Ring direct current microgrid line fault diagnosis method based on fuzzy granulation interval

Abstract

The application provides a ring direct-current micro-grid line fault diagnosis method based on a fuzzy granulation interval, a pole-to-pole / positive pole grounding fault classification strategy based on upper and lower bounds of the granulation interval and output voltage estimation, and realization of pole-to-pole / positive pole grounding fault classification by using characteristics of the granulation interval itself and characteristics of a capacitor discharge process, so that the pole-to-pole / positive pole grounding fault classification is not affected by voltage / current similarity caused by unknown fault resistance, and the pole-to-pole / positive pole grounding fault can be accurately distinguished in less than 0.5 ms without setting a threshold value. A new negative pole grounding fault detection index interval range is provided, abnormal fluctuations of voltage after fault are amplified, and the negative pole grounding fault can be detected in 1.5 ms. After obtaining diagnosis results at the ends of various converters, a neighbor cooperation mechanism can be used to identify a fault line. The diagnosis method has good performance when the fault position and fault resistance change, can accurately diagnose different types of line faults, and has adaptability to different network topologies.

Description

Technical Field

[0001] This invention belongs to the field of DC microgrid fault diagnosis technology, specifically a method for diagnosing line faults in a ring DC microgrid based on fuzzy granular intervals. Background Technology

[0002] With the rapid development of DC / DC conversion technology and the increasing proportion of DC power load, DC microgrids have become a promising option for renewable energy consumption. DC microgrids offer advantages such as strong controllability, ease of deployment, high flexibility, high reliability, and high transmission efficiency, providing customers with high-quality power. However, failures are difficult to avoid due to various factors such as equipment lifespan, weather conditions, and natural disasters. The lack of relevant protection schemes and standards poses a significant challenge to the promotion and application of DC microgrid technology.

[0003] DC microgrids can have radial, ring, and mesh topologies. Ring topologies can be simplified to radial topologies or extended to mesh topologies. Power sources in a ring microgrid can supply power in any direction to ensure stable power supply during fault conditions. These characteristics make ring microgrid applications attractive. However, in a ring structure, there are always two paths between a faulty node and any other node, which increases the difficulty of fault analysis and diagnosis.

[0004] Existing fault diagnosis methods for DC microgrids can be mainly divided into methods based on single-end line information and methods based on double-end line information. Fault diagnosis methods based on single-end line information identify faults by detecting the electrical state at one end of the line, such as detecting single-end current, current change rate, power, or current-limiting reactor voltage. These methods rely solely on local information for fault detection, requiring no communication and offering high reliability. However, they are susceptible to short-circuit impedance and struggle to detect high-resistance faults. Fault diagnosis methods based on double-end line information detect faults by comparing differences in electrical quantities such as current and power at both ends of the line. They typically exhibit high sensitivity and selectivity; however, they rely on real-time communication between the two ends of the line and are significantly affected by communication delays. Furthermore, for line fault detection in bipolar DC microgrids, existing methods require the simultaneous installation of corresponding measuring devices on both the positive and negative lines. As the scale of the microgrid increases, this leads to protection cost issues. Additionally, most current methods based on basic local measurement information have their own characteristics and applicable scenarios; the selection of fault features and the setting of diagnostic thresholds require careful consideration.

[0005] Compared with the prior art, this application has the following advantages;

[0006] Comparison with patent CN1163383804 "A fault diagnosis method for DC microgrids based on minimal sensors";

[0007] Patent CN103178538A uses the output voltage and current of a local converter to construct characteristic quantities, and employs a voltage-current differential sliding window summation to amplify voltage and current changes after a fault, thereby achieving the detection of inter-electrode and ground faults. This method requires setting thresholds ε for both the current characteristic quantity SI and the voltage characteristic quantity SV. SI and ε SV Threshold selection needs to be discussed separately. Furthermore, the window length of the sliding window cumulative sum and the interval of multi-step differencing affect the detection time and algorithm performance, and also need to be selected in advance. The algorithm has many parameters, requiring extensive prior knowledge analysis.

[0008] The proposed "Fault Diagnosis Method for Ring DC Microgrids Based on Fuzzy Granular Intervals" utilizes fuzzy granular interval theory. Firstly, for inter-pole and positive-pole grounding faults with significant fault current changes, detection is achieved by setting ε1 using current differentiation. Then, classification is realized through the relationship between interval boundaries and the estimated voltage of the converter output. For negative-pole grounding faults with insignificant fault currents, a characteristic interval range IR is constructed based on interval features to amplify the voltage change after the fault, reducing the difficulty of selecting the negative-pole grounding detection threshold ε2. The threshold ε1 used in this invention is only for preliminary differentiation, while the selection of threshold ε2 only needs to consider the case of high-resistance faults at the end of the line. The threshold selection difficulty is low, and the algorithm is not affected by parameters such as differential intervals.

[0009] In patent CN103178538A, converter output voltage estimation is achieved based on fault dynamic analysis of the converter output-side capacitor. The fault classification threshold ε is set according to the different errors between the output voltage and the measured voltage under different fault conditions. v This system classifies faults into inter-pole, positive-pole grounding, and negative-pole grounding types. The selection of classification thresholds requires separate discussion.

[0010] The fault classification method proposed in this invention fully utilizes the characteristics of fuzzy granular intervals. For inter-electrode faults, the error between the estimated and actual output voltage values ​​is very small, and the estimated value lies within the interval. For positive-to-ground faults, the estimated voltage value will be less than the actual value, and the estimated value will be less than the lower bound of the interval. This method does not require setting a classification threshold when classifying inter-electrode / positive-to-ground faults, and it can also solve the problem that inter-electrode and positive-to-ground faults have similar characteristics when the fault resistance is twice that of the actual fault.

[0011] Comparison with patent CN 113702767 B "Fault Diagnosis Method for Islanded DC Microgrids Based on Wavelet Sliding Window Energy";

[0012] Patent CN 113702767 B addresses a parallel-type DC microgrid, where the power supply provides power to the load through a tree-like structure. There is no coupling between the lines, making the analysis relatively simple. However, the applicability of this invention to other topologies has not been verified.

