A power distribution network edge side fault research and judgment method and system containing distributed photovoltaics

Abstract

This application relates to the field of power system fault detection technology, specifically to a method and system for fault assessment at the edge of a distribution network including distributed photovoltaic power. The method includes: acquiring three-phase voltage and current sequences from each node; obtaining the current distortion rate, fundamental positive-sequence active power, fundamental direction angle of the positive-sequence current, and fundamental phase of the positive-sequence voltage for each node; obtaining a first distortion degree based on the correlation between the fundamental positive-sequence active power and the current distortion rate, and determining a reference distortion rate; obtaining a second distortion degree based on the change in current distortion rate relative to the reference distortion rate after a fault occurs; obtaining the distortion attenuation ratio based on the attenuation degree of the current distortion change; obtaining fault impact offset characteristics based on the directional offset relative to the forward and reverse feed references; and concatenating the three-phase characteristics into a fault feature vector and comparing it with the intermediate measurement vectors for each fault type to determine the fault type. This application improves the accuracy of fault assessment at the edge of the distribution network.

Description

Technical Field

[0001] This application relates to the field of power system fault detection technology, specifically to a method and system for fault assessment on the edge side of a distribution network containing distributed photovoltaic power. Background Technology

[0002] Large-scale integration of distributed photovoltaic (PV) power into a 10kV low-current grounding multi-phase distribution grid enables local consumption of renewable energy and self-healing of faults in the active distribution network. However, due to the weak single-phase grounding fault current and insignificant fault characteristics in such distribution networks, traditional fault assessment techniques cannot meet the fault diagnosis needs after PV grid connection, resulting in low accuracy and high misjudgment rate, which can lead to fault retention in the distribution network.

[0003] Existing fault assessment methods typically utilize the fundamental component, fifth harmonic component, and intrinsic mode energy obtained from empirical mode decomposition of zero-sequence current as fault features, and then combine them with neural networks to achieve fault assessment; or construct a fault identification model through multi-dimensional fault features in the time domain, frequency domain, and time-frequency domain, and then achieve fault assessment.

[0004] In the scenario of distributed photovoltaic power being connected to a 10kV low-current grounded multi-power grid, the current limiting characteristics (such as current limiting threshold, waveform holding mode, etc.) of photovoltaic inverters at different connection points will differ, resulting in differences in the degree of current distortion injected into the feeder. Moreover, this difference will change in real time with the photovoltaic power output. At the same time, after the distributed power source is connected, the power flow direction on the edge side will also change, thereby generating bidirectional power.

[0005] Therefore, existing methods ignore the differences in current limiting characteristics at the edge and the impact of changes in power flow direction, which leads to deviations in fault feature extraction during fault detection and inaccurate fault assessment results. Summary of the Invention

[0006] To address the technical problem of inaccurate fault assessment results, this application provides a fault assessment method for the edge side of a distribution network containing distributed photovoltaic power. The specific technical solution adopted is as follows:

[0007] Firstly, this application proposes a method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power, which includes the following steps:

[0008] The three-phase voltage and three-phase current sequences of each node are collected, and the current distortion rate, fundamental positive sequence active power, positive sequence current fundamental direction angle and positive sequence voltage fundamental phase of each node are obtained through preprocessing and harmonic decomposition.

[0009] The first distortion degree is obtained based on the correlation between the fundamental positive sequence active power and the current distortion rate of each node, and the reference distortion degree and reference distortion rate are determined by the first distortion degree and the mean current distortion rate when there is no fault.

[0010] After a fault occurs, the second distortion degree of each node is obtained based on the change of the current distortion rate after the fault relative to the reference distortion rate, and the correlation between the fundamental positive sequence active power fluctuation and the change of current distortion after the fault.

[0011] The distortion attenuation ratio of each node is obtained based on the degree of attenuation of the current distortion change over time after the fault occurs.

[0012] During normal, fault-free operation, a directional reference is established between the direction angle of the positive sequence current and the phase of the positive sequence voltage under both power forward and power reverse conditions. The fault impact offset characteristics are obtained based on the degree of offset of the directionality of each node relative to the directional reference of its respective operating condition after a fault.

