(SC)故障。短路故障破坏性较大，因而子模块驱动电路中一般配备了短路保护模块，当短路故障发生时，由本地闭锁该子模块，确保系统仍可正常运行。开路故障危害相对较小，因而不易被立即检测到，从而造成电压电流波形畸变等后果，威胁到系统的正常运行。子模块开路(OC)故障诊断通常需要故障检测和故障定位。在故障检测中，会生成错误警报以通知子模块开路(OC)故障的发生。为了在 子模块开路(OC)故障时实现MMC的不间断运行，将通过故障定位来识别故障SM，以确保正确配置拓扑。
目前学界提出了多种故障检测和定位方法。基于硬件的方法，如采用集成监控传感器和配备具有故障检测功能的驱动模块来实现SM的故障和定位，然而基于硬件的方法需要添加额外的故障检测电路，不仅增加了成本，又带来了新的潜在故障点。基于观测器的方法，如通过基于卡尔曼滤波器的观测器实现故障检测，通过电容电压的比较来进行故障定位；基于滑模观测器的分别执行用于故障检测和故障定位，该类方法需要构建MMC系统的精确的数学模型，并需要人为设置经验阈值。基于人工智能的方法，如基于滑动时间窗口与卷积神经网络的故障检测与定位方法，该类方法通常需要进行大量数据分析和样本训练，计算过程复杂且计算量大。
发明内容
本发明的目的在于提供基于孤立森林的模块化多电平换流器开路故障诊断方法，解决了上述技术问题，同时实现子模块开路故障检测和定位，无需额外的硬件资源。
本发明的目的可以通过以下技术方案实现：
基于孤立森林的模块化多电平换流器开路故障诊断方法，包括以下步骤：
步骤1：对模块化多电平换流器的电容电压与时间的无限长流量信号进行采样，采样频率为fs，采样间隔为Ts；
步骤2：基于每个时刻采样的电容电压数据，对一个桥臂的电容电压数据构建一棵孤立树；
步骤3：基于所构建的孤立树，计算孤立树中每个子模块SMi的深度D(i)；
步骤4：每隔Ts为一个采样时刻，基于连续m个采样时刻所构建的m棵孤立树，组成孤立森林，计算每个子模块SMi在孤立森林中的平均深度AD(i)；
步骤5：将孤立森林中平均深度AD(i)最小的子模块的序号i作为当前孤立森林的输出，记为IFO，有序存储在可容纳k个IFO的输出缓冲区中；
步骤6：根据输出缓冲区的情况输出子模块故障定位标志Flag，并判断是否出现故障。
进一步地，所述孤立树是一种非线性的数据结构，具有一定数量的层数，用以根据子模块电容电压大小关系对子模块进行分类。
所述孤立树的构建方法为：第0层的根节点N0包含一个桥臂的n个子模块以及相对应的n个电容电压值，从根节点起，随机选择一个电压分割值u0，将电容电压小于等于u0的子模块分入节点N1_1，将电容电压大于u0的子模块分入节点N1_2；对N1_1、N1_2及之后的每个节点N重复上述过程，直到所有节点都不可再分，则一棵孤立树构建完成，此时孤立树含有n个不可再分的外部节点，每个外部节点仅包含一个子模块，表示为TN(SMi)，其中SMi对应于桥臂中的第i个子模块且1≤i≤n。
进一步地，所述步骤1中采样频率fs＝100kHz，采样间隔Ts＝1ms。
进一步地，所述步骤3中深度D(i)，定义为：子模块SMi所在外部节点在孤立树中的层数，计算公式为：
D(i)＝IT[Level(TN(SMi))]。
进一步地，所述步骤4中m＝100。
进一步地，所述步骤4中每个子模块SMi在m棵孤立树中的平均深度AD(i)，计算公式为：
其中，D(i,j)表示孤立森林的第j(1≤j≤m)棵孤立树中子模块SMi所在外部节点的深度。