[0013] The diagnostic method proposed in this invention is applicable to ring topologies as well as parallel and mesh topologies, as verified in Example 3. In a ring topology, the power supply can be directed in any direction to ensure stable power supply under fault conditions. A ring topology can be simplified to a radial topology or extended to a mesh topology. However, a ring topology contains two paths between the fault and any node, making fault analysis and diagnosis more challenging compared to a parallel topology.

[0014] Patent CN 113702767 B describes a data-driven diagnostic method that collects the output current from the converter network side for fault diagnosis. This method decomposes the raw current using a wavelet sliding window energy method, extracts fault features, trains a support vector machine model, and then uses it for fault diagnosis. This method requires pre-constructing a training dataset through simulation; the quality of the training data directly affects the algorithm's diagnostic performance. However, in practical engineering, it is often difficult to obtain a sufficient training dataset.

[0015] The diagnostic method proposed in this invention belongs to the mechanism analysis diagnostic method. By conducting prior analysis on different fault characteristics, it constructs feature quantities based on the original voltage and current sampling signals and uses them for diagnosis. It does not require model training or a large amount of data as support, has a wider range of applications, and lower implementation costs.

[0016] Comparison with patent CN111722054A "A method for analyzing single-pole grounding faults in MMCHVDC transmission lines based on capacitive fuzzy recognition" and patent CN111722055A "A method for identifying single-pole grounding faults in MMC DC transmission lines based on inductive fuzzy recognition".

[0017] Patents CN111722054A and CN111722055A target DC transmission lines, which differ significantly from the DC microgrids to which this invention applies. Compared to DC transmission lines, DC microgrids contain numerous source-load-storage units, resulting in strong system random uncertainty. The low inertia of DC microgrids and bidirectional fault currents lead to weak overcurrent capacity and high requirements for rapid protection. Furthermore, DC microgrid lines are shorter, resulting in limited sampling information, typically only localized data. In addition, protection and grounding standards related to DC microgrids are immature, leaving many protection issues unresolved. These characteristics make the diagnostic method proposed in this paper significant.

[0018] Although patents CN111722054A and CN111722055A contain the word "fuzzy" in their titles, they are essentially based on the ratio of the correlation coefficients of the voltage, current and voltage change rate and current change rate of the sampled positive and negative poles of the line to diagnose faults. The fuzzy coefficient in the above patents is essentially a ratio of correlation, which is calculated based on the Pearson correlation coefficient and has no connection with fuzzy theory or fuzzy sets.

[0019] The "Fault Diagnosis Method for Ring DC Microgrids Based on Fuzzy Granulated Intervals" proposed in this invention is based on fuzzy set theory and fuzzy membership functions. By constructing granular intervals for voltage and current sampling signals, it amplifies the dynamics after a fault and designs diagnostic feature quantities accordingly. This method is significantly different from the aforementioned patents.

[0020] Patents CN111722054A and CN111722055A require additional sensors to be installed on the line to collect the positive and negative voltages and currents at both ends of the line for diagnosis, which increases investment costs to some extent.

[0021] The method proposed in this paper only requires obtaining the grid-side voltage and positive current of the converter to complete the diagnosis. The signal used is also used for system control and shares the input with the system controller. There is no need to install additional sensors on the line, which has significant advantages.

[0022] Comparison with patent CN 114325213 A "A fault detection method for DC transmission system".

[0023] Patent CN 114325213 A addresses DC transmission systems, which differs significantly from the DC microgrids to which this invention applies. Compared to DC transmission systems, DC microgrids contain numerous source-load-storage units, resulting in strong system random uncertainty. The low inertia of DC microgrids and bidirectional fault currents lead to weak overcurrent capacity and high requirements for rapid protection. Furthermore, the shorter line lengths of DC microgrids limit sampling information, typically providing only localized data. In addition, relevant protection and grounding standards for DC microgrids are immature, leaving many protection issues unresolved. These characteristics make the diagnostic method proposed in this paper significant.

[0024] Although patent CN 114325213 A contains the word "fuzzy" in its title, it is essentially based on the correlation between the negative current and voltage collected from the rectifier and inverter sides of a DC transmission system to achieve diagnosis. The fuzziness coefficient in this patent is essentially a ratio of correlation, calculated based on the Pearson correlation coefficient, and has no connection with fuzzy theory or fuzzy sets.

[0025] The "Fault Diagnosis Method for Ring DC Microgrids Based on Fuzzy Granulated Intervals" proposed in this invention is based on fuzzy set theory and fuzzy membership functions. By constructing granular intervals for voltage and current sampling signals, it amplifies the dynamics after a fault and designs diagnostic feature quantities accordingly. This method is significantly different from the aforementioned patents.

[0026] Patent CN 114325213 A requires the installation of additional sensors on the line to collect the positive and negative voltages and currents at both ends of the line for diagnosis, which increases the investment cost to some extent.

[0027] The method proposed in this paper only requires obtaining the grid-side voltage and positive current of the converter to complete the diagnosis. The signal used is also used for system control and shares the input with the system controller. There is no need to install additional sensors on the line, which has significant advantages.

[0028] Comparison with patent CN116068333A "Fault Selection Device and Method Based on Fuzzy Theory and Multi-Criterion Fusion" and patent CN116068333B "Fault Selection Device and Method Based on Fuzzy Theory and Multi-Criterion Fusion".

[0029] Patents CN111722054A and CN111722055A target distribution networks, which differ significantly from the DC microgrids to which this invention applies. Compared to traditional AC distribution networks, DC microgrids lack natural zero-crossing points, making it impossible to utilize existing AC protection technologies. Furthermore, DC microgrids lack frequency and phase information, resulting in fewer available fault characteristic quantities. In addition, fault currents change rapidly and have short durations, making diagnosis more difficult. These characteristics highlight the significance of the diagnostic method proposed in this paper.

[0030] The methods proposed in patents CN111722054A and CN111722055A apply fuzzy theory to the selection of overvoltage fault lines in distribution networks. After diagnosing a fault based on overvoltage, they utilize multiple criteria based on fuzzy theory to fuse the results of three basic fault selection methods: one based on high-frequency mode energy, one based on the fifth harmonic component, and one based on attenuated DC components. This process selects the faulty line and represents a follow-up step after fault detection. Furthermore, this method only considers single-phase grounding faults.

