WO2024104037A1 - 基于数学形态学和模式识别的直流电弧检测方法 - Google Patents

基于数学形态学和模式识别的直流电弧检测方法 Download PDF

Info

Publication number
WO2024104037A1
WO2024104037A1 PCT/CN2023/125029 CN2023125029W WO2024104037A1 WO 2024104037 A1 WO2024104037 A1 WO 2024104037A1 CN 2023125029 W CN2023125029 W CN 2023125029W WO 2024104037 A1 WO2024104037 A1 WO 2024104037A1
Authority
WO
WIPO (PCT)
Prior art keywords
pattern recognition
spectrum
arc detection
arc
mathematical morphology
Prior art date
Application number
PCT/CN2023/125029
Other languages
English (en)
French (fr)
Inventor
杨博
王世恩
张玉林
周旭
Original Assignee
上海正泰电源系统有限公司
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 上海正泰电源系统有限公司 filed Critical 上海正泰电源系统有限公司
Publication of WO2024104037A1 publication Critical patent/WO2024104037A1/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
    • 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
    • G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/10—Pre-processing; Data cleansing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
    • 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 invention relates to the field of direct current power system protection, and in particular to a direct current arc detection method based on mathematical morphology and pattern recognition.
  • arc faults may occur due to loose connectors, aging of cable materials, etc.
  • DC arcs tend to maintain stable combustion, and the heat generated can accumulate to produce temperatures of thousands of degrees, thus causing fires.
  • Photovoltaic power generation systems are widely used in DC power systems. Its typical structure is shown in Figure 1, which mainly includes photovoltaic cell arrays, photovoltaic grid-connected inverters and power grids. Photovoltaic cell arrays usually contain one or more photovoltaic cell strings ( Figure 1 only shows one), and photovoltaic grid-connected inverters usually include one or more MPPT units ( Figure 1 only shows one) and grid-connected inverters. There are many DC-side terminals in photovoltaic power generation systems, so the possibility of arc hazards is high, which is a key issue affecting the safe operation of the system.
  • fault arcs can be divided into three categories according to the location of the fault arc, including series arc, parallel arc and ground arc.
  • A is a series arc within the group
  • B is a parallel arc within the group
  • C and D are ground arcs.
  • the possibility of parallel arc and ground arc is relatively small, and the system current increases rapidly when it occurs, which can be detected using the current threshold method.
  • the possibility and frequency of series arc occurrence are the highest.
  • the DC voltage and DC current amplitude will not change significantly, which makes the detection of series arc extremely difficult.
  • DC arc detection and positioning methods can be divided into methods based on radiation characteristics and methods based on current characteristics.
  • a DC arc occurs, it is usually accompanied by heat, light, noise and electromagnetic radiation.
  • Arc detection based on radiation characteristics can be performed based on this.
  • this type of method is greatly affected by environmental factors and is usually suitable for confined spaces. Photovoltaic power generation systems have a large range and are usually outdoors, so this type of method is not applicable.
  • the arc can be equivalent to a nonlinear time-varying resistor, which produces disturbances in the current.
  • the method based on current characteristics collects current signals for a period of time and generates a disturbance in the current.
  • the signal is processed by fast Fourier transform, wavelet transform and other mathematical methods to extract the time domain characteristics, frequency domain characteristics and time-frequency domain characteristics of the signal.
  • Common time domain characteristics include the variance, peak-to-peak value, rate of change, kurtosis, waveform factor, entropy, etc. of the current; the impact of the arc on the spectrum is usually from several kilohertz to tens of kilohertz, and common frequency domain characteristics include the maximum value, energy proportion and variance of the characteristic frequency band; common time-frequency domain characteristics include the energy proportion of each wavelet, singular value, modulus maximum value and entropy.
  • the arc fault After extracting the characteristic value of the current signal, the arc fault can be judged according to the threshold value set in advance and the change of the characteristic value over time.
  • This type of method is greatly affected by the working state of the photovoltaic inverter, the threshold is difficult to set, and it is easy to cause false alarms or missed alarms.
  • the purpose of the present invention is to overcome the defects of the prior art and provide a DC arc detection method based on mathematical morphology and pattern recognition, which has the advantages of being less affected by the outside world, convenient detection and high accuracy.
  • the present invention adopts the following technical solutions:
  • a DC arc detection method based on mathematical morphology and pattern recognition comprises the following steps:
  • Step S1 establishing a training data set based on the acquired electrical quantities and working state data of the DC power system under arc fault and normal working states, and establishing and training an arc detection pattern recognition model based on a pattern recognition algorithm and the training data set;
  • Step S2 Acquire the electrical quantity of the DC power system in the operating state
  • Step S3 extracting the energy proportion of the base value component and/or the fluctuation component in each characteristic frequency band in the electrical quantity spectrum under the operating state based on mathematical transformation-mathematical morphology calculation; wherein the mathematical transformation-mathematical morphology calculation comprises the following steps:
  • Step a The obtained electrical quantities of the DC power system are transformed mathematically to obtain the frequency spectrum of the electrical quantities
  • Step b performing mathematical morphological operations on the spectrum of the electrical quantity to obtain the base value component and/or fluctuation component of the spectrum;
  • Step c according to the selected multiple characteristic frequency bands, respectively calculating the energy proportion of the base value component and/or the fluctuation component of the spectrum in each characteristic frequency band;
  • Step S4 inputting the energy proportion in the operating state into the arc detection pattern recognition model, and determining whether a series arc fault occurs through the arc detection pattern recognition model.
  • step S1 the arc detection pattern recognition model is established and trained based on a support vector machine or an artificial neural network algorithm and a training data set.
  • the calculation process of the arc detection pattern recognition model established and trained by the support vector machine based on the linear kernel function is simplified as follows: A T X+b
  • X is the input data vector
  • A is the coefficient vector with the same dimension as X
  • b is a constant.
  • the electrical quantity of the DC power system includes a DC current, a DC voltage, a component of a DC current or a component of a DC voltage.
  • step a the DC component is filtered out before obtaining the spectrum of the electrical quantity.
  • step a the average value of the electrical quantity is first obtained, the DC component is filtered out by subtracting the average value from each electrical quantity, and then a mathematical transformation is performed to obtain the frequency spectrum of the electrical quantity.
  • step a after the frequency spectrum of the electrical quantity is obtained, the frequency spectrum is normalized.
  • step a the effective value of the frequency spectrum of the electrical quantity is calculated, and the frequency point amplitude is divided by the effective value to obtain a normalized frequency spectrum.
  • the frequency spectrum of the electrical quantity is obtained by fast Fourier transform.
  • the spectrum is processed using an erosion operation, a dilation operation, or a composite operation of an erosion operation and a dilation operation in mathematical morphology.
  • step b a composite operation of an erosion operation and a dilation operation is performed on the spectrum or the normalized spectrum to obtain a base value component and/or a fluctuation component of the spectrum or the normalized spectrum.
  • step S3 20kHz-28kHz, 36kHz-44kHz and 52kHz-60kHz are selected as characteristic frequency bands for arc occurrence.
  • a threshold is set for the counter of the arc detection pattern recognition model.
  • the counter accumulates the count. If the count value of the counter is greater than the threshold, it is determined that an arc fault has occurred in the DC power system.