[0013] The second distortion degree, the distortion attenuation ratio, the first distortion degree, and the fault influence offset feature of each node's three phases are concatenated into a fault feature vector; the fault feature vector is compared with the pre-constructed measurement intermediate vectors of each fault type to determine the fault type of each node.

[0014] In one embodiment, the acquisition of the three-phase voltage sequence and three-phase current sequence of each node is performed at the measurement terminal deployed at the grid connection point of each distributed photovoltaic system, with each measurement terminal serving as a node; the sampling frequency of the three-phase voltage sequence and three-phase current sequence is 12.8kHz, and the number of sampling points within each power frequency cycle is 256.

[0015] In one embodiment, the preprocessing and harmonic decomposition include:

[0016] The three-phase voltage sequence and the three-phase current sequence under various operating conditions are low-pass filtered to eliminate noise interference. Then, Fourier transforms are performed to obtain the complex current and voltage numbers of the fundamental and each harmonic. The current distortion rate is calculated based on the complex current numbers. The positive sequence and zero sequence components of the three-phase voltage sequence and the three-phase current sequence are obtained based on the symmetrical component method. Then, the fundamental positive sequence active power, the fundamental direction angle of the positive sequence current, and the fundamental phase of the positive sequence voltage are obtained. The number of sampling points for the Fourier transform is set to 256, the frequency band is set to 13, and the windowing method is Hanning window.

[0017] In one embodiment, obtaining the first distortion degree based on the correlation between the fundamental positive-sequence active power and the current distortion rate at each node includes:

[0018] During normal, fault-free operation, the fundamental positive-sequence active power and the current distortion rate for all cycles are normalized to their respective maximum values. The first distortion degree is obtained by using the deviation of the normalized value of the fundamental positive-sequence active power from its mean as the independent variable and the deviation of the normalized value of the current distortion rate from its mean as the dependent variable, and by employing a univariate linear regression slope relationship. The larger the first distortion degree, the more sensitive the current distortion rate is to changes in the fundamental positive-sequence active power.

[0019] In one embodiment, obtaining the second distortion degree of each node includes:

[0020] The current distortion rate after the fault is obtained in chronological order relative to the reference distortion rate for each cycle. The fundamental positive-sequence active power of each cycle after the fault is formed into a first sequence, and the current distortion rate after the fault is formed into a second sequence. The first sequence and the second sequence are then subjected to first-order difference and normalized. The inner product of the difference value sequence of the normalized first sequence and the difference value sequence of the normalized second sequence is calculated. The inner product is converted into a normalized parameter that is positively related to the degree of fault impact according to a preset formula, and the rate of change between the first distortion degree and the reference distortion degree after the fault is obtained. The maximum value between the normalized parameter and the rate of change is taken as the second distortion degree of the node.

[0021] In one embodiment, obtaining the distortion attenuation ratio of each node based on the attenuation degree of current distortion change over time after a fault occurs includes:

[0022] The average current distortion rate of the first four cycles and the average current distortion rate of the last four cycles after the fault are obtained in chronological order. The ratio of the difference between the two to the average current distortion rate of the first four cycles is taken as the distortion attenuation ratio of the node. The larger the distortion attenuation ratio, the more obvious the current distortion attenuation after the fault and the stronger the influence of the actual fault on the phase where the node is located.

[0023] In one embodiment, the method for obtaining the directional reference includes:

[0024] During normal, fault-free operation, the sign of the fundamental positive-sequence active power of each cycle is obtained, and the directivity between the fundamental direction angle of the positive-sequence current and the fundamental phase of the positive-sequence voltage within that cycle is calculated. Based on the sign, the directivity of each cycle during normal, fault-free operation is divided into a power forward set and a power reverse set. The mean value of the directivity in the power forward set and the mean value of the directivity in the power reverse set are used as the forward direction reference and the reverse direction reference, respectively. The acquisition of the fault impact offset characteristic includes: after a fault occurs, selecting the corresponding directivity reference based on the sign of the fundamental positive-sequence active power of that cycle, calculating the degree of offset of the directivity of that cycle relative to the selected directivity reference, and taking the largest offset among all cycles after the fault as the fault impact offset characteristic.