[0031] The "Fault Diagnosis Method for Ring DC Microgrids Based on Fuzzy Granulated Intervals" proposed in this invention directly applies fuzzy set theory and membership functions to fault diagnosis. By constructing granular intervals, it can accurately diagnose both inter-pole faults and grounding faults, demonstrating higher reliability compared to the overvoltage threshold method of the aforementioned patent. The fault line determination method in this paper employs a collaborative approach between adjacent diagnostic units, requiring only distributed non-real-time communication, thus achieving even higher reliability.

[0032] The methods proposed in patents CN111722054A and CN111722055A require additional sensors to be installed on the line to collect the zero-sequence voltage and zero-sequence current of the bus at both ends of the line for diagnosis, which increases the investment cost to some extent.

[0033] The method proposed in this paper only requires obtaining the grid-side voltage and positive current of the converter to complete the diagnosis. The signal used is also used for system control and shares the input with the system controller. There is no need to install additional sensors on the line, which has significant advantages.

[0034] Comparison with patent CN105911414A "A multi-criteria fusion fault selection method for distribution networks based on fuzzy theory".

[0035] Patent CN105911414A addresses distribution networks, which differs significantly from the DC microgrids to which this invention applies. Compared to traditional AC distribution networks, DC microgrids lack natural zero-crossing points, making it impossible to utilize existing AC protection technologies. Furthermore, DC microgrids lack frequency and phase information, resulting in fewer available fault characteristic quantities. In addition, fault currents change rapidly and have short durations, making diagnosis more difficult. These characteristics highlight the significance of the diagnostic method proposed in this paper.

[0036] The method proposed in patent CN105911414A applies fuzzy theory to the selection of overvoltage fault lines in distribution networks. After diagnosing the fault based on overvoltage, it intelligently integrates complementary fault selection methods based on high-frequency mode energy, the fifth harmonic component, and the attenuated DC component using multiple criteria based on fuzzy theory, thereby ultimately selecting the faulty line. This is a follow-up process after fault detection. Furthermore, this method only considers single-phase grounding faults.

[0037] The "Fault Diagnosis Method for Ring DC Microgrids Based on Fuzzy Granulated Intervals" proposed in this invention directly applies fuzzy set theory and membership functions to fault diagnosis. By constructing granular intervals, it can accurately diagnose both inter-pole faults and grounding faults, demonstrating higher reliability compared to the overvoltage threshold method of the aforementioned patent. The fault line determination method in this paper employs a collaborative approach between adjacent diagnostic units, requiring only distributed non-real-time communication, thus achieving even higher reliability.

[0038] The method proposed in patent CN105911414A requires the installation of additional sensors on the line to collect the zero-sequence voltage and zero-sequence current of the bus at both ends of the line for diagnosis, which increases the investment cost to some extent.

[0039] The method proposed in this paper only requires obtaining the grid-side voltage and positive current of the converter to complete the diagnosis. The signal used is also used for system control and shares the input with the system controller. There is no need to install additional sensors on the line, which has significant advantages.

[0040] Comparison with patent CN102305910A "A method for diagnosing large-scale DC analog circuits based on fuzzy neural networks".

[0041] Patent CN102305910A targets large-scale DC analog circuits, belonging to the field of integrated circuits. It is mainly used to simulate and test the operation of power systems. It is large in scale and uses power system simulation software for calculation and simulation. Its main faults include soft faults and hard faults in the circuit.

[0042] The DC microgrid targeted in this invention is a small-scale, independent power system in actual operation, belonging to the field of power systems. It consists of renewable energy generation facilities and energy storage systems, used for power supply to specific regions or buildings. Its main faults include inter-pole faults and grounding faults.

[0043] The two differ significantly in purpose, scale, operation mode, connection method and application field.

[0044] Although the term "interval" also appears in patent CN102305910A, it divides the voltage interval [0,+∞] into multiple sub-intervals based on the tolerance range of the component, which is used to describe the fault state of soft faults and as input to a fuzzy neural network.

[0045] The interval of this invention is based on the fuzzy membership function to construct the upper and lower bounds of the voltage fluctuation range, and the interval is used for the design of fault classification and detection feature range range. The voltage interval is always around the working voltage of 400V. It is a fluctuation interval constructed based on data, which has stronger real-time performance and accuracy.

[0046] Although the term "fuzzy" also appears in patent CN102305910A, it uses a fuzzy neural network to take the membership degree corresponding to the measured voltage range obtained above as input, and the 0 / 1 binary output is used to indicate the fault condition, which is a data-driven method.

[0047] The "Fault Diagnosis Method for Ring DC Microgrid Based on Fuzzy Granulated Intervals" proposed in this invention directly applies fuzzy set theory and membership functions to fault diagnosis. By constructing granulated intervals, it is a diagnostic method based on mechanism analysis. Through prior analysis of different fault characteristics, it constructs feature quantities based on the original voltage and current sampling signals and uses them for diagnosis. It does not require model training or a large amount of data as support, has a wider range of applications, and lower implementation costs. Summary of the Invention

[0048] To address the aforementioned technical problems, this invention proposes a fault diagnosis method for ring DC microgrid lines based on fuzzy granular intervals. This method aims to perform diagnosis using the output voltage and positive current sampling information of the DC-DC converter, even with limited sensors. It primarily solves two major challenges in sensor-limited diagnosis: classifying inter-electrode / positive grounding faults under unknown fault resistance and detecting negative grounding faults with indistinct positive current characteristics.

[0049] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0050] The fault diagnosis method for ring DC microgrid lines based on fuzzy granular intervals includes the following steps:

[0051] Step 1) Determine the fuzzy granulation interval and the interval range;

[0052] The purpose of fuzzy granulation is to extract abnormal fluctuations in data and quantify them in the form of intervals. By granulating the model input through fuzzy set theory, an input-output mapping based on granulation is established, which effectively captures voltage fluctuations after a fault and reflects them in interval quality indicators such as width and boundary.

