Unified threshold fault detection method for DC microgrid based on multiple mean square error features

Abstract

The present invention is a DC microgrid unified threshold fault detection method based on multiple mean square error features. Under this method, the characteristics of the fault voltage signal are amplified and highlighted by MMSD feature value extraction, and the probability distribution changes of the fault voltage data on the converter bus side at different stages (stable, fault moment, and after the fault) are captured. Based on these ubiquitous probability distribution changes of the fault voltage data, a unified threshold is proposed by combining data statistics methods and standardization methods. The converter fault detection under this method is based on the feature value extraction and threshold comparison of voltage data, without the need for additional sensors, and the fault converter positioning is achieved through distributed communication, which has low communication bandwidth requirements and does not require real-time. Compared with traditional detection methods, the present invention has higher rapidity, reliability and cost-effectiveness.

Description

Technical Field

[0001] The present invention belongs to the technical field of DC microgrid fault detection, and specifically provides a DC microgrid unified threshold fault detection method based on multiple mean square error features. Background Art

[0002] As traditional fossil fuels are gradually depleted and their prices continue to rise, people are stepping up efforts to increase the penetration rate of renewable energy to meet the growing demand for electricity. As an economical and efficient means of integrating distributed DC power sources and loads, DC microgrids have the advantages of fewer energy conversion stages, less feeder energy loss, and large transmission capacity, and have become an important part of the new power system. In DC microgrids, existing fault detection research mainly focuses on line faults, and the detection methods for converter faults are relatively limited. However, all DC power sources and loads are connected to the DC bus through power electronic converters. Converter failures can bring serious risks, affect system stability, reduce energy conversion efficiency, and cause potential equipment damage. The most common fault types in converters are IGBT switch short-circuit faults and open-circuit faults. For short-circuit faults, fast fuses can be implanted in the converter switches to convert them into open-circuit faults for detection.

[0003] Existing methods have proposed a variety of solutions for converter fault detection, including: an overcurrent protection scheme based on converter output current signal detection, which collects current and detects overcurrent signals through relays. When the input current signal exceeds the threshold and lasts for more than the set time, the system assumes that a converter fault has occurred. This method has fast detection speed and simple operation, but its detection threshold depends on empirical settings and has limited applicability in different application scenarios. Another existing method combines overcurrent and current directional / differential comparison protection schemes, sets different thresholds for different fault scenarios and types, has high detection accuracy and sensitivity, but has high requirements for parameter settings and data acquisition accuracy. In order to overcome the challenge of threshold selection, another existing method detects faults by comparing the similarity of the current change rate at both ends of the line, avoiding the use of fixed thresholds while being able to make protection decisions quickly without delay. However, this method still has high requirements on the reliability of the communication network, and requires data exchange and decision derivation between multiple nodes, which increases the cost of fault detection.

[0004] The patents that have been applied for are as follows:

[0005] Patent number: CN202110783338.9, patent name A fault component identification method suitable for a fault-clearing DC distribution network. A fault component identification method is proposed for a DC distribution mode of a fault-clearing converter combined with a fast disconnector. According to the fault characteristics of each bus at different fault locations, a bus virtual voltage characteristic quantity is constructed to identify the fault line; according to the switching process of some sub-modules of the hybrid converter, an auxiliary identification criterion is constructed based on the capacitor charge transfer amount to complete the identification of the fault bus. The inventive method can realize the accurate identification of fault components and accurate fault ranging, has a strong ability to withstand transition resistance, and reduces the investment in line sensors.

[0006] It constructs a bus virtual voltage characteristic based on the fault characteristics of each bus at different fault locations in the DC distribution network to identify the fault line. The calculation of this characteristic requires obtaining the voltage information of multiple buses and the output current information of multiple converters through real-time communication.

[0007] The MMSD fault characteristic value constructed by the present invention can be calculated by only acquiring the output voltage of the local converter, and the fault detection does not require communication.

[0008] Patent number: CN 201811362494.2, patent name: Ring DC distribution network fault detection method based on voltage prediction, belongs to the field of safe operation of DC distribution network, and protection devices are installed at both ends of the DC distribution network, and the protection devices include circuit breakers, sensors, relays, and reactors; it is characterized in that: the predicted voltage value of the line side at the next moment is predicted according to the sampled voltage values ​​at any three sampling time points on the line side, and the difference between the predicted voltage value and the measured voltage value on the line side at the moment of the fault is used as the action basis of the protection. The present invention solves the problems of low speed of traditional DC distribution network protection, excessive dependence on communication and the damping state of the circuit. The present invention only relies on the voltage value on the line side, and effectively avoids the influence of excessive dependence on the damping state of the circuit at the moment of the fault when setting protection using elements such as current, first-order derivative and second-order derivative of current, and first-order derivative of voltage.