进一步地，所述步骤5中k＝5。
进一步地，所述步骤5中输出缓冲区的工作方式为：每隔mTs的时间对输出缓冲区中的数据进行一次更新，将其中最早生成的IFO删除，同时加入最新生 成的IFO。
进一步地，所述子模块为半桥结构，包括两个功率开关Su、Sl，两个二极管Du、Dl和一个直流电容C，其中，功率开关Su和二极管Du组成上管，功率开关Sl和二极管Dl组成下管；二极管Du的阴极连接功率开关Su的集电极，二极管Du的阳极连接功率开关Su的发射极，二极管Dl的阴极连接功率开关Sl的集电极，二极管Dl的阳极连接功率开关Sl的发射极，功率开关Su的发射极、功率开关Sl的集电极分别与子模块桥臂电流流入侧连接，功率开关Su的栅极、功率开关Sl的栅极分别与控制功率开关开通与关断的控制电路连接，功率开关Sl的发射极与子模块桥臂电流流出侧连接，功率开关Su的集电极经直流电容与子模块桥臂电流流出侧连接。
进一步地，所述步骤6中故障的具体判断方法为：
若输出缓冲区中的所有IFO都相同，则Flag＝1，对应的子模块SMi确定为故障子模块；否则，Flag＝0，系统视为正常，重复步骤1-5，继续对系统进行检测。
本发明的有益效果：
1、本发明提出的换流器开路故障诊断方法可同时实现故障检测和定位，具备很高的实用价值；摆脱了传统子模块开路故障诊断算法中将开路故障的检测和定位分离，使故障诊断过程复杂化，延长故障诊断时间的问题；
2、本发明开路故障诊断方法不涉及系统参数，无需构建系统数学模型和人为设置经验阈值，因此不受系统参数不确定性的影响，具有很高的鲁棒性；
3、本发明开路故障诊断方法不需要对硬件电路做任何改动，不增加额外的硬件成本，易于实施；
4、本发明开路故障诊断方法无需在系统中引入环流，不改变系统的输出特 性；
5、本发明开路故障诊断方法基于无监督学习，利用异常数据的稀疏性和差异性定位故障，无需进行大量的数据分析和样本训练，具有线性时间复杂度、数据量小、计算过程简单、计算成本低等优点。
附图说明
下面结合附图对本发明作进一步的说明。
图1是本发明中三相MMC及子模块拓扑结构图；
图2是本发明故障诊断方法的具体实施流程图；
图3是本发明中所构建的孤立树(IT)的示例图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其它实施例，都属于本发明保护的范围。
本发明针对子模块发生开路故障的问题，提出了适用于MMC的子模块故障诊断方法，其中MMC拓扑结构由六个桥臂组成，如图1所示，每个桥臂上包含了n个相同的子模块(Submodule，SM)以及一个桥臂电感Ls，子模块采用半桥结构，每个子模块由两个功率开关Su、Sl，两个二极管Du、Dl和一个直流电容C组成；
电容电压平衡方法为：根据桥臂参考电压与载波比较得到一个桥臂上需要投入的子模块个数为p，当桥臂电流大于0，投入电容电压最低的p个子模块，当桥臂电流小于0，投入电容电压最高的p个子模块。
如图2所示，一种基于孤立森林的模块化多电平换流器开路故障诊断方法，包括：采样子模块电容电压，并根据每个时刻采样得到的电容电压数据构建孤立树，用以描述子模块电容电压的数据特征，并计算每个子模块在孤立树中的 深度。基于所构建的孤立树组成孤立森林，计算每个子模块在孤立森林中的平均深度，并将平均深度最小的子模块视为当前孤立森林输出，最后通过输出缓冲区定位故障子模块，可在短时间内实现故障子模块的准确定位。具体包括以下步骤：