[0053] Step 2) Fault diagnosis based on interval characteristics;

[0054] By utilizing the inherent characteristics of the granulation interval and the features of the capacitor discharge process, inter-electrode / positive-electrode grounding fault classification is achieved, making it unaffected by the voltage / current similarity caused by unknown fault resistance; a fault feature width range is designed based on the interval width to amplify the voltage change after the fault and detect negative-electrode grounding faults with indistinct current characteristics.

[0055] Step 3) Fault line identification;

[0056] Using the above strategy, the fault types occurring in the two lines connected to each protection unit are diagnosed, and a neighboring source cooperation method is proposed to further identify faulty lines based on the diagnostic results of the local protection unit and its neighboring protection units.

[0057] 2. The method for fault diagnosis of ring DC microgrid lines based on fuzzy granular intervals according to claim 1, characterized in that: step 1) specifically involves the following steps;

[0058] The fuzzy granulation process consists of three steps:

[0059] First, the sampling sequence is divided into multiple subsequences, called the operation window. The operation window n contains the sampled voltage U. n Based on the previous w-1 data, in order to capture voltage fluctuations, each operating window is sorted and further divided into two subsequences, as follows:

[0060]

[0061] Where b = 1 when w is even and b = 2 when w is odd, and [w / 2] represents the largest integer not greater than w / 2;

[0062] Then, for each operation window, granular intervals are constructed using fuzzy membership functions. Choosing an appropriate membership function can significantly improve the quality of the constructed granular intervals. Triangular membership functions are used, and the upper and lower bounds of the intervals are defined as follows:

[0063]

[0064] Finally, repeat step 2) until all operation windows have completed the granulation operation;

[0065] After granulation, the original voltage signal is converted into various granular intervals containing important features for fault diagnosis. The granular interval of a certain granulation window is represented as follows:

[0066]

[0067] Among them G upn G lown It constructs the lower and upper bounds of the interval, and the interval width measures the range of data fluctuation. IW n Defined as:

[0068]

[0069] A wider range reflects a faster drop in the output voltage of the converter near the faulty line. Based on this characteristic, a new fault diagnosis index, the range difference IR, is proposed. n Defined as:

[0070] IR n =maxG upn -minG lown (5)

[0071] The range represents the difference between the upper limit of the maximum range and the lower limit of the minimum range. By utilizing the characteristics of fuzzy granular ranges, abnormal voltage fluctuations caused by inter-pole / positive grounding and negative grounding faults can be accurately detected and quantified.

[0072] As a further improvement to the present invention, step 2) is specifically as follows;

[0073] Considering that the sampled current is discrete, backward differential is used to approximate the first derivative of the current and is applied to detect inter-electrode / positive-electrode grounding faults with significant changes in fault current. Δt is the sampling step size, and Δi is the sampling step size. n It is the difference between the nth and (n-1)th current samples;

[0074]

[0075] Once the sampling frequency is determined, Δt becomes a constant, in which case Δi is used. n To estimate the first derivative of the current, the detection strategy for inter-electrode and positive-electrode grounding faults is defined as follows:

[0076]

[0077] Where t is the time point and ε1 is the detection threshold;

[0078] To avoid misclassification of faults, a fault classification method based on interval features is proposed. Considering the capacitor discharge process in the circuit, the following relationship is derived:

[0079] i in (t)≈i in (t 0- )≈i o (t 0- (8)

[0080] Where i in (t) and i in (t 0_ ) represents the input-side feed current before and after the fault, i o (t 0_ The current is the output-side feed current before the fault. According to Kirchhoff's current law, the capacitor discharge current is expressed as follows:

[0081] i C (t)=i o (t)-i in (t)≈i o (t)-i o (t 0- (9)

[0082] Considering the voltage drop caused by capacitor discharge, via i o (t) and initial output voltage U o (t0) is used to calculate the estimated output voltage:

[0083]

[0084] U o (t0) is the initial voltage across the capacitor. Considering discrete sampling, equation (6) can be rewritten as:

[0085]

[0086] Where N is the number of samples after the fault, and T s It is the sampling period, which is the estimated voltage when an inter-electrode fault occurs. Approximate measurement voltage U o And it will fall within the constructed granular range; however, when a positive ground fault occurs, the size of the discharge capacitor will double. The descent speed will be faster than U o Therefore, the estimated voltage will be outside the granular range and below the lower limit. Based on the above analysis, the classification strategy for inter-electrode and positive-electrode grounding faults is designed as follows:

[0087]

[0088] When an inter-pole & positive-pole ground fault occurs, only the current in the two units adjacent to the faulty line will rise rapidly and exceed ε1. Voltage estimation is performed on these two units. In addition, since the calculation of the upper and lower limits is in real time, this classification method does not require setting thresholds.

[0089] If in window W n If equation (7) is not satisfied, it may correspond to a negative grounding or non-fault condition. Since the current hardly changes after a negative grounding fault occurs, it is difficult to accurately distinguish between negative grounding fault and non-fault conditions using only current samples. For a negative grounding fault, the voltage of the unit adjacent to the fault line drops faster. Therefore, the negative grounding fault detection method uses the abnormal fluctuation of the voltage after the fault to perform fuzzy granular interval detection.

[0090] For each voltage sampling data window W n Calculate the interval range IR n It is then compared with a predefined threshold ε2 to detect negative grounding faults:

[0091]

[0092] As a further improvement to the present invention, step 3) is as follows;

[0093] For determining the faulty line Fi, the principle of the proposed method is shown in equation (15), and:

[0094]

[0095] Where i,j=1,2…6,|ij|=1.D={between poles, positive pole grounded, negative pole grounded},T i and T jIt is a trip signal. When the protection unit at one end of the line detects a fault, it sends the corresponding diagnostic results to the adjacent unit. If the same diagnostic results are received from the adjacent unit, it indicates that the line connected to the two converters has a fault. The relevant solid-state circuit breaker is used to identify and isolate the faulty line. Since only the fault type rather than a specific value is sent to the adjacent unit for judgment, only distributed low-bandwidth communication is needed in this process.

[0096] Therefore, load switching and photovoltaic intermittency can also cause non-faulty transient changes. When the load fluctuates significantly, only the protection unit adjacent to the load will recognize this change. Since adjacent source units receive different diagnostic results, they will not issue a trip command.