[0009] It identifies faults by the difference between the predicted voltage and the measured voltage value on the line side at the moment of fault. Its voltage prediction method is designed based on the second-order derivative of voltage. The premise that this fault detection method can quickly and accurately identify faults is that there is a very large voltage drop at the moment of fault. When the voltage drops slowly, it is difficult to ensure the speed and accuracy of fault detection using this method.

[0010] The fault detection method proposed in the present invention is achieved by extracting the output voltage on the converter bus side and calculating the MMSD characteristic value, which involves window sliding and data power exponential amplification, has a significant amplification effect on the fault characteristics, and performs detection based on the threshold, with good rapidity. Summary of the invention

[0011] In order to solve the above problems, the present invention proposes a DC microgrid unified threshold fault detection method based on multiple mean square error features, with a unified threshold set to 7, which overcomes the challenge of threshold setting and simplifies the detection process of the DC microgrid converter. By standardizing the fault data set and exploring its correlation with the MMSD eigenvalue, the efficiency and speed of feature extraction are improved, and fault detection can be completed within 2ms.

[0012] To achieve the above object, the technical solution adopted by the present invention is:

[0013] A DC microgrid unified threshold fault detection method based on multiple mean square error features is characterized by comprising the following steps:

[0014] Step 1) The output voltage of the converter bus side of each power supply end is sampled by a voltage sensor at a sampling frequency of 20kHz;

[0015] Step 2) Based on the data windowing operation, the MMSD eigenvalue of the bus side voltage data is calculated. The calculation steps are as follows:

[0016] First, calculate the mean μ(n) and variance σ(n) of the voltage data:

[0017]

[0018]

[0019] Where U(x) is the output voltage sampling value of the converter bus side at the power supply end at time x, n is the length of a single data window, and N is the total number of sampling points. It satisfies the following relationship with the total sampling time T and the sampling frequency f:

[0020] Nf=T (3)

[0021] Then, substitute the above values ​​into the MMSD feature extraction formula to obtain the eigenvalue S M (n):

[0022]

[0023] Step 3) The fault characteristic value S M (n) Compare with the unified threshold to detect whether a converter open circuit fault occurs.

[0024] Step 4) Under distributed communication, each unit sends fault detection results to each other, compares and determines the location of the faulty converter.

[0025] Step 5) Send a signal to the circuit breaker to isolate the faulty converter.

[0026] As a further improvement of the present invention, the specific steps of step 3) are as follows:

[0027] The determination of the unified threshold value relies on the analysis of the statistical characteristics of the bus-side voltage, with the probability distribution of the reference data set satisfying the standard normal distribution as the basic requirement.

[0028] First, taking the standard normal distribution data set U(x) as a reference, and knowing that its mean μ(n′) is 0 and variance σ(n′) is 1, formula (4) is simplified as follows:

[0029]

[0030] Then, in the context of data statistical characteristics analysis, U(x) is replaced by the standard normal distribution probability density function f(x), which satisfies:

[0031]

[0032] Then, the MMSD eigenvalue corresponding to the standard normal distribution of the data set is obtained and approximated by the integral equation:

[0033]

[0034] Therefore, 6.383 is the standard value. In order to ensure the reliability of the fault detection method and retain about 10% detection margin, the unified threshold value for the output voltage on the converter bus side under this method is 7;

[0035] Finally, based on the threshold method, according to the converter fault voltage characteristic value S M (n) Perform fault detection, if S M If (n) is greater than the threshold 7, it is determined that a converter failure has occurred.

[0036] As a further improvement of the present invention, the specific steps of step 4) are as follows:

[0037] Distributed communication is performed between adjacent distributed generation units, and fault detection results are sent to each other. Because the open-circuit fault of the converter will propagate through the bus and affect the voltage of other source and load nodes, but because the propagation and influence process takes time, the voltage change direction caused by the Boost converter fault and the Buck converter fault is also different. Therefore, by observing the time sequence and value change direction of each node MMSD characteristic value change exceeding the threshold, the location of the converter fault can be located.