(1)对模块化多电平换流器的电容电压与时间的无限长流量信号进行采样，采样频率为fs，采样间隔为Ts(如fs＝100kHz，Ts＝1ms)。
(2)基于每个时刻采样的电容电压数据，对一个桥臂构建一棵孤立树(Isolation Tree，IT)，孤立树是一种非线性的数据结构，具有一定数量的层数，用以根据子模块电容电压大小关系对子模块进行分类。
孤立树的构建方法为：第0层的根节点N0包含一个桥臂的n个子模块以及相对应的n个电容电压值，从根节点起，随机选择一个电压分割值u0(介于当前时刻子模块电容电压最大值与最小值之间)，将电容电压小于等于u0的子模块分入节点N1_1，将电容电压大于u0的子模块分入节点N1_2。对N1_1、N1_2及之后的每个节点N重复上述过程，直到所有节点都不可再分，则一棵孤立树构建完成，此时孤立树含有n个不可再分的外部节点(Terminal Node，TN)，每个外部节点仅包含一个子模块，表示为TN(SMi)，其中SMi对应于桥臂中的第i个子模块且1≤i≤n。
(3)基于所构建的孤立树，计算孤立树中每个子模块SMi的深度D(i)。
(4)每隔Ts为一个采样时刻，基于连续m个采样时刻所构建的m棵孤立树(如m＝100)，组成孤立森林(Isolation Forest，IF)，计算每个子模块SMi在孤立森林中的平均深度AD(i)。
(5)将孤立森林中平均深度AD(i)最小的子模块的序号i作为当前孤立森林的输出，记为IFO，有序存储在可容纳k个IFO的输出缓冲区中(如k＝5)。
(6)根据输出缓冲区的情况输出子模块故障定位标志Flag，并判断是否出现故障。
子模块开路故障标准，具体如下：若输出缓冲区中的所有IFO都相同，则Flag＝1，对应的子模块SMi确定为故障子模块；反之，Flag＝0，系统视为正常， 子模块正常工作，未出现开路故障。
本发明尤其适用于子模块数目众多的MMC系统，与传统的子模块故障诊断方法相比，其能显著减小诊断算法的计算量。所提出的方法对电容电压进行分析，利用异常电容电压数据的稀疏性和差异性进行故障诊断。在子模块开路(OC)故障时，故障子模块的电容电压变化会与正常子模块有所不同。因此，在提出的方法中监测电容的电压变化。由于仅涉及电容电压，因此所提出的方法不需要额外的硬件资源，不增加额外的硬件成本。由于不涉及系统参数，不需要构建系统数学模型，不需要人为设置经验阈值，因此该方法不受系统参数不确定性的影响，并且具有很高的鲁棒性。与其他基于人工智能的方法相比，该方法基于无监督学习，不需要进行大量数据分析和样本训练，具有线性时间复杂度、数据量小、计算过程简单、计算成本低等优点。
在本说明书的描述中，参考术语“一个实施例”、“示例”、“具体示例”等的描述意指结合该实施例或示例描述的具体特征、结构、材料或者特点包含于本发明的至少一个实施例或示例中。在本说明书中，对上述术语的示意性表述不一定指的是相同的实施例或示例。而且，描述的具体特征、结构、材料或者特点可以在任何的一个或多个实施例或示例中以合适的方式结合。
以上显示和描述了本发明的基本原理、主要特征和本发明的优点。本行业的技术人员应该了解，本发明不受上述实施例的限制，上述实施例和说明书中描述的只是说明本发明的原理，在不脱离本发明精神和范围的前提下，本发明还会有各种变化和改进，这些变化和改进都落入要求保护的本发明范围内。