[0025] In one embodiment, the construction of the measurement intermediate vector includes:

[0026] For each fault type, multiple simulations are performed, and the mean of the fault feature vectors of each node obtained from each simulation is used as the measurement intermediate vector for that fault type. The cosine similarity between the fault feature vectors obtained from each simulation and the measurement intermediate vector is obtained. The cosine similarity is processed based on the box curve method to obtain a lower limit threshold and an upper limit threshold. The lower limit threshold is used as the fault measurement threshold for that fault type. During normal fault-free operation, the second distortion degree, the distortion attenuation ratio, and the fault influence offset feature of each node are all set to zero. The first distortion degree of each fault type is calculated from eight cycles after the fault occurs.

[0027] In one embodiment, determining the fault type of each node includes:

[0028] Obtain the cosine similarity between the fault feature vector of the node and the measurement intermediate vector of each fault type. The fault type with the largest cosine similarity value and a cosine similarity value that is not lower than the fault measurement threshold of the corresponding fault type is taken as the fault type of the node. The number of simulations for each fault type is not less than 30. The acquisition period during normal fault-free operation contains 300 cycles, and the acquisition period after the fault occurs contains 8 cycles.

[0029] On the other hand, this application provides a fault assessment system for the edge side of a distribution network containing distributed photovoltaics, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the above-mentioned fault assessment method for the edge side of a distribution network containing distributed photovoltaics.

[0030] The beneficial effects of this application are as follows:

[0031] This application highlights the correlation between photovoltaic power output fluctuations and current distortion rate by measuring data such as current phasors in the electrical parameters of the edge side, reflecting the degree to which the current distortion rate changes with photovoltaic power output fluctuations during normal, fault-free periods. Furthermore, by measuring the differences in electrical parameters of the edge side before and after a fault, it highlights the degree to which the current distortion rate deviates from the pre-fault correlation with photovoltaic power output fluctuations due to the fault, reflecting the severity of the current distortion caused by the fault.

[0032] By measuring the differences in electrical parameters caused by different power directions when distributed power sources are connected at the edge, the characteristics of bidirectional changes in current direction with active power forward and reverse transmission are highlighted. This solves the problem that existing fault assessment methods ignore the differences in current limiting characteristics at the edge and the influence of changes in power flow direction, which leads to inaccurate assessment results. It enables accurate assessment of fault types at the edge of distributed photovoltaic distribution networks. Attached Figure Description

[0033] Figure 1 A flowchart illustrating a fault assessment method for a distribution network edge-side containing distributed photovoltaic power, provided as an embodiment of this application;

[0034] Figure 2 This is a schematic diagram illustrating a scenario for fault assessment at the edge of a distribution network containing distributed photovoltaic power, provided in one embodiment of this application.

[0035] Figure 3 This is a schematic diagram illustrating the deviation of current direction under power forward and power reverse operation conditions, provided in one embodiment of this application.

[0036] Explanation of reference numerals in the attached figures

[0037] M represents busbar, K represents circuit breaker, T represents measurement terminal, PV represents distributed photovoltaic; U represents the phase direction of voltage, I+ represents the current direction when power is transmitted forward, I- represents the current direction when power is transmitted backward, and I~ represents the direction of current deviation caused by fault current. Detailed Implementation

[0038] The following description, in conjunction with the accompanying drawings, details a specific scheme for a fault assessment method for a distribution network edge side containing distributed photovoltaic power, as provided in this application.

[0039] Please see Figure 1 The diagram illustrates a flowchart of a fault assessment method for a distribution network edge side containing distributed photovoltaic power, according to an embodiment of this application. The method includes the following steps:

[0040] Step S001: Collect the three-phase voltage sequence and three-phase current sequence of each node, and obtain the current distortion rate, fundamental positive sequence active power, positive sequence current fundamental direction angle and positive sequence voltage fundamental phase of each node through preprocessing and harmonic decomposition.