[0038] Beneficial effects:

[0039] The present invention proposes a DC microgrid unified threshold fault detection method based on multiple mean square error features. This method is based on the probability distribution analysis of the output voltage on the converter bus side under fault conditions under data statistics, and can extract and quantify the corresponding local statistical features of the data. By comparing with the unified threshold, it can achieve rapid and accurate detection of various types of converter open circuit faults. In addition, this method does not require complex sensor configuration and real-time communication, has high reliability and economy, and can sensitively respond to dynamic changes under different fault scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of the present invention;

[0041] Figure 2 It is a schematic diagram of the DC microgrid structure of the present invention;

[0042] Figure 3 This is a schematic diagram of the open circuit fault of the Boost converter;

[0043] Figure 4 It is a schematic diagram of the output voltage change on the bus side of each power supply end when the Boost converter has an open circuit fault;

[0044] Figure 5 It is a schematic diagram of the open circuit fault detection result of the Boost converter of the present invention;

[0045] Figure 6 This is a schematic diagram of the Buck converter open circuit fault;

[0046] Figure 7 This is a schematic diagram of the output voltage change on the bus side of each power supply end when the Buck converter has an open circuit fault;

[0047] Figure 8 This is a schematic diagram of the open circuit fault detection result of the Buck converter of the present invention;

[0048] Fig. 9 This is a schematic diagram of the output voltage change on the bus side of each power supply end under normal line load switching;

[0049] Fig.10 It is a schematic diagram of the detection result under normal switching of the line load of the present invention. DETAILED DESCRIPTION

[0050] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The flow chart of the present invention is as follows: Figure 1 As shown:

[0051] Embodiment 1:

[0052] This embodiment and the following embodiments are all based on Figure 2 The DC microgrid shown is used as an object for explanation. Figure 2The DC microgrid shown is a ring bipolar DC microgrid with a voltage level of 400V (±200V), including 1 photovoltaic power generation unit, 2 energy storage units and 1 DC load unit. Each distributed power source and load is connected to the bus through a DC / DC converter. Among them, the photovoltaic power generation unit adopts a maximum power tracking control strategy to make full use of renewable energy, and the energy storage unit adopts a droop control strategy to adjust the system voltage. The switching frequency of the converter in the system is set to 20kHz, and the voltage and current sampling frequency is set to 20kHz.

[0053] when Figure 2 When an open circuit fault occurs in the IGBT switch tube of the Boost converter of the DC microgrid photovoltaic array shown in FIG. Figure 3 As shown in the figure, because the Boost converter relies on the switching action of the IGBT switch tube to achieve voltage boost, the corresponding boost basic principle will not work properly under fault conditions, resulting in a significant drop in the output voltage, which may be close to the input voltage level. The inductor is continuously in a discharge state, causing the inductor current to continue to decrease.

[0054] This embodiment describes a detection method for an open circuit fault of an IGBT switch tube of a Boost converter. Figure 4 As shown, the output voltage on the bus side of the Boost converter drops rapidly in the transient state (within a few milliseconds). Although the bus side voltages of other nodes will also be affected to a certain extent, compared with the faulty nodes, their response time is longer and the voltage change rate is smaller in the transient state.

[0055] The schematic diagram of the Boost converter open circuit fault detection of the present invention is as follows Figure 5 As shown in the figure, the MMSD eigenvalues ​​are extracted based on the sliding time window strategy. Before the fault occurs, the MMSD eigenvalues ​​of all nodes fluctuate around a fixed value and do not exceed the threshold, which indicates that the system is operating normally; when a fault occurs, the MMSD eigenvalues ​​of the photovoltaic nodes corresponding to the Boost converter first change significantly and exceed the threshold. This is because this method is highly sensitive to the change in fault voltage and can quickly detect anomalies; after the fault occurs, the eigenvalues ​​of other nodes also gradually change and exceed the threshold, because the fault will propagate through the bus and gradually affect the output voltage values ​​on the bus side of other power supply ends. However, since the propagation and influence process takes time, there will be a time difference in the change of the eigenvalues ​​of different nodes, which can determine the location of the faulty converter.

[0056] Embodiment 2:

[0057] when Figure 2 When an open-circuit fault occurs in the IGBT switch tube of the Buck converter of the DC load of the DC microgrid shown in the figure, the output voltage of the bus side of each power supply end under the fault is as follows: Figure 6As shown in the figure. Under the corresponding fault, the power demand on the DC load side will decrease, causing the voltage across the load to gradually drop to zero. The inductor will continue to discharge, and the inductor current will gradually decrease to zero. Therefore, the current on the bus side of the DC load converter will decrease, while the voltage will increase.