  • the arc detection pattern recognition model in step S1 is a pre-acquired recognition model; or, step S1 includes establishing a training data set based on mathematical transformation-mathematical morphology, and establishing and training an arc detection pattern recognition model based on a pattern recognition algorithm and the training data set, wherein the training data set includes the energy proportions under arc fault and normal working conditions.
  • step S1 the energy proportion of each characteristic frequency band is converted and calculated.
  • the training data set includes the characteristic values of each characteristic frequency band under the two states of arc fault and normal operation.
  • an arc detection pattern recognition model is established and trained.
  • the energy proportion of the base value component and/or fluctuation component of the characteristic frequency bands of multiple arcs is extracted from the electrical quantity spectrum and input into the arc detection pattern recognition model to determine whether a series arc fault has occurred. Since the base value component of the spectrum is removed by mathematical morphology, it is less affected by the fluctuation of the absolute value of signal sampling. Therefore, compared with the existing detection method of extracting the characteristic value of the current signal and comparing it with the threshold, it has the advantages of less external influence and high accuracy.
  • each characteristic frequency band is converted into the characteristic value of each characteristic frequency band, which is beneficial to model calculation and improves calculation accuracy.
  • FIG1 is a schematic diagram of a DC power system in the background art
  • FIG2 is a schematic diagram of a DC power system in the present invention.
  • FIG3 is a flow chart of an arc detection algorithm of the present invention.
  • FIG4 is a set of normal current waveform diagrams with the DC current component filtered out
  • FIG5 is a set of arc fault current waveforms with the DC current component filtered out
  • Figure 6 is the spectrum of normal current and fault current after FFT analysis
  • Figure 7 is the spectrum of the fluctuation components of normal current and fault current in the full frequency band
  • FIG8 is a distribution diagram of characteristic values of multiple groups of normal currents and fault currents in three frequency bands (20kHz-28kHz, 36kHz-44kHz, 52kHz-60kHz).
  • a DC arc detection method based on mathematical morphology and pattern recognition comprises the following steps:
  • Step S1 obtaining electrical quantities of a DC power system under arc fault and normal working conditions, selecting several characteristic frequency bands of arc occurrence, and calculating and extracting the energy proportion of the base value component and/or fluctuation component of the spectrum of the electrical quantity under arc fault and normal working conditions in each characteristic frequency band based on mathematical transformation-mathematical morphology, establishing a training data set, the training data set including the energy proportion under arc fault and normal working conditions, and establishing and training an arc detection pattern recognition model based on a pattern recognition algorithm and the training data set;
  • the mathematical transformation-mathematical morphology calculation comprises the following steps:
  • Step a The obtained electrical quantities of the DC power system are transformed mathematically to obtain the frequency spectrum of the electrical quantities
  • Step b performing mathematical morphological operations on the spectrum of the electrical quantity to obtain the base value component and/or fluctuation component of the spectrum;
  • Step c according to the selected multiple characteristic frequency bands, respectively calculating the energy proportion of the base value component and/or the fluctuation component of the spectrum in each characteristic frequency band;
  • Step S2 Acquire the electrical quantity of the DC power system in the operating state
  • Step S3 Based on the mathematical transformation-mathematical morphology of step S1, calculate and extract the energy proportion of the base value component and/or fluctuation component in each characteristic frequency band in the spectrum of the electrical quantity under the operating state;
  • Step S4 Input the energy proportion in the operating state into the arc detection pattern recognition model, and determine whether a series arc fault occurs through the arc detection pattern recognition model.
  • the present invention establishes and trains an arc detection pattern recognition model based on mathematical morphology and pattern recognition. By performing mathematical transformation-mathematical morphology calculation on electrical quantities, the energy proportions of the base value components and/or fluctuation components of the characteristic frequency bands of multiple arc occurrences are extracted from the electrical quantity spectrum and input into the arc detection pattern recognition model, thereby determining whether a series arc fault occurs. Since the base value component of the spectrum is removed by mathematical morphology, it is less affected by the fluctuation of the absolute value of signal sampling. Therefore, compared with the existing detection method of extracting the characteristic value of the current signal compared with the threshold, it has the advantages of less external influence and high precision.
  • FIGS. 2 and 3 A specific embodiment is provided in conjunction with FIGS. 2 and 3 .
  • the present invention can be applied in the fields of photovoltaic power generation systems, energy storage systems, DC microgrids, etc.
  • This embodiment takes a photovoltaic power generation system as an example.
  • the photovoltaic power generation system mainly includes a photovoltaic cell array, a photovoltaic grid-connected inverter and a power grid.
  • the photovoltaic cell array includes one or more photovoltaic cell strings
  • the photovoltaic grid-connected inverter includes one or more MPPT units. Yuanhe grid-connected inverter, as shown in Figure 2, the photovoltaic power generation system of this embodiment includes a photovoltaic battery string and a photovoltaic grid-connected inverter connected to the photovoltaic battery string.
  • the photovoltaic grid-connected inverter includes a switch S, an MPPT unit, an inverter and a controller. The switch S is connected between the photovoltaic battery string and the MPPT.
  • the controller includes a sampling module or a sampling circuit.
  • the electrical quantity data of the photovoltaic grid-connected inverter is collected through the sampling module or the sampling circuit.
  • the switch S of the photovoltaic grid-connected inverter is closed, and the controller collects the input current I pv of the MPPT unit, the input voltage is U pv , and the voltage output to the inverter is U bus .
  • the DC arc detection method based on mathematical morphology and pattern recognition includes the following steps:
  • Step S1 First, a large amount of electrical quantities of the DC power system under arc fault and normal working conditions and working state data of the DC power system are collected. Subsequently, the collected electrical quantities of the DC power system under arc fault and normal working conditions are calculated by mathematical transformation-mathematical morphology, and the base value component and/or fluctuation component are extracted from the electrical quantity spectrum under arc fault and normal working conditions. Several characteristic frequency bands of arc occurrence are selected, and the energy proportion of each characteristic frequency band is calculated.
  • the training data set includes the energy proportion of the fluctuating component under the two states of arc fault and normal operation, or the training data set includes the energy proportion of the base value component under the two states of arc fault and normal operation, or the training data set includes the energy proportion of the fluctuating component and the energy proportion of the base value component under the two states of arc fault and normal operation, and in addition, the training data set may also include the working state data of the DC power system;
  • an arc detection pattern recognition model is established and trained based on a pattern recognition algorithm and a training data set, and the pattern recognition algorithm may adopt algorithms such as support vector machines and artificial neural networks.
  • the training data set is preferably stored in a database, and of course it may also be stored in an XML text or other manner.
  • the arc detection pattern recognition model established and trained by the support vector machine based on the linear kernel function is preferably used to reduce the amount of calculation, and the calculation process of the model is simplified as follows: A T X+b
  • the mathematical transformation-mathematical morphology calculation method includes the following steps:
  • Step a The electrical quantities in the arc fault state and the electrical quantities in the normal working state are respectively transformed by mathematical transformation, that is, the transformation of the time domain and frequency domain of the electrical quantities, and the spectrum and frequency domain of the electrical quantities in the arc fault state are obtained accordingly.