[0041] In scenarios where distributed photovoltaic (PV) power is connected to a 10kV low-current grounding multi-control grid, the current-limiting characteristics of PV inverters at different connection points vary, and the power flow direction on the edge side changes depending on the matching relationship between PV output and local load. This results in varying and inconsistent current distortion levels at different nodes. Therefore, it is necessary to first acquire electrical parameters at the edge side that simultaneously reflect both PV output and current waveform distortion levels as the data foundation for subsequent fault feature extraction.

[0042] Specifically, each measurement terminal is treated as a node, and these terminals are deployed at the grid connection points of distributed photovoltaic systems, with one measurement terminal corresponding to one distributed photovoltaic system. For example... Figure 2 As shown, it illustrates a scenario diagram of fault assessment on the edge side of a distribution network containing distributed photovoltaics provided in an embodiment of this application. M represents the bus, K represents the circuit breaker, T represents the measurement terminal (edge ​​side), and PV represents distributed photovoltaics. Each measurement terminal is deployed at the grid connection point of each distributed photovoltaic, thereby enabling the direct acquisition of electrical parameters at this node on the edge side.

[0043] In the power distribution network simulation system, the three-phase voltage and three-phase current sequences at each measurement terminal location are collected under normal, fault-free operation and under various fault types. Each case is simulated N times, and the value of N must at least meet the basic measurement requirements. In a specific embodiment of this application, the empirical value of N is 30. The sampling frequency of the three-phase voltage and three-phase current is set to 12.8kHz, meaning there are 256 sampling points within each power frequency cycle. The sampling period under normal, fault-free operation includes 300 cycles; the sampling period after a fault occurs is 8 cycles after the fault occurs.

[0044] Preprocessing: The three-phase voltage and current sequences under each operating condition are used as inputs to a low-pass filtering algorithm. The input data are filtered to eliminate noise interference, resulting in noise-free three-phase voltage and current sequences. The low-pass filtering algorithm is a well-known technique and will not be described in detail here.

[0045] Harmonic decomposition: The preprocessed three-phase voltage and current sequences are used as inputs to the Fourier transform algorithm to obtain the complex current and voltage values ​​of the fundamental and each harmonic. The sampling points are set to 256, the frequency band to 13, the windowing method to Hanning, and the sampling frequency to 12.8 kHz. Based on the complex current values ​​of the fundamental and each harmonic, the current distortion rate is calculated. Furthermore, based on the symmetrical component method, the positive-sequence and zero-sequence components of the three-phase voltage and current sequences are obtained, and the fundamental current direction angle, fundamental voltage phase, and fundamental positive-sequence active power are calculated. The Fourier transform algorithm, the calculation of the current distortion rate, and the symmetrical component method are all well-known techniques and will not be elaborated further in this application.

[0046] It should be noted that the electrical parameter measurements in this application are performed on any one phase of a three-phase circuit, and the measurement methods and logic are the same for each phase.

[0047] Thus, the current distortion rate, fundamental positive sequence active power, positive sequence current fundamental direction angle, and positive sequence voltage fundamental phase of each node were obtained.

[0048] Step S002: Obtain the first distortion degree and determine the reference distortion rate based on the correlation between the fundamental positive sequence active power and the current distortion rate.

[0049] The degree of current distortion generated by a photovoltaic inverter is determined by factors such as the photovoltaic output and the inverter's current-limiting characteristics. The fundamental positive-sequence active power reflects the photovoltaic output at a given node, while the current distortion rate reflects the distortion of the current waveform at that node. During normal operation, the degree of current distortion at different nodes varies with the photovoltaic output, and there is no fixed current distortion value. Therefore, this application does not use a fixed distortion threshold, but instead uses the linear sensitivity of the node's current distortion rate to changes in the fundamental positive-sequence active power during normal, fault-free operation as a benchmark. This internalizes the differences in the current-limiting characteristics of photovoltaic inverters at each node into the node's own benchmark characteristic.

[0050] Specifically, the fundamental positive sequence active power of all cycles during normal and fault-free operation is used as input, and the fundamental positive sequence active power is normalized based on the maximum value among all fundamental positive sequence active power, so that the normalized value of the fundamental positive sequence active power is specified to be between 0 and 1; similarly, the current distortion rate of all cycles during normal and fault-free operation is normalized.