Claims (10)

  1. 基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，包括以下步骤：
    步骤1：对模块化多电平换流器的电容电压与时间的无限长流量信号进行采样，采样频率为fs，采样间隔为Ts；
    步骤2：基于每个时刻采样的电容电压数据，对一个桥臂的电容电压数据构建一棵孤立树；
    步骤3：基于所构建的孤立树，计算孤立树中每个子模块SM i的深度D(i)；
    步骤4：每隔Ts为一个采样时刻，基于连续m个采样时刻所构建的m棵孤立树，组成孤立森林，计算每个子模块SM i在孤立森林中的平均深度AD(i)；
    步骤5：将孤立森林中平均深度AD(i)最小的子模块的序号i作为当前孤立森林的输出，记为IFO，有序存储在可容纳k个IFO的输出缓冲区中；
    步骤6：根据输出缓冲区的情况输出子模块故障定位标志Flag，并判断是否出现故障。
  2. 根据权利要求1所述基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，所述孤立树是一种非线性的数据结构，具有一定数量的层数，用以根据子模块电容电压大小关系对子模块进行分类；
    所述孤立树的构建方法为：第0层的根节点N0包含一个桥臂的n个子模块以及相对应的n个电容电压值，从根节点起，随机选择一个电压分割值u0，将电容电压小于等于u0的子模块分入节点N1_1，将电容电压大于u0的子模块分入节点N1_2；对N1_1、N1_2及之后的每个节点N重复上述过程，直到所有节点都不可再分，则一棵孤立树构建完成，此时孤立树含有n个不可再分的外部节点，每个外部节点仅包含一个子模块，表示为TN(SM i)，其中SM i对应于桥臂中的第i个子模块且1≤i≤n。
  3. 根据权利要求1所述基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，所述步骤1中采样频率fs＝100kHz，采样间隔Ts＝1ms。
  4. 根据权利要求1所述基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，所述步骤3中深度D(i)，定义为：子模块SM i所在外部节点在孤立树中的层数，计算公式为：
    D(i)＝IT[Level(TN(SM i))]。
  5. 根据权利要求1所述基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，所述步骤4中m＝100。
  6. 根据权利要求5所述基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，所述步骤4中每个子模块SM i在m棵孤立树中的平均深度AD(i)，计算公式为：
    其中，D(i,j)表示孤立森林的第j(1≤j≤m)棵孤立树中子模块SM i所在外部节点的深度。
  7. 根据权利要求1所述基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，所述步骤5中k＝5。
  8. 根据权利要求1所述基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，所述步骤5中输出缓冲区的工作方式为：每隔mTs的时间对输出缓冲区中的数据进行一次更新，将其中最早生成的IFO删除，同时加入最新生成的IFO。
  9. 根据权利要求1所述基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，所述子模块为半桥结构，包括两个功率开关Su、Sl，两个二极管Du、Dl和一个直流电容C，其中，功率开关Su和二极管Du组成上管，功率开关Sl和二极管Dl组成下管；二极管Du的阴极连接功率开关Su的集电极，二极管Du的阳极连接功率开关Su的发射极，二极管Dl的阴极连接功率开关Sl的集电极，二极管Dl的阳极连接功率开关Sl的发射极，功率开关Su的发射极、功率开关Sl的集电极分别与子模块桥臂电流流入侧连接，功率开关Su的栅极、功率开关Sl 的栅极分别与控制功率开关开通与关断的控制电路连接，功率开关Sl的发射极与子模块桥臂电流流出侧连接，功率开关Su的集电极经直流电容与子模块桥臂电流流出侧连接。
  10. 根据权利要求1所述的基于孤立森林的模块化多电平换流器开路故障诊断方法，其特征在于，所述步骤6中故障的具体判断方法为：
    若输出缓冲区中的所有IFO都相同，则Flag＝1，对应的子模块SM i确定为故障子模块；否则，Flag＝0，系统视为正常，重复步骤1-5，继续对系统进行检测。
PCT/CN2023/073937 2022-03-10 2023-01-31 基于孤立森林的模块化多电平换流器开路故障诊断方法 WO2023169098A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210234148.6 2022-03-10
CN202210234148.6A CN114609546B (zh) 2022-03-10 2022-03-10 基于孤立森林的模块化多电平换流器开路故障诊断方法