  • the spectrum of the electrical quantity is represented by F.
  • the spectrum of the normal current and the fault current after FFT analysis is shown in FIG6 .
  • the DC component is first filtered out, specifically by obtaining the current average value I ave and subtracting I ave from each collected current value.
  • the DC filtering may also be omitted, because the step of removing the DC component is not necessary, but this step is beneficial to the subsequent signal processing process.
  • the spectrum F After obtaining the spectrum F of the electrical quantity, it is preferred to normalize the spectrum F, specifically, calculating the effective value of the spectrum F, and dividing the amplitude of each frequency point in the spectrum F by the effective value to obtain the normalized spectrum Fn .
  • the spectrum F can also be directly used for the next calculation, but normalization is conducive to improving the accuracy of the algorithm.
  • Step b Perform mathematical morphology operation on the spectrum of the electrical quantity to obtain the base value component and/or fluctuation component of the spectrum.
  • Mathematical morphology preferably includes corrosion operation, expansion operation and composite operation of corrosion operation and expansion operation.
  • the spectrum of the arc fault state and the normal working state are processed respectively, and the base value component and fluctuation component of the spectrum of the DC power system in the two states are correspondingly obtained, wherein the spectrum is represented by F, the base value component is represented by Fb , and the fluctuation component is represented by Fm .
  • the corrosion operation and the expansion operation are:
  • the normalized spectrum Fn is subjected to erosion-dilation-dilation-erosion operation, and the base value component Fb1 of the spectrum is obtained as follows:
  • the normalized spectrum Fn is subjected to dilation-erosion-erosion-dilation operation to obtain the base value component Fb2 of the spectrum:
  • the average value of the base value components F b1 and F b2 is calculated to remove the dimension and normalize the base value component of the spectrum F n.
  • the quantity F b is:
  • FIG7 The fluctuation components of normal current and fault current in the whole frequency band obtained in this embodiment are shown in FIG7. It should be noted that FIG7 of this embodiment is illustrated by extracting the energy proportion of the fluctuation component in each characteristic frequency band. As other embodiments, it is also possible to extract the energy proportion of the base value component. Of course, the energy proportion of the base value component and the energy proportion of the fluctuation component can also be used in combination. In this case, a corresponding training data set needs to be established.
  • Step c According to the selected multiple characteristic frequency bands of arc occurrence, respectively calculate the energy proportion of the base value component and/or the fluctuation component of the spectrum in each characteristic frequency band, and then calculate the characteristic value based on the energy proportion of each characteristic frequency band, that is, respectively calculate the energy proportion of the base value component and/or the fluctuation component in the arc fault state and the energy proportion of the base value component and/or the fluctuation component in the normal working state.
  • the energy proportion of the fluctuation component in the arc fault state and the normal working state are calculated respectively, that is, the training data set is composed of the energy proportion of the fluctuation component in the arc fault state and the normal working state.
  • the training data set may also include the working state data of the DC power system in these two states.
  • This embodiment selects 20kHz-28kHz, 36kHz-44kHz and 52kHz-60kHz as the characteristic frequency bands for arc occurrence, but the characteristic frequency bands for arc occurrence are not limited to the above frequency bands. Subsequently, the energy proportion of the fluctuation component Fm in each characteristic frequency band is calculated respectively, and the calculation method is the sum of the squares of the spectrum amplitudes of the characteristic frequency band divided by the sum of the squares of the entire spectrum.
  • the energy proportion of the fluctuation component of each characteristic frequency band is converted into the characteristic value of each characteristic frequency band.
  • the training data set includes the characteristic values converted from the energy proportion of the fluctuation component under the two states of arc fault and normal operation, which is conducive to model calculation and improves accuracy.
  • k is the adjustment factor
  • the training data set may also include working state data, that is, operating data under two states of arc fault or normal operation.
  • working state data that is, operating data under two states of arc fault or normal operation.
  • the energy proportion of each frequency band, the input current I pv of the MPPT unit, the input voltage U pv , the voltage output to the inverter U bus and other converter working information in this embodiment are input into the arc detection mode recognition model to calculate the output of the model.
  • the working state data in this embodiment includes the input current I pv of the MPPT unit, the input voltage U pv , and the voltage output to the inverter U bus .
  • its working state data is different, and the working state data of the same electrical equipment may also be different.
  • the training data set may also include its working state data, and the working state data includes environmental data such as noise generated by the isolating switch during operation. If the isolating switch is not included, the working state data of the DC power system only needs to consider the current and voltage data of the equipment. According to different electrical equipment, its working state data is different, and the working state data of the same electrical equipment may also be different.
  • Step S2 Acquire the electrical quantity of the DC power system operation state, that is, the sampling module or sampling circuit of the controller that executes the arc detection pattern recognition model collects the input current I of the MPPT unit.
  • the training data set also includes working state data, it is also necessary to collect working state data of the DC power system operation state, such as environmental data such as noise.
  • the electrical quantities of the DC power system obtained include DC current, DC voltage, a component of DC current or a component of DC voltage, that is, the signal collected by the sampling module or the sampling circuit can be a DC current or a DC voltage, or a partial component of the DC current or DC voltage, such as a high-frequency component or a component in a specific frequency band.
  • the DC voltage is slightly less effective than the DC current, and the preferred collected signal is the DC current.
  • the normal current waveform diagram is shown in FIG4
  • the arc fault current waveform diagram is shown in FIG5 .
  • the signal lengths of FIGS. 4 and 5 are both 1024 acquisition signals.
  • Step S3 Based on the mathematical transformation-mathematical morphology in step S1, calculate and extract the energy proportion of the fluctuation component of each characteristic frequency band in the electrical quantity spectrum of the operating state.
  • step S3 and step S1 the selected multiple characteristic frequency bands of the arc occurrence are the same, and the mathematical transformation-mathematical morphology calculation method is the same.
  • Step S4 Input the energy proportion or characteristic value of each characteristic frequency band into the arc detection pattern recognition model, and the arc detection pattern recognition model determines whether a series arc fault occurs.
  • the key to the present invention is to extract the energy proportion of each characteristic frequency band from at least the electrical quantity (or the characteristic value adjusted by the energy proportion of each characteristic frequency band).
  • the counter of the arc detection pattern recognition model is set with a threshold value.
  • the counter counts cumulatively, and the counter is increased by 1. Otherwise, the counter is decreased until it reaches 0, that is, the counter is decreased by 1 until it reaches 0. If the count value of the counter is greater than the threshold value, it is determined that a series arc fault has occurred in the DC power system to improve the detection accuracy.
  • the threshold value is selected as 10. Of course, the threshold value may not be set.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Power Engineering (AREA)
  • Testing Relating To Insulation (AREA)
  • Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)