[0058] This embodiment describes a detection method for an open circuit fault of an IGBT switch tube of a Buck converter. Figure 7 As shown, when a fault occurs, the output voltage on the bus side of the converters at each power supply end will not be greatly affected in the transient state (within a few milliseconds) because their response requires a certain amount of time; after the fault occurs, the output voltage value on the node bus side gradually increases.

[0059] The schematic diagram of the Buck converter open circuit fault detection of the present invention is as follows Figure 8 As shown in the figure, the MMSD eigenvalue is also extracted based on the sliding time window strategy to perform fault detection. When a fault occurs, the voltage MMSD eigenvalue on the bus side of other power sources does not change temporarily; after the fault occurs, the voltage MMSD eigenvalue on the bus side of other power sources will gradually change and exceed the threshold. The direction of change of the eigenvalue is consistent with the direction of voltage change, showing an increasing trend. Therefore, it can be judged that a converter fault has occurred, and it is a Buck converter open circuit fault.

[0060] Embodiment 3:

[0061] This embodiment illustrates the results of the detection method under normal load switching. The line load switching under normal conditions is simulated, wherein the load power is increased by 10%, to verify that the method will not give false alarms under normal system operation.

[0062] The bus voltage at each power supply end during line load switching is as follows: Fig. 9 When line load switching occurs, the output voltage on the bus side of the power supply will not respond immediately in a transient state. This is because the overall impact of load switching on the system is small, and voltage changes also require a certain propagation time.

[0063] The schematic diagram of the normal switching detection of the line load of the present invention is as follows Fig.10 As shown, in the case of line load switching, the MMSD feature value extracted based on the sliding time window strategy will not react to it, so that no misjudgment will occur.

[0064] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A DC microgrid unified threshold fault detection method based on multiple mean square error features, characterized in that: The following steps are involved: Step 1) The output voltage of the converter bus side of each power supply end is sampled by a voltage sensor at a sampling frequency of 20kHz; Step 2) Based on the data windowing operation, the MMSD eigenvalue of the bus side voltage data is calculated. The calculation steps are as follows: First, calculate the mean μ(n) and variance σ(n) of the voltage data: Where U(x) is the output voltage sampling value of the converter bus side at the power supply end at time x, n is the length of a single data window, and N is the total number of sampling points, which satisfies the following relationship with the total sampling time T and the sampling frequency f: Nf=T (3) Then, substitute the above values ​​into the MMSD feature extraction formula to obtain the eigenvalue S M (n): Step 3) The fault characteristic value S M (n) comparing with a unified threshold to detect whether a converter open circuit fault occurs; Step 4) Under distributed communication, each unit sends fault detection results to each other, compares and determines the location of the faulty converter; Step 5) Send a signal to the circuit breaker to isolate the faulty converter.

2. The DC microgrid unified threshold fault detection method based on multiple mean square error features according to claim 1 is characterized in that: Step 3) The specific steps are as follows: The determination of the unified threshold depends on the analysis of the statistical characteristics of the bus-side voltage, with the probability distribution of the reference data set meeting the standard normal distribution as the basic requirement; First, taking the standard normal distribution data set U(x) as a reference, and knowing that its mean μ(n′) is 0 and variance σ(n′) is 1, formula (4) is simplified as follows: Then, in the context of data statistical characteristics analysis, U(x) is replaced by the standard normal distribution probability density function f(x), which satisfies: Then, the MMSD eigenvalue corresponding to the standard normal distribution of the data set is obtained and approximated by the integral equation: In order to ensure the reliability of the fault detection method and retain about 10% of the detection margin, the unified threshold for the output voltage on the converter bus side under this method is 7; Finally, based on the threshold method, according to the converter fault voltage characteristic value S M (n) Perform fault detection, if S M (n) is greater than the threshold value 7, it is determined that a converter failure has occurred; Step 4) The specific steps are as follows: Distributed communication is performed between adjacent distributed generation units, and fault detection results are sent to each other. Because the open-circuit fault of the converter will propagate through the bus and affect the voltage of other source and load nodes, but since the propagation and influence process takes time, the voltage change direction caused by the Boost converter fault and the Buck converter fault is also different. Therefore, by observing the time sequence and value change direction of each node MMSD eigenvalue change exceeding the threshold, the location of the converter fault can be located.

Images