Publications (1)

Publication Number Publication Date
WO2023169098A1 true WO2023169098A1 (zh) 2023-09-14

Family

ID=81862957

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/073937 WO2023169098A1 (zh) 2022-03-10 2023-01-31 基于孤立森林的模块化多电平换流器开路故障诊断方法

Country Status (2)

Country Link
CN (1) CN114609546B (zh)
WO (1) WO2023169098A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116995909A (zh) * 2023-09-27 2023-11-03 安徽大学 储能mmc系统开路故障诊断与容错运行的方法和系统
CN117077067A (zh) * 2023-10-18 2023-11-17 北京亚康万玮信息技术股份有限公司 一种基于智能匹配的信息系统自动部署规划方法
CN117407822A (zh) * 2023-12-12 2024-01-16 江苏新希望生态科技有限公司 一种全自动芽苗机及其控制方法
CN117874653A (zh) * 2024-03-11 2024-04-12 武汉佳华创新电气有限公司 一种基于多源数据的电力系统安全监测方法及系统

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114609546B (zh) * 2022-03-10 2023-07-28 东南大学 基于孤立森林的模块化多电平换流器开路故障诊断方法
CN117421687B (zh) * 2023-12-18 2024-03-15 圣道天德电气(山东)有限公司 一种数字化电力环网柜运行状态监测方法

Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110719044A (zh) * 2019-09-12 2020-01-21 东南大学 模块化多电平换流器子模块下管开路故障的定位方法
CN110988742A (zh) * 2019-10-31 2020-04-10 东南大学 基于四分位数的模块化多电平换流器开路故障诊断方法
US20200210854A1 (en) * 2018-12-27 2020-07-02 Utopus Insights, Inc. System and method for fault detection of components using information fusion technique
CN111707973A (zh) * 2020-06-04 2020-09-25 东南大学 基于格拉布斯准则的模块化多电平换流器开路故障诊断方法
CN112348071A (zh) * 2020-10-30 2021-02-09 浙江大学 一种模块化多电平换流器子模块开关管开路故障诊断方法
CN113341345A (zh) * 2021-06-04 2021-09-03 浙江大学 一种基于特征提取和随机森林的mmc开关管开路故障诊断方法
CN113420073A (zh) * 2021-08-23 2021-09-21 平安科技（深圳）有限公司 基于改进的孤立森林的异常样本检测方法及相关设备
CN113762350A (zh) * 2021-08-12 2021-12-07 南京星云数字技术有限公司 异常数据检测方法、装置、计算机设备和存储介质
US20210384816A1 (en) * 2020-06-03 2021-12-09 Delta Electronics (Shanghai) Co., Ltd. Modular multilevel converter system and voltage detection method and open-circuit fault diagnosis method thereof
WO2022043872A1 (en) * 2020-08-26 2022-03-03 Prescinto Technologies Private Limited System for monitoring under-performance of solar power plant
CN114609546A (zh) * 2022-03-10 2022-06-10 东南大学 基于孤立森林的模块化多电平换流器开路故障诊断方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100070212A1 (en) * 2008-09-15 2010-03-18 Caterpillar Inc. Method and apparatus for determining state of inverter capacitor in an electric drive
CN107271925B (zh) * 2017-06-26 2019-11-05 湘潭大学 基于深度卷积网络的模块化五电平换流器故障定位方法
CN110322368A (zh) * 2019-07-03 2019-10-11 厦门理工学院 一种谐波数据异常检测方法、终端设备及存储介质
CN112234576B (zh) * 2020-09-29 2022-07-05 东南大学 基于滑窗特征提取的模块化多电平换流器故障诊断方法
CN112611982B (zh) * 2020-11-30 2022-07-12 西华大学 模块化多电平换流器的深度小波孪生网络故障诊断方法
CN113159119A (zh) * 2021-03-11 2021-07-23 大连理工大学 一种中压模块化多电平变流器故障诊断方法
CN114089181B (zh) * 2021-11-19 2022-11-08 浙江大学 基于XG-Boost的MMC开关管开路故障检测方法

Patent Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200210854A1 (en) * 2018-12-27 2020-07-02 Utopus Insights, Inc. System and method for fault detection of components using information fusion technique
CN110719044A (zh) * 2019-09-12 2020-01-21 东南大学 模块化多电平换流器子模块下管开路故障的定位方法
CN110988742A (zh) * 2019-10-31 2020-04-10 东南大学 基于四分位数的模块化多电平换流器开路故障诊断方法
US20210384816A1 (en) * 2020-06-03 2021-12-09 Delta Electronics (Shanghai) Co., Ltd. Modular multilevel converter system and voltage detection method and open-circuit fault diagnosis method thereof
CN111707973A (zh) * 2020-06-04 2020-09-25 东南大学 基于格拉布斯准则的模块化多电平换流器开路故障诊断方法
WO2022043872A1 (en) * 2020-08-26 2022-03-03 Prescinto Technologies Private Limited System for monitoring under-performance of solar power plant
CN112348071A (zh) * 2020-10-30 2021-02-09 浙江大学 一种模块化多电平换流器子模块开关管开路故障诊断方法
CN113341345A (zh) * 2021-06-04 2021-09-03 浙江大学 一种基于特征提取和随机森林的mmc开关管开路故障诊断方法
CN113762350A (zh) * 2021-08-12 2021-12-07 南京星云数字技术有限公司 异常数据检测方法、装置、计算机设备和存储介质
CN113420073A (zh) * 2021-08-23 2021-09-21 平安科技（深圳）有限公司 基于改进的孤立森林的异常样本检测方法及相关设备
CN114609546A (zh) * 2022-03-10 2022-06-10 东南大学 基于孤立森林的模块化多电平换流器开路故障诊断方法