[0097] Total diagnosis time T d The calculation is as follows:

[0098] T d =T op +T cd +T SSCB (15)

[0099] Where T op It is the execution time of the diagnostic algorithm, T cd This is the communication delay time. Considering the change in fault distance, the protection unit closer to the fault point will trigger the detection first, therefore T op Take the larger value between the detection times of the two protection units.

[0100] As a further improvement of the present invention, the communication delay T cd According to the IEC 61850 standard, T cd It is determined by four factors:

[0101] Processing delay, propagation delay, queuing delay, and transmission delay.

[0102] As a further improvement to the present invention, the detection threshold ε1 is specifically selected as follows:

[0103] In the proposed method, threshold ε1 is used to detect inter-electrode / positive ground faults based on current characteristics, while ε2 is used to detect negative ground faults based on voltage characteristics, by changing the fault location p and the fault resistance R. f To simulate inter-pole, positive-pole grounding, and negative-pole grounding faults, the following characteristic values ​​were calculated using the first derivative of the corresponding current and the peak IR within 2ms after the fault to determine the threshold.

[0104]

[0105] For all scenarios, analyze the feature values ​​under extreme conditions to set diagnostic thresholds.

[0106] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0107] This invention proposes a fault diagnosis method for ring DC microgrid lines based on fuzzy granular intervals. It employs a pole-to-pole / positive-to-ground fault classification strategy based on the upper and lower bounds of the granular intervals and output voltage estimation. Utilizing the inherent characteristics of the granular intervals and the features of the capacitor discharge process, it achieves pole-to-pole / positive-to-ground fault classification, making it unaffected by voltage / current similarity caused by unknown fault resistance. It can accurately distinguish pole-to-pole / positive-to-ground faults in less than 0.5 ms without setting a threshold. A novel negative-to-ground fault detection index, the interval range, is proposed, amplifying abnormal voltage fluctuations after a fault and detecting negative-to-ground faults within 1.5 ms. After obtaining the diagnostic results at each converter end, a neighbor cooperation mechanism can be used to identify the faulty line. The invented diagnostic method exhibits good performance under varying fault location and fault resistance conditions, accurately diagnoses different types of line faults, and is adaptable to different network topologies. Attached Figure Description

[0108] Figure 1 This is a flowchart of the present invention;

[0109] Figure 2 This is a schematic diagram of the DC microgrid structure of the present invention;

[0110] Figure 3 This is a simplified model of the impact between adjacent lines under fault conditions according to the present invention;

[0111] Figure 4 This is a schematic diagram of the inter-electrode fault / positive electrode grounding fault current of the present invention;

[0112] Figure 5 This is a schematic diagram of the inter-electrode fault / positive grounding fault detection results of the present invention;

[0113] Figure 6 This is a schematic diagram of the negative electrode grounding fault detection results of the present invention;

[0114] Figure 7 This is a schematic diagram of the radial DC microgrid and the grid-line DC microgrid of the present invention;

[0115] Figure 8 This is a schematic diagram showing the positive / negative grounding fault detection and diagnosis results under different topologies of the present invention. Detailed Implementation

[0116] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0117] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0118] A fault diagnosis method for a ring DC microgrid in a fuzzy granular region includes the following steps, as shown in the appendix. Figure 1 As shown:

[0119] Step 1) Determine the fuzzy granulation interval and the interval range

[0120] The purpose of fuzzy granulation is to extract abnormal fluctuations in data and quantify them in the form of intervals. By granulating the model input using fuzzy set theory, a granular input-output mapping can be established, effectively capturing post-fault voltage fluctuations and reflecting them in interval quality indicators such as width and boundaries. The fuzzy granulation process can be divided into three main steps:

[0121] First, the sampling sequence is divided into multiple subsequences, called operating windows. The operating window n contains the sampled voltage U. n And its previous w-1 data. To capture voltage fluctuations, each operating window is sorted and further divided into two subsequences, which can be represented as:

[0122]

[0123] Where b = 1 when w is even and b = 2 when w is odd, and [w / 2] represents the largest integer not greater than w / 2. In this invention, the value of w is chosen to be 5.

[0124] Then, for each operation window, granular intervals are constructed using fuzzy membership functions. Choosing an appropriate membership function significantly improves the quality of the constructed granular intervals. In this paper, the triangular membership function is used because it has been shown to better capture fluctuating data trends. The upper and lower bounds of the intervals are defined as follows:

[0125]

[0126] Finally, repeat step 2) until all operation windows have completed the granulation operation.

[0127] After granulation, the original voltage signal is converted into various granular intervals containing important features for fault diagnosis. The granular interval of a certain granulation window can be represented as:

[0128]

[0129] Among them G upn G lown These are the lower and upper bounds of the constructed interval. The interval width measures the range of data fluctuation, IW. n Defined as:

[0130]

[0131] A wider range can reflect a faster drop in the output voltage of the converter near the faulty line. Based on this characteristic, a new fault diagnosis index, the range difference IR, is developed. n It can be defined as:

[0132] IR n =maxG upn -minG lown (5)

[0133] The range represents the difference between the upper bound of the maximum range and the lower bound of the minimum range. By utilizing the characteristics of fuzzy granular ranges, abnormal voltage fluctuations caused by inter-pole / positive grounding and negative grounding faults can be accurately detected and quantified.

[0134] Step 2). Fault diagnosis based on interval characteristics

[0135] Considering that the sampled current is discrete, backward differential is used to approximate the first derivative of the current, and this is applied to detect inter-electrode / positive-electrode grounding faults with significant changes in fault current. Δt is the sampling step size, Δi... n It is the difference between the nth and (n-1)th current samples.

[0136]

[0137] Once the sampling frequency is determined, Δt becomes a constant. In this case, Δi can be used. n To estimate the first derivative of the current. The detection strategy for inter-electrode and positive-electrode grounding faults can be defined as follows:

[0138]

[0139] Where t is the time point and ε1 is the detection threshold.