Abstract

一种基于数学形态学和模式识别的直流电弧检测方法，包括选取特征频段，提取电弧故障和正常工作状态下电气量频谱的基值分量和/或波动分量在各特征频段的能量占比，建立并训练电弧检测模式识别模型；获取直流电力系统运行状态的电气量；提取运行状态的电气量频谱中基值分量或波动分量在各特征频段的能量占比；将各特征频段的能量占比以及运行状态的直流电力系统工作状态数据输入至电弧检测模式识别模型判断。本发明通过对电气量进行数学变换-数学形态学计算，从电气量频谱中提取多个电弧发生的特征频段的基值分量或波动分量的能量占比输入到电弧检测模式识别模型，进而判断是否发生串联电弧故障，受信号采样绝对值波动影响较小，精度高。

Description

基于数学形态学和模式识别的直流电弧检测方法 技术领域
本发明涉及直流电力系统保护领域，具体涉及基于数学形态学和模式识别的直流电弧检测方法。
背景技术
在直流电力系统中，由于接插件松脱、线缆材料老化等原因可能会引发电弧故障，而直流电弧容易维持稳定燃烧，其产生的热量经过积累可产生上千度高温，从而引发火灾。
目前，直流电力系统中得到较多应用的是光伏发电系统，其典型结构如图1所示，主要包括光伏电池阵列、光伏并网逆变器和电网。光伏电池阵列通常包含一路或多路光伏电池组串(图1仅示出一路)，光伏并网逆变器通常包括一个或多个MPPT单元(图1仅示出一个)和并网逆变器。光伏发电系统直流侧接线端子很多，因此发生电弧危害的可能性较高，是影响系统安全运行的关键问题。
在光伏发电系统中，按故障电弧发生位置不同，可将故障电弧分为三类，包括串联电弧、并联电弧和对地电弧，在图1中A为组内串联电弧，B为组内并联电弧，C和D为接地电弧，其中并联电弧和对地电弧发生的可能性较小，而且发生时系统电流迅速增大，可以利用电流阈值法检出，然而，串联电弧发生的可能性和频次最高，在出现串联电弧的系统中，直流电压和直流电流幅值不会发生明显变化，这使得串联电弧的检测变得极为困难。
针对不同的物理现象，直流电弧检测和定位方法可以分为基于辐射特性和基于电流特性的方法。在直流电弧发生时，通常伴随发热、发光、噪声和电磁辐射的现象，可以据此进行基于辐射特性的电弧检测，但该类方法受环境因素影响较大，通常适用于密闭空间，而光伏发电系统的范围较大且通常在室外，因此该类方法并不适用。在串联电弧发生后，电弧可等效为一个非线性时变电阻，在电流中产生扰动。基于电流特性的方法会采集一段时间的电流信号，并 对信号进行快速傅里叶变换、小波变换等数学处理手段，从中提取信号的时域特征、频域特征和时频域特征等。常见的时域特征包括电流的方差、峰峰值、变化率、峭度、波形因子、熵等；电弧对频谱的影响通常在几千赫兹至几十千赫兹，常见的频域特征包括特征频段的最大值、能量占比和方差等；常见的时频域特征包括各小波能量占比、奇异值、模极大值和熵等。提取到电流信号的特征值后可以根据提前整定好的阈值，以及特征值随时间的变化情况对电弧故障进行判断。这类方法受光伏逆变器工作状态的影响较大，阈值整定困难，容易造成误报或漏报。
发明内容
本发明的目的在于克服现有技术的缺陷，提供一种基于数学形态学和模式识别的直流电弧检测方法，该方法具备受外界影响小、检测方便且精度较高的优点。
为实现上述目的，本发明采用了如下技术方案：
一种基于数学形态学和模式识别的直流电弧检测方法，包括如下步骤：
步骤S1：基于获取的电弧故障和正常工作两种状态下直流电力系统的电气量和工作状态数据建立训练数据集，并基于模式识别算法和训练数据集建立并训练得到电弧检测模式识别模型；
步骤S2：获取运行状态下直流电力系统的电气量；
步骤S3：基于数学变换-数学形态学计算提取运行状态下电气量频谱中基值分量和/或波动分量在各特征频段的能量占比；其中，所述数学变换-数学形态学计算包括如下步骤：
步骤a：将所获直流电力系统的电气量通过数学变换分别得到电气量的频谱；
步骤b：对电气量的频谱进行数学形态学运算获得频谱的基值分量和/或波动分量；
步骤c：根据所选取的多个特征频段，分别计算频谱的基值分量和/或波动分量在各特征频段的能量占比；
步骤S4：将运行状态下的所述能量占比输入至所述电弧检测模式识别模型，通过所述电弧检测模式识别模型判断是否发生串联电弧故障。
优选的，在步骤S1中，基于支持向量机或人工神经网络算法以及训练数据集建立并训练得到所述电弧检测模式识别模型。
优选的，基于线性核函数的支持向量机建立并训练的电弧检测模式识别模型的计算过程简化为：
ATX+b
其中，X为输入数据向量，A是和X维数相同的系数向量，b是一个常数。
优选的，所述直流电力系统的电气量包括直流电流、直流电压、直流电流的分量或直流电压的分量。
优选的，在步骤a中，在获得电气量的频谱之前先滤除直流分量。
优选的，在步骤a中，首先求取电气量的平均值，通过使每个电气量减去平均值以滤除直流分量，然后再进行数学变换得到电气量的频谱。
优选的，在步骤a中，在获得电气量的频谱后，对频谱进行归一化处理。
优选的，在步骤a中，计算电气量的频谱的有效值，将频点幅值除以有效值得到归一化频谱。
优选的，在步骤a中，通过快速傅里叶变换得到电气量的频谱。
优选的，在步骤b中，利用数学形态学的腐蚀运算、或膨胀运算、或由腐蚀运算和膨胀运算组合的复合运算处理频谱。
优选的，在步骤b中，对频谱或归一化频谱进行腐蚀运算和膨胀运算组合的复合运算，得到频谱或归一化频谱的基值分量和/或波动分量。
优选的，在步骤S3中，选取20kHz-28kHz、36kHz-44kHz以及52kHz-60kHz作为电弧发生的特征频段。
优选的，在步骤S4中，电弧检测模式识别模型的计数器设置有阈值，当计算结果符合电弧故障特征，计数器累加计数，若计数器的计数值大于阈值，则判断直流电力系统中发生了电弧故障。
优选的，步骤S1中的所述电弧检测模式识别模型为预先获得的识别模型；或者，步骤S1包括基于数学变换-数学形态学，建立训练数据集，并基于模式识别算法和训练数据集建立并训练得到电弧检测模式识别模型，其中所述训练数据集包括电弧故障和正常工作两种状态下的能量占比。