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116995909A (zh) * 2023-09-27 2023-11-03 安徽大学 储能mmc系统开路故障诊断与容错运行的方法和系统
CN116995909B (zh) * 2023-09-27 2023-12-29 安徽大学 储能mmc系统开路故障诊断与容错运行的方法和系统
CN117077067A (zh) * 2023-10-18 2023-11-17 北京亚康万玮信息技术股份有限公司 一种基于智能匹配的信息系统自动部署规划方法
CN117077067B (zh) * 2023-10-18 2023-12-22 北京亚康万玮信息技术股份有限公司 一种基于智能匹配的信息系统自动部署规划方法
CN117407822A (zh) * 2023-12-12 2024-01-16 江苏新希望生态科技有限公司 一种全自动芽苗机及其控制方法
CN117407822B (zh) * 2023-12-12 2024-02-20 江苏新希望生态科技有限公司 一种全自动芽苗机及其控制方法
CN117874653A (zh) * 2024-03-11 2024-04-12 武汉佳华创新电气有限公司 一种基于多源数据的电力系统安全监测方法及系统
CN117874653B (zh) * 2024-03-11 2024-05-31 武汉佳华创新电气有限公司 一种基于多源数据的电力系统安全监测方法及系统

Also Published As

Publication number Publication date
CN114609546B (zh) 2023-07-28
CN114609546A (zh) 2022-06-10

Similar Documents

Publication Publication Date Title
WO2023169098A1 (zh) 基于孤立森林的模块化多电平换流器开路故障诊断方法
US10234495B2 (en) Decision tree SVM fault diagnosis method of photovoltaic diode-clamped three-level inverter
WO2017041428A1 (zh) 二极管箝位型级联多电平换流器的建模方法及系统
CN110988742A (zh) 基于四分位数的模块化多电平换流器开路故障诊断方法
CN105675957A (zh) 一种基于状态监测的mmc模块电压测量和故障定位方法
CN111707973B (zh) 基于格拉布斯准则的模块化多电平换流器开路故障诊断方法
CN107609283A (zh) 基于桥臂等值电容的模块化多电平换流器高效建模方法
US20230083078A1 (en) Method for intelligent fault detection and location of power distribution network
Yuan et al. Open-circuit fault diagnosis of NPC inverter based on improved 1-D CNN network
Deng et al. Isolation forest based submodule open-circuit fault localization method for modular multilevel converters
CN112540261B (zh) 一种快速mmc子模块器件开路故障的检测方法及系统
Xin et al. A deep learning and softmax regression fault diagnosis method for multi-level converter
WO2022227415A1 (zh) 一种应用于同期装置的信号异常检测方法
CN108897908A (zh) 一种含均压控制功能的mmc实时仿真建模方法
CN116706863A (zh) 一种海底直流系统暂态稳定性优化方法
AU2021102724A4 (en) A method for modular multilevel converter (inverter) open-circuit fault identification and positioning
CN112234576B (zh) 基于滑窗特征提取的模块化多电平换流器故障诊断方法
CN114545133B (zh) 一种基于电流检测的单相级联h桥整流器的故障诊断方法
CN114879087A (zh) 一种用于vienna整流器开路故障综合辨识的方法
CN107994798B (zh) 一种含在线故障诊断的双向双buck逆变器及其工作方法
CN115097184A (zh) 基于无阈值递归图和卷积神经网络的电流信号分析方法
Hu et al. An Open Circuit Faults Diagnosis Method Based on XG-Boost Used in MMC
CN111413583A (zh) 一种配电网区段定位的实时线性整数规划方法
CN115267474B (zh) 一种基于特勒根定理的电力变流器故障诊断方法及系统
Ye et al. A fault diagnosis method for microgrid five-level inverter based on adversarial transfer learning

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: 23765678

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18561221

Country of ref document: US