[0140] To avoid misclassification of faults, a fault classification method based on interval features is proposed. Consider the capacitor discharge process in a circuit. The following relationship can be derived:

[0141] i in (t)≈i in (t 0- )≈i o (t 0- (8) where i in (t) and i in (t 0_ ) represents the input-side feed current before and after the fault, i o (t 0_ This is the output-side feed current before the fault. According to Kirchhoff's Current Law, the capacitor discharge current can be expressed as follows:

[0142] i C(t)=i o (t)-i in (t)≈i o (t)-i o (t 0- (9)

[0143] Considering the voltage drop caused by capacitor discharge, it can be achieved through i o (t) and initial output voltage U o (t0) is used to calculate the estimated output voltage:

[0144]

[0145] U o (t0) is the initial voltage across the capacitor. Considering discrete sampling, (6) can be written as:

[0146]

[0147] Where N is the number of samples after the fault, and T s This is the sampling period. When an inter-electrode fault occurs, the estimated voltage is... Approximate measurement voltage U o And it will fall within the constructed granular range. However, when a positive ground fault occurs, the size of the discharge capacitor will double. The descent speed will be faster than U o Therefore, the estimated voltage will be outside the granular range and below the lower limit. Based on the above analysis, the classification strategy for inter-electrode and positive-to-ground faults can be designed as follows:

[0148]

[0149] When an inter-pole & positive-pole ground fault occurs, only the current in the two units adjacent to the faulty line will rise rapidly and exceed ε1, and voltage estimation will be performed on these two units. Furthermore, since the upper and lower limits are calculated in real time, this classification method does not require setting thresholds.

[0150] If in window W n If equation (7) is not satisfied, it may correspond to a negative ground fault or a non-fault condition. Since the current hardly changes after a negative ground fault occurs, it is difficult to accurately distinguish between negative ground faults and non-fault conditions using only current samples. For negative ground faults, the voltage of units adjacent to the faulty line drops faster. Therefore, the negative ground fault detection method uses abnormal fluctuations in the voltage after the fault to perform fuzzy granular interval detection.

[0151] For each voltage sampling data window W n Calculate the interval range IR nIt is then compared with a predefined threshold ε2 to detect negative grounding faults:

[0152]

[0153] Step 3). Fault line identification

[0154] Using the above strategy, the type of fault occurring in the two lines connected to each unit can be diagnosed. The invention further proposes a neighboring source cooperation method to further identify faulty lines. For determining the faulty line Fi, the principle of the proposed method is shown in equation (15), and:

[0155]

[0156] Where i,j=1,2…6,|ij|=1.D={between poles, positive pole grounded, negative pole grounded},T i and T j This is a trip signal. When a protection unit at one end of the line detects a fault, it sends the corresponding diagnostic results to the adjacent unit. If the same diagnostic results are received from the adjacent unit, it indicates that a fault has occurred on the line connected to these two converters, and the relevant solid-state circuit breaker can be used to identify and isolate the faulty line. Since only the fault type, rather than a specific value, is sent to the adjacent unit for judgment, only distributed low-bandwidth communication is needed in this process.

[0157] The proposed method also demonstrates high sensitivity to faults. The DC microgrid studied in this invention includes photovoltaics, batteries, and loads. Therefore, load switching and photovoltaic intermittency can also cause non-faulty transient changes. When the load fluctuates significantly, only the protection unit adjacent to the load will recognize this change. Since adjacent source units receive different diagnostic results, no tripping command will be issued.

[0158] Total diagnosis time T d It can be calculated as follows:

[0159] T d =T op +T cd +T SSCB (15)

[0160] Where T op This refers to the diagnostic algorithm execution time. Considering the variation in fault distance, protection units closer to the fault point will trigger detection first, therefore T... op Take the larger of the detection times of the two protection units. Communication delay T cd It can be determined according to the IEC 61850 standard. cdThe delay is primarily determined by four factors: processing delay, propagation delay, queuing delay, and transmission delay. The processing delay in this paper is small because it is affected by the size of the transmitted data and the bandwidth of the communication channel. In DC microgrids, the distance between the two terminals is small, resulting in a small propagation delay. For advanced communication systems, transmission and queuing delays are negligible. Communication delay is typically in the range of 400 to 600 μs. Therefore, the communication delay can be considered to be 450 μs. The circuit breaker operating time T... SSCB It is typically considered to be 50 μs. Based on the above analysis, the total communication delay T cd It is considered to be 500 μs.

[0161] Further threshold selection was performed, specifically as follows:

[0162] In the proposed method, threshold ε1 is used to detect inter-electrode / positive ground faults based on current characteristics, while ε2 is used to detect negative ground faults based on voltage characteristics. Due to the limited availability of local measurement units, it is difficult to accurately calculate the thresholds using analytical methods. In practice, the threshold values ​​are determined through repeated testing under different fault scenarios. This is achieved by varying the fault location p and the fault resistance R. f ., in the appendix Figure 2 At point F5, inter-electrode, positive-to-ground, and negative-to-ground faults were simulated respectively. To determine the threshold, the following eigenvalues ​​were calculated using the corresponding first derivative of the current and the peak IR within 2 ms after the fault.

[0163]

[0164] For all scenarios, analyze characteristic values ​​under extreme conditions to set diagnostic thresholds. When a high-resistance ground fault (R...) occurs... f When a fault occurs at the far end of the line (p = 0.95 pu) with a resistance of 10Ω, the minimum FV1 and FV2 for inter-pole and positive-pole ground faults are 9.0408 and 4.6791, respectively. When R... f When the resistance is 0.01Ω, the maximum FV3 for a negative ground fault is 2.1094 at the near end (p = 0.05pu). Even with a low fault resistance, the first derivative of the current for a negative ground fault is much smaller than that for inter-electrode and positive ground faults under high-resistance ground faults. To detect inter-electrode / positive ground faults and distinguish them from negative ground faults, ε1 should be in the range of [2.1094, 4.6791]. On the other hand, the post-fault voltage drop and range will decrease with increasing fault resistance and distance, especially when p = 0.95pu and R... f When the resistance is 10Ω, the minimum FV4 is 1.2187. Therefore, in order to detect a negative ground fault, ε2 should be less than 1.2187.