优选的，在步骤S1中，基于各特征频段的能量占比转化计算各特征频段的 特征值，训练数据集包括电弧故障和正常工作两种状态下的各特征频段的特征值，
计算公式为：
其中k为调节系数。
基于数学形态学和模式识别建立并训练电弧检测模式识别模型，通过对电气量进行数学变换-数学形态学计算，从电气量频谱中提取多个电弧发生的特征频段的基值分量和/或波动分量的能量占比输入到电弧检测模式识别模型，进而判断是否发生串联电弧故障。由于通过数学形态学去除了频谱的基值成分，受信号采样绝对值波动影响较小，因此相比现有提取电流信号特征值与阈值相比的检测方法，具备受外界影响小、精度高的优点。
此外，各特征频段的能量占比转化为各特征频段的特征值，利于模型计算，利于提高计算精度。
附图说明
图1是背景技术中直流电力系统的示意图；
图2是本发明中直流电力系统的示意图；
图3是本发明的电弧检测算法流程图；
图4是一组将直流电流分量滤除的正常电流波形图；
图5是一组将直流电流分量滤除的电弧故障电流波形图；
图6是正常电流、故障电流经过FFT分析后的频谱；
图7是正常电流、故障电流在全频段的波动分量频谱；
图8是多组正常电流、故障电流在三个频段(20kHz-28kHz、36kHz-44kHz、52kHz-60kHz)的特征值分布图。
具体实施方式
以下结合附图给出的实施例，进一步说明本发明的基于数学形态学和模式识别的直流电弧检测方法的具体实施方式。本发明的基于数学形态学和模式识 别的直流电弧检测方法不限于以下实施例的描述。
一种基于数学形态学和模式识别的直流电弧检测方法，包括如下步骤：
步骤S1：获取电弧故障和正常工作两种状态下直流电力系统的电气量，选取若干电弧发生的特征频段，基于数学变换-数学形态学，计算提取电弧故障和正常工作两种状态下电气量频谱的基值分量和/或波动分量在各特征频段的能量占比，建立训练数据集，训练数据集包括电弧故障和正常工作两种状态下的能量占比，并基于模式识别算法和训练数据集建立并训练得到电弧检测模式识别模型；
所述数学变换-数学形态学计算包括如下步骤：
步骤a：将所获直流电力系统的电气量通过数学变换分别得到电气量的频谱；
步骤b：对电气量的频谱进行数学形态学运算获得频谱的基值分量和/或波动分量；
步骤c：根据所选取的多个特征频段，分别计算频谱的基值分量和/或波动分量在各特征频段的能量占比；
步骤S2：获取运行状态下直流电力系统的电气量；
步骤S3：基于步骤S1的数学变换-数学形态学，计算提取运行状态下电气量频谱中基值分量和/或波动分量在各特征频段的能量占比；
步骤S4：将运行状态下的所述能量占比输入至所述电弧检测模式识别模型，通过所述电弧检测模式识别模型判断是否发生串联电弧故障。本发明基于数学形态学和模式识别建立并训练电弧检测模式识别模型，通过对电气量进行数学变换-数学形态学计算，从电气量频谱中提取多个电弧发生的特征频段的基值分量和/或波动分量的能量占比输入到电弧检测模式识别模型，进而判断是否发生串联电弧故障。由于通过数学形态学去除了频谱的基值成分，受信号采样绝对值波动影响较小，因此相比现有提取电流信号特征值与阈值相比的检测方法，具备受外界影响小、精度高的优点。
结合图2、3提供一个具体实施例，本发明可以应用在光伏发电系统、储能系统、直流微电网等领域，本实施例以光伏发电系统为例。
光伏发电系统主要包括光伏电池阵列、光伏并网逆变器和电网，光伏电池阵列包括一路或多路光伏电池组串，光伏并网逆变器包括一个或多个MPPT单 元和并网逆变器，如图2所示，本实施例的光伏发电系统包括一路光伏电池组串以及与一路光伏电池组串连接的光伏并网逆变器，所述光伏并网逆变器包括开关S、一个MPPT单元、逆变器以及控制器，开关S连接于光伏电池组串与MPPT之间，控制器包括采样模块或采样电路，通过采样模块或采样电路分别采集光伏并网逆变器的电气量数据，当光伏发电系统工作时，闭合光伏并网逆变器的开关S，控制器采集到MPPT单元的输入电流Ipv，输入电压为Upv，输出给逆变器的电压为Ubus。
基于数学形态学和模式识别的直流电弧检测方法包括如下步骤：
步骤S1：首先，采集大量的电弧故障和正常工作两种状态下的直流电力系统的电气量和直流电力系统的工作状态数据，随后，对采集到的电弧故障和正常工作两种状态下的直流电力系统的电气量通过数学变换-数学形态学计算，从电弧故障和正常工作两种状态的电气量频谱中提取基值分量和/或波动分量，选取若干电弧发生的特征频段，计算得到各特征频段的能量占比，
建立训练数据集，训练数据集包括电弧故障和正常工作两种状态下的波动分量能量占比，或训练数据集包括电弧故障和正常工作两种状态下的基值分量能量占比，或，训练数据集包括电弧故障和正常工作两种状态下的波动分量能量占比和基值分量能量占比，另外，训练数据集还可以包括直流电力系统的工作状态数据；
最后，基于模式识别算法以及训练数据集建立并训练电弧检测模式识别模型，模式识别算法可以采用支持向量机和人工神经网络等算法。训练数据集优选通过数据库方式存储，当然也可以通过xml文本或其它方式存储。
在本实施例中，优选基于线性核函数的支持向量机建立并训练的电弧检测模式识别模型以减小计算量，模型的计算过程简化为：
ATX+b
其中，X为输入数据向量，A是和X维数相同的系数向量，b是一个常数。A和b均可以提前通过模型训练得到。具体的，数学变换-数学形态学计算方法包括如下步骤：
步骤a：电弧故障状态的电气量、正常工作状态的电气量分别通过数学变换，也就是电气量的时域与频域的变换，对应地得到电弧故障状态下电气量的频谱、 正常工作状态下电气量的频谱，其中数学变换优选为快速傅里叶变换。在本实施例中，电气量的频谱以F表示。在本实施例中，正常电流、故障电流经过FFT分析后的频谱参见图6。
优选的，在进行快速傅里叶变换之前，首先滤除直流分量，具体过程为，求取电流平均值Iave，每个采集到的电流值减去Iave。当然，也可以省略滤除直流，因为去除直流分量的步骤并不是必须的，但是该步骤利于之后的信号处理过程。
在得到电气量的频谱F后，优选对频谱F进行归一化处理，具体为，计算频谱F的有效值，频谱F中各频点幅值除以有效值得到归一化的频谱Fn。当然，也可以直接利用频谱F进行下一步计算，但经过归一化处理后利于提升算法的准确性。