[0165] In industrial applications, factors such as measurement error and noise should also be considered when selecting the threshold. According to the IEC 61869-3 standard, the measurement error for voltage and current should be less than 3%. Furthermore, measuring equipment is typically designed with a signal-to-noise ratio (SNR) of approximately 80 dB. Therefore, measurement error and noise in the sampled signal are also considered when determining the threshold. Taking these factors into account, the values ​​of ε1 and ε2 are chosen to be 4 and 1, respectively.

[0166] Figure 2 The illustrated DC microgrid is a ring-type bipolar DC microgrid with a voltage level of 400V (±200V). It includes one photovoltaic power generation unit, three energy storage units, and two DC load units. Each distributed power source and load is connected to the bus via a DC / DC converter. The performance of the proposed fault diagnosis method in terms of sensitivity, selectivity, speed, and reliability is evaluated. Sensors are installed on the output side of each converter, with a sampling frequency f. s =20kHz.

[0167] Example 1:

[0168] This embodiment illustrates the detection and classification of inter-electrode / positive ground faults. When an inter-electrode or positive ground fault occurs, it should first be distinguished from a negative ground fault and a non-fault transient based on a threshold ε1. To verify the classification capability of the proposed method, two typical fault resistors were simulated in the middle of line 61 (F6) at t = 0.5s. and The measured current and corresponding first derivative of different units are as follows: Figure 4 As shown. The signals of two adjacent units of the fault line are marked with dashed lines. Although the fault lines are different, the fault current characteristics are consistent with the analysis in Section 2. The first-stage derivative of the current for inter-pole and positive-to-ground faults will exceed ε1 within 0.5001s.

[0169] At the same time, the similarity between the two types of faults can also be seen in Figure 4 It was observed that it is difficult to distinguish between these two types of faults simply by setting a threshold. To address this issue, the classification strategy proposed in Equation (12) is adopted. The classification algorithm is initiated after an inter-electrode / positive ground fault is detected. It is only necessary to estimate the voltage of two adjacent units, since only their current rises rapidly and exceeds ε1. The estimated voltage, measured voltage, and voltage range of unit 6 (load 2) and unit 1 (photovoltaic) are as follows. Figure 5 As shown.

[0170] For positive-to-ground faults, the second voltage estimate will be outside the granulation interval, and fault classification will be completed within 0.50015s. Furthermore, since the voltage estimate for inter-electrode faults is always within the granulation interval before the capacitor discharge ends, to ensure algorithm speed, the calculation results within the granulation window (five samples) will be used as the basis for classification. Considering the total time delay T... td =500μs, the diagnosis time for inter-electrode faults and positive-electrode grounding faults is approximately 0.80ms and 0.65ms, respectively. Inter-electrode faults can generally be classified within 0.50030s.

[0171] Example 2:

[0172] This embodiment illustrates the detection and classification of negative grounding faults. At t = 0.5s, a negative grounding fault is simulated in the middle of line 12 (F1). R f =10Ω. The measured current and corresponding first derivative of different units are as follows: Figure 6 As shown in (a)-(b), the first derivative of the current in a negative ground fault is much smaller compared to inter-pole and positive ground faults. Even with a low fault resistance, it will not exceed ε1, thus preventing false triggering by the classification algorithm. The fault current does not exhibit such a significant characteristic when a negative ground fault occurs, making detection difficult. However, the proposed fuzzy granularization method can amplify the post-fault voltage dip by continuously calculating the range of all cells, thereby improving the accuracy of negative ground fault detection.

[0173] To further verify the performance of the proposed method under different line fault conditions, simulations were also performed on F6 and R. f A negative ground fault with a voltage rating of 10Ω. When a negative ground fault occurs, the voltage range characteristic of different converters is as follows: Figure 6 As shown in (c)-(d), despite the different fault lines, the range between adjacent units increases faster than in other converters. For faults F1 and F6, the maximum time for the range between adjacent units to exceed the threshold ε2 is approximately 0.50070 and 0.50065 s, respectively. Considering a total time delay of 500 μs, the diagnosis time for the negative ground fault is approximately 1.20 and 1.15 ms.

[0174] Example 3:

[0175] To demonstrate the feasibility and adaptability of the proposed method to different network topologies, the following considerations were made. Figure 7 The radial network and mesh network are shown. Figure 7 In (a), line 23 is removed from the system due to maintenance or shutdown. At t = 0.5s, the intermediate simulated R of line 23 (F2) is... f A negative ground fault with a resistance of 10Ω was identified. The diagnostic results are as follows: Figure 8As shown in (a)-(b), both converters adjacent to the faulty line can classify a positive-to-ground fault in 0.50015s. Considering T... td =500μs, diagnosis time is 0.65ms.

[0176] In addition, two new lines, Line 13 (1.5 km) and Line 26 (1.2 km), have been added to form... Figure 7 (b) shows the mesh network. At t = 0.5s, R is simulated between line 13 (F7) and line 26 (F8). f =10Ω negative ground fault. As shown in 8(c)-(d). Despite the different gate structures, the range difference between adjacent cells rises faster than in other converters. For faults F7 and F8 The longest detection times for negative grounding faults are approximately 0.50070 s and 0.50065 s, respectively. When T... td When the time is 500 μs, the diagnosis time for negative ground faults is approximately 1.20 ms and 1.15 ms. These results demonstrate that this method can be applied to DC microgrids with various network topologies.