步骤b：对电气量的频谱进行数学形态学运算获得频谱的基值分量和/或波动分量，数学形态学优选包括腐蚀运算、膨胀运算以及腐蚀运算和膨胀运算组合的复合运算，本实施例中，也就是分别处理电弧故障状态、正常工作状态的频谱，对应的获得直流电力系统在两种状态下的频谱的基值分量和波动分量，其中频谱以F表示，基值分量以Fb表示，波动分量以Fm表示，确定数学形态学运算的结构元素g(m)＝[0，0，0，0，0]T以及偏移量为L，腐蚀运算和膨胀运算为：
腐蚀
膨胀
具体提供一种具体数学形态学运算，本实施例优选扁平状结构元素g(m)＝[0，0，0，0，0]T，偏移量L为2；
对归一化的频谱Fn进行腐蚀-膨胀-膨胀-腐蚀运算，得到频谱的基值分量Fb1为：
对归一化的频谱Fn进行膨胀-腐蚀-腐蚀-膨胀运算，得到频谱的基值分量Fb2为：
求取基值分量Fb1和Fb2的平均值，以去除量纲，归一化的频谱Fn的基值分 量Fb为：
Fb＝0.5·(Fb1+Fb2)；归一化的频谱Fn的波动分量Fm为：
Fm＝Fn-Fb。
需要说明的是，数学形态学运算、组合方式以及采用的结构元素不做具体限制，另外，也可以直接对频谱F进行上述数学形态学运算。本实施例获得的正常电流、故障电流在全频段的波动分量参见图7。需要说明的是，本实施例图7以提取波动分量在各特征频段的能量占比进行说明，作为其它实施例，提取基值分量能量占比也是可以的，当然，也可以同时利用基值分量能量占比与波动分量能量占结合使用，此时需要建立相应的训练数据集。
步骤c：根据所选取的电弧发生的多个特征频段，分别计算频谱的基值分量和/或波动分量在各特征频段的能量占比，再基于各特征频段的能量占比计算特征值，也就是，分别计算电弧故障状态下基值分量和/或波动分量的能量占比、正常工作状态下基值分量和/或波动分量的能量占比，本实施例中，分别计算电弧故障状态以及正常工作状态下的波动分量能量占比，也就是由电弧故障和正常工作两种状态下的波动分量能量占比组建成训练数据集，当然训练数据集也可以包括该两种状态下直流电力系统的工作状态数据。
本实施例选取20kHz-28kHz、36kHz-44kHz以及52kHz-60kHz作为电弧发生的特征频段，但电弧发生的特征频段并不限于上述频段，随后分别计算波动分量Fm在每个特征频段的能量占比，其计算方法为特征频段的频谱幅值的平方和除以整个频谱的平方和。
优选的，将各特征频段的波动分量能量占比转化为各特征频段的特征值，此时训练数据集包括由电弧故障和正常工作两种状态下的波动分量能量占比分别转化的特征值，利于模型计算，提高精确度，当然，直接使用各特征频段的能量占比进行模型训练以及进行判断是否发生串联电弧故障也是可以的。
结合图8提供本实施例在建立模型时多组正常电流、故障电流在三个频段的特征值分布，其中以“○”表示正常电流特征值，以“*”表示故障电流特征值，其中各频段的特征值的计算方法是
其中k为调节系 数，优选k大于0，且小于或等于0.3，本实施例中k＝0.1。
另外，训练数据集还可以包括工作状态数据，也就是在电弧故障或正常工作两种状态下的运行数据，具体在本实施例中，将本实施例中各个频段的能量占比、MPPT单元的输入电流Ipv，输入电压为Upv，输出给逆变器的电压为Ubus等变换器工作信息输入到电弧检测模式识别模型，计算模型的输出，也就是说，本实施例中工作状态数据包括MPPT单元的输入电流Ipv，输入电压为Upv，输出给逆变器的电压为Ubus，根据不同的电器设备，其工作状态数据有所不同，同一电器设备其工作状态数据也可以不同，例如，若直流电力系统包括隔离开关，训练数据集还可以包括其工作状态数据，工作状态数据包括隔离开关在运行过程中产生的噪音等环境数据，若不包括隔离开关，则直流电力系统的工作状态数据只需考虑设备的电流电压等数据。根据不同的电器设备，其工作状态数据有所不同，同一电器设备其工作状态数据也可以不同。
步骤S2：获取直流电力系统运行状态的电气量，也就是由执行电弧检测模式识别模型的控制器的采样模块或采样电路采集MPPT单元的输入电流I。当训练数据集还包括工作状态数据时，还需要采集直流电力系统运行状态的工作状态数据，例如噪音等环境数据。
需要说明的是，在本实施例中，在正常状态、电弧故障状态或运行状态下，所获的直流电力系统的电气量包括直流电流、直流电压、直流电流的分量或直流电压的分量，也就是由采样模块或采样电路采集的信号可以是直流电流或者直流电压，或者采集直流电流或直流电压中的部分分量，例如高频分量、特定频段的分量。在本实施例中，直流电压相比直流电流效果略差，优选采集信号为直流电流。
本实施例中，正常电流波形图参见图4，电弧故障电流波形图参见图5，图4、5的信号长度均为1024个采集信号。
步骤S3：基于步骤S1中的数学变换-数学形态学，计算提取运行状态的电气量频谱中各特征频段的波动分量能量占比，在步骤S3和步骤S1中，所选取的电弧发生的多个特征频段相同，数学变换-数学形态学计算方法相同。
当然，在步骤S1的建模过程中将各特征频段的能量占比转化为各特征频段的特征值时，步骤S3中也需要将运行状态时各特征频段的能量占比转化为各特 征频段的特征值。
步骤S4：将各特征频段的能量占比或特征值输入至电弧检测模式识别模型，由电弧检测模式识别模型判断是否发生串联电弧故障。本发明的关键在于至少需电气量中提取的各特征频段的能量占比(或由各特征频段的能量占比调整的特征值)。
在本实施例中，电弧检测模式识别模型的计数器设置有阈值，当电弧检测模式识别模型判定符合电弧故障特征，计数器累加计数，计数器加1，否则计数器累减直到为0，也就是计数器减1直到为0，若计数器的计数值大于阈值，则判断直流电力系统发生的串联电弧故障，以提高检测准确性，例如本实施例中，阈值选取10。当然，也可以不设置阈值，当电弧检测模式识别模型判定符合电弧故障特征时，直接就认为发生串联电弧故障。
以上内容是结合具体的优选实施方式对本发明所作的进一步详细说明，不能认定本发明的具体实施只局限于这些说明。对于本发明所属技术领域的普通技术人员来说，在不脱离本发明构思的前提下，还可以做出若干简单推演或替换，都应当视为属于本发明的保护范围。