[0177] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A fault diagnosis method for ring DC microgrid lines based on fuzzy granular intervals, characterized in that: Includes the following steps: Step 1) Determine the fuzzy granulation interval and the interval range; The purpose of fuzzy granulation is to extract abnormal fluctuations in data and quantify them in the form of intervals. By granulating the model input through fuzzy set theory, an input-output mapping based on granulation is established, which effectively captures voltage fluctuations after a fault and reflects them in interval quality indicators such as width and boundary. The fuzzy granulation process consists of three steps: Step 1.1 divides the sampling sequence into multiple sub-sequences, called the operation window. The operation window n contains the sampled voltage U. n and its predecessors To capture voltage fluctuations, the data is sorted for each operating window and further divided into two subsequences, as follows: （1） Among them when When b is even, b = 1; when b is even, b = 1. When b is odd, b = 2. Indicates not greater than The largest integer that is 2 / 2; Step 1.2 For each operation window, granular intervals are constructed using fuzzy membership functions. Selecting appropriate membership functions significantly improves the quality of the constructed granular intervals. Triangular membership functions are used, and the upper and lower bounds of the intervals are defined as follows: （2） Step 1.3 Repeat step 1.2 until all operation windows are granular; After granulation, the original voltage signal is converted into various granular intervals containing important features for fault diagnosis. The granular interval of a certain granulation window is represented as follows: （3） in It is the lower and upper bounds of the constructed interval, and the interval width. Defined as: （4） A wider range reflects a faster drop in the output voltage of the converter near the faulty line. Based on this characteristic, a new fault diagnosis index, the range difference IR, is proposed. n Defined as: （5） The range represents the difference between the upper limit of the maximum range and the lower limit of the minimum range. By utilizing the characteristics of fuzzy granular ranges, abnormal voltage fluctuations caused by inter-pole / positive grounding and negative grounding faults can be accurately detected and quantified. Step 2) Fault diagnosis based on interval characteristics; By utilizing the inherent characteristics of the granulation range and the features of the capacitor discharge process, inter-electrode / positive-electrode grounding fault classification can be achieved, making it unaffected by the voltage / current similarity caused by unknown fault resistance; Considering that the sampled current is discrete, backward differential is used to approximate the first derivative of the current, and this method is applied to detect inter-electrode / positive-electrode grounding faults with significant changes in fault current. It is the sampling step size. It is the difference between the nth and (n-1)th current samples; （6） Once the sampling frequency is determined, It then becomes a constant, in which case, using To estimate the first derivative of the current, the detection strategy for inter-electrode and positive-electrode grounding faults is defined as follows: （7） Where t is a time point. It is the detection threshold; To avoid misclassification of faults, a fault classification method based on interval features is proposed. Considering the capacitor discharge process in the circuit, the following relationship is derived: （8） Where i in (t) and i in (t 0_ ) represents the input-side feed current before and after the fault, i o (t 0_ The current is the output-side feed current before the fault. According to Kirchhoff's current law, the capacitor discharge current is expressed as follows: （9） Considering the voltage drop caused by capacitor discharge, via i o (t) and initial output voltage U o (t0) is used to calculate the estimated output voltage: （10） U o (t0) is the initial voltage across the capacitor. Considering discrete sampling, (6) can be written as: （11） Where N is the number of samples after the fault, and T s It is the sampling period, which is the estimated voltage when an inter-electrode fault occurs. Approximate measurement voltage U o And it will fall within the constructed granular range; however, when a positive ground fault occurs, the size of the discharge capacitor will double. The descent speed will be faster than U o Therefore, the estimated voltage will be outside the granular range and below the lower limit. Based on the above analysis, the classification strategy for inter-electrode and positive-electrode grounding faults is designed as follows: （12） When an inter-pole & positive-pole ground fault occurs, only the current in the two units adjacent to the faulty line will rise rapidly and exceed [a certain value]. Voltage estimation is performed on these two units. Furthermore, since the calculation of the upper and lower limits is real-time, this classification method does not require setting thresholds. If in window W n If equation (7) is not satisfied, it corresponds to the negative grounding or non-fault situation. Since the current hardly changes after the negative grounding fault occurs, it is difficult to accurately distinguish the negative grounding fault and non-fault state using only current samples. For the negative grounding fault, the voltage of the unit adjacent to the fault line drops faster. Therefore, the negative grounding fault detection method uses the abnormal fluctuation of the voltage after the fault to perform fuzzy granular interval detection. For each voltage sampling data window W n Calculate the interval range IR n and compare it with a predefined threshold. Comparison is used to detect negative grounding faults: （13） Step 3) Fault line identification; Using the above strategy, the fault types occurring in the two lines connected to each protection unit are diagnosed, and a neighboring source cooperation method is proposed to further identify faulty lines based on the diagnostic results of the local protection unit and its neighboring protection units.

2. The method for fault diagnosis of ring DC microgrid lines based on fuzzy granular intervals according to claim 1, characterized in that: Step 3) The specific steps are as follows; For determining the faulty line Fi, the principle of the proposed method is shown in equation (14), and: （14） Where i,j=1,2…6, |ij|=1. D= {between poles, positive pole grounded, negative pole grounded}, T i and T j It is a trip signal. When the protection unit at one end of the line detects a fault, it sends the corresponding diagnostic results to the adjacent unit. If the same diagnostic results are received from the adjacent unit, it indicates that the line connected to the two converters has a fault. The relevant solid-state circuit breaker is used to identify and isolate the faulty line. Since only the fault type rather than a specific value is sent to the adjacent unit for judgment, only distributed low-bandwidth communication is needed in this process. Therefore, load switching and photovoltaic intermittency can also cause non-faulty transient changes. When the load fluctuates significantly, only the protection unit adjacent to the load will recognize this change. Since adjacent source units receive different diagnostic results, they will not issue a trip command. Total diagnosis time T d The calculation is as follows: （15） Where T op It is the execution time of the diagnostic algorithm, T cd This is the communication delay time. Considering the change in fault distance, the protection unit closer to the fault point will trigger the detection first, therefore T op Take the larger value between the detection times of the two protection units.

3. The method for fault diagnosis of ring DC microgrid lines based on fuzzy granular intervals according to claim 2, characterized in that: The communication delay T cd According to the IEC 61850 standard, T cd It is determined by four factors: Processing delay, propagation delay, queuing delay, and transmission delay.

4. The method for fault diagnosis of ring DC microgrid lines based on fuzzy granular intervals according to claim 3, characterized in that: Detection threshold The specific choices are as follows: In the proposed method, the threshold Used for detecting inter-electrode / positive grounding faults based on current characteristics, Used for detecting negative grounding faults based on voltage characteristics, by changing the fault location p and the fault resistance R. f To simulate inter-pole, positive-pole grounding, and negative-pole grounding faults, the following characteristic values ​​were calculated using the first derivative of the corresponding current and the peak IR within 2ms after the fault to determine the threshold. （16） For all scenarios, analyze the feature values ​​under extreme conditions to set diagnostic thresholds.

Images