Claims (15)

  1. 一种基于数学形态学和模式识别的直流电弧检测方法，其特征在于：包括如下步骤：
    步骤S1：基于获取的电弧故障和正常工作两种状态下直流电力系统的电气量和工作状态数据建立训练数据集，并基于模式识别算法和训练数据集建立并训练得到电弧检测模式识别模型；
    步骤S2：获取运行状态下直流电力系统的电气量；
    步骤S3：基于数学变换-数学形态学计算提取运行状态下电气量频谱中基值分量和/或波动分量在各特征频段的能量占比；其中，所述数学变换-数学形态学计算包括如下步骤：
    步骤a：将所获直流电力系统的电气量通过数学变换分别得到电气量的频谱；
    步骤b：对电气量的频谱进行数学形态学运算获得频谱的基值分量和/或波动分量；
    步骤c：根据所选取的多个特征频段，分别计算频谱的基值分量和/或波动分量在各特征频段的能量占比；
    步骤S4：将运行状态下的所述能量占比输入至所述电弧检测模式识别模型，通过所述电弧检测模式识别模型判断是否发生串联电弧故障。
  2. 根据权利要求1所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：在步骤S1中，基于支持向量机或人工神经网络算法以及训练数据集建立并训练得到所述电弧检测模式识别模型。
  3. 根据权利要求2所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：基于线性核函数的支持向量机建立并训练的电弧检测模式识别模型的计算过程简化为：
    ATX+b
    其中，X为输入数据向量，A是和X维数相同的系数向量，b是一个常数。
  4. 根据权利要求1所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：所述直流电力系统的电气量包括直流电流、直流电压、直流电流的分量或直流电压的分量。
  5. 根据权利要求1所述的基于数学形态学和模式识别的直流电弧检测方法， 其特征在于：在步骤a中，在获得电气量的频谱之前先滤除直流分量。
  6. 根据权利要求5所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：在步骤a中，首先求取电气量的平均值，通过使每个电气量减去平均值以滤除直流分量，然后再进行数学变换得到电气量的频谱。
  7. 根据权利要求1所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：在步骤a中，在获得电气量的频谱后，对频谱进行归一化处理。
  8. 根据权利要求7所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：在步骤a中，计算电气量的频谱的有效值，将频点幅值除以有效值得到归一化频谱。
  9. 根据权利要求1、5、6、7或8所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：在步骤a中，通过快速傅里叶变换得到电气量的频谱。
  10. 根据权利要求1所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：在步骤b中，利用数学形态学的腐蚀运算、或膨胀运算、或由腐蚀运算和膨胀运算组合的复合运算处理频谱。
  11. 根据权利要求10所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：在步骤b中，对频谱或归一化频谱进行腐蚀运算和膨胀运算组合的复合运算，得到频谱或归一化频谱的基值分量和/或波动分量。
  12. 根据权利要求1所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：在步骤S3中，选取20kHz-28kHz、36kHz-44kHz以及52kHz-60kHz作为电弧发生的特征频段。
  13. 根据权利要求1所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：在步骤S4中，电弧检测模式识别模型的计数器设置有阈值，当计算结果符合电弧故障特征，计数器累加计数，若计数器的计数值大于阈值，则判断直流电力系统中发生了电弧故障。
  14. 根据权利要求1-13任一项所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：
    步骤S1中的所述电弧检测模式识别模型为预先获得的识别模型；或者，
    步骤S1包括基于数学变换-数学形态学，建立训练数据集，并基于模式识别 算法和训练数据集建立并训练得到电弧检测模式识别模型，其中所述训练数据集包括电弧故障和正常工作两种状态下的能量占比。
  15. 根据权利要求1所述的基于数学形态学和模式识别的直流电弧检测方法，其特征在于：
    在步骤S1中，基于各特征频段的能量占比转化计算各特征频段的特征值，训练数据集包括电弧故障和正常工作两种状态下的各特征频段的特征值，
    计算公式为：
    其中k为调节系数。
PCT/CN2023/125029 2022-11-14 2023-10-17 基于数学形态学和模式识别的直流电弧检测方法 WO2024104037A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211421377.5A CN115808600B (zh) 2022-11-14 2022-11-14 基于数学形态学和模式识别的直流电弧检测方法
CN202211421377.5 2022-11-14

Publications (1)

Publication Number Publication Date
WO2024104037A1 true WO2024104037A1 (zh) 2024-05-23

Family

ID=85483054

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/125029 WO2024104037A1 (zh) 2022-11-14 2023-10-17 基于数学形态学和模式识别的直流电弧检测方法

Country Status (2)

Country Link
CN (1) CN115808600B (zh)
WO (1) WO2024104037A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115808600B (zh) * 2022-11-14 2024-04-30 上海正泰电源系统有限公司 基于数学形态学和模式识别的直流电弧检测方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101706527A (zh) * 2009-10-30 2010-05-12 西安交通大学 基于电流高频分量时频特征的电弧故障检测方法
CN102854426A (zh) * 2012-10-09 2013-01-02 邵俊松 基于实时测量多频段频率分量占比判断直流电弧故障的方法
US20180351505A1 (en) * 2017-06-05 2018-12-06 The Texas A&M University System Systems and Methods for Determining Arc Events
CN113962879A (zh) * 2021-09-08 2022-01-21 国网河北省电力有限公司雄安新区供电公司 一种低压直流电弧能量分布可视化研究方法及装置
CN114487689A (zh) * 2021-11-24 2022-05-13 正泰集团研发中心(上海)有限公司 串联电弧故障检测方法、装置、设备和存储介质
CN115808600A (zh) * 2022-11-14 2023-03-17 正泰集团研发中心(上海)有限公司 基于数学形态学和模式识别的直流电弧检测方法

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9431816B2 (en) * 2012-09-28 2016-08-30 Eaton Corporation Direct current arc fault detector and circuit interrupter, and method of detecting an arc in a direct current power circuit
CN107870286A (zh) * 2017-11-06 2018-04-03 福州大学 一种直流故障电弧检测方法及装置
CN110542835B (zh) * 2018-05-29 2021-11-02 上海海拉电子有限公司 一种用于车辆电弧故障的检测方法、检测系统及试验系统
US11656263B2 (en) * 2019-06-11 2023-05-23 Arizona Board Of Regents On Behalf Of Arizona State University Effective feature set-based high impedance fault detection
CN115267450A (zh) * 2022-07-21 2022-11-01 上海电机学院 一种直流电弧故障检测方法及装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101706527A (zh) * 2009-10-30 2010-05-12 西安交通大学 基于电流高频分量时频特征的电弧故障检测方法
CN102854426A (zh) * 2012-10-09 2013-01-02 邵俊松 基于实时测量多频段频率分量占比判断直流电弧故障的方法
US20180351505A1 (en) * 2017-06-05 2018-12-06 The Texas A&M University System Systems and Methods for Determining Arc Events
CN113962879A (zh) * 2021-09-08 2022-01-21 国网河北省电力有限公司雄安新区供电公司 一种低压直流电弧能量分布可视化研究方法及装置
CN114487689A (zh) * 2021-11-24 2022-05-13 正泰集团研发中心(上海)有限公司 串联电弧故障检测方法、装置、设备和存储介质
CN115808600A (zh) * 2022-11-14 2023-03-17 正泰集团研发中心(上海)有限公司 基于数学形态学和模式识别的直流电弧检测方法

Also Published As

Publication number Publication date
CN115808600A (zh) 2023-03-17
CN115808600B (zh) 2024-04-30

Similar Documents

Publication Publication Date Title
WO2024104037A1 (zh) 基于数学形态学和模式识别的直流电弧检测方法
Yılmaz et al. A real-time UWT-based intelligent fault detection method for PV-based microgrids
Allan et al. A new passive islanding detection approach using wavelets and deep learning for grid-connected photovoltaic systems
CN103605016B (zh) 一种电能质量数据处理方法和装置
Chen et al. Intelligent identification of voltage variation events based on IEEE Std 1159-2009 for SCADA of distributed energy system
CN109617096B (zh) 一种基于遍历阻抗的区域电网宽频带扰动稳定分析方法
CN111007364A (zh) 一种电缆早期自恢复故障的识别方法
CN107505519B (zh) 一种分布式电源接入电网电能质量分析方法及装置
CN109038574A (zh) 一种光伏电站的组串并联失配损耗的计算方法
CN105550450B (zh) 一种电能质量干扰源特征谐波建模方法
Damala et al. A simple decision tree-based disturbance monitoring system for VSC-based HVDC transmission link integrating a DFIG wind farm
CN102478601A (zh) 一种用于电能质量中64个采样点的谐波计算方法
Liu et al. Identification of major power quality disturbance sources in regional grid based on monitoring data correlation analysis
CN113128153A (zh) 一种光伏电站中复合阈值触发的主动变频故障录波方法
CN116298509A (zh) 一种电力系统谐波谐振的在线识别方法
CN107271916B (zh) 一种电池板组串健康状态检测方法
Kumar et al. Voltage and current actuated hybrid protection scheme for utility grid with high penetration levels of renewable energy
CN102467614A (zh) 一种用于电能质量的数据采集与计算方法
CN110221168B (zh) 一种主导谐波源定位和谐波污染传播路径追踪方法
Behzadi et al. Identification of combined power quality disturbances in the presence of distributed generations using variational mode decomposition and K-nearest neighbors classifier
Patil et al. Improved fault detection and location scheme for photovoltaic system
Patel et al. Wavelet and Machine learning based approach for Fault classification in AC Micro-grid system
Xi et al. Research on islanding detection of solar distributed generation based on best wavelet packet and neural network
Nayak et al. Change detection filter technique of HVDC transmission link fed by a wind farm
CN114047372B (zh) 一种基于电压特征的台区拓扑辨识系统