[0051] The first distortion degree at each node is constructed using the normalized values ​​of both, and its calculation relationship is as follows:

[0052]

[0053] This represents the first distortion degree of the i-th node. This represents the cycle at each node. This indicates the total number of cycles during normal, trouble-free operation. This represents the normalized value of the fundamental positive-sequence active power for each cycle. This represents the normalized mean of all fundamental positive-sequence active power at each node. This represents the normalized value of the current distortion rate per cycle. This represents the normalized mean of all current distortion rates at each node.

[0054] In this process, after a fault occurs, the fundamental positive sequence active power of the eight cycles following the fault is used as input. The fundamental positive sequence active power is normalized based on the maximum value among all fundamental positive sequence active powers. The first distortion degree of the node at the time of the fault is calculated in the same way.

[0055] If the fundamental positive sequence active power is detected to remain constant within the sampling time window during the eight cycles following the fault, resulting in the variance denominator term being zero or the calculation result being invalid, the first distortion parameter at the time of the fault is assigned as the baseline distortion before the fault.

[0056] The calculation formula for the first distortion degree is based on the slope equation in univariate linear regression, which can reflect the linear sensitivity of the current distortion rate of the node to changes in photovoltaic power output. The larger the first distortion degree, the more sensitive the current distortion caused by the photovoltaic inverter at the node is to changes in photovoltaic power output.

[0057] Since the photovoltaic output fluctuations, current limiting characteristics, and power flow direction at each node may change after a fault, for each node, the first distortion degree during normal fault-free operation is used as the benchmark distortion degree, the average current distortion rate of all cycles during normal fault-free operation is used as the benchmark distortion rate, and the electrical parameter characteristics of the node itself during normal fault-free operation are used as the comparison benchmark for subsequent fault feature extraction.

[0058] At this point, the first distortion degree, the baseline distortion degree, and the baseline distortion rate of each node have been obtained.

[0059] Step S003: After the fault occurs, obtain the second distortion degree based on the change of the current distortion rate relative to the reference distortion rate.

[0060] Photovoltaic output will still fluctuate before and after a fault. Therefore, after a fault occurs, current distortion will be affected by both photovoltaic output and the fault, causing the degree of current distortion to differ from that during normal, fault-free operation. Before and after a fault, the current distortion rate includes both the current distortion rate generated by distributed photovoltaic power and the current distortion rate caused by the fault. The occurrence of current distortion rate caused by the fault will reduce the linear relationship between current distortion rate and photovoltaic output; the more severely this linear relationship is disrupted, the stronger the impact of the fault on that phase circuit.

[0061] Therefore, firstly, for the current distortion rate of the eight cycles after the fault, the rate of change of current distortion before and after the fault is measured sequentially in time for each cycle, and the calculation relationship is as follows:

[0062]

[0063] Let represent the rate of change of current distortion at node i in the j-th cycle after the fault occurs, where j is an integer not less than 1 and not greater than 8. This represents the current distortion rate of the i-th node in the j-th cycle after the fault occurs. This represents the baseline distortion rate for each node during normal, fault-free operation.

[0064] After a fault occurs, the greater the degree of change in the current distortion at the node, the greater the rate of change in the current distortion. This indicates that the current distortion caused by the fault and the photovoltaic output in that phase is greater, and the impact of the fault on that phase circuit may be stronger.

[0065] Furthermore, the fundamental positive-sequence active power within eight cycles is arranged into a first sequence according to time order, and the current distortion rate of change within eight cycles is arranged into a second sequence. The first and second sequences are used as inputs to a first-order difference algorithm to obtain the difference value sequences of the first and second sequences, respectively. The first-order difference algorithm is a known technique and will not be described in detail in this application.

[0066] Using the difference value sequence of the first sequence as input, all difference values ​​generated by the first sequence are normalized based on the maximum value, so that the normalized difference values ​​of the first sequence are within the range of -1 to 1. Similarly, the difference value sequence of the second sequence is normalized. Here, the maximum value represents the maximum absolute value of the difference values ​​within all nodes.

[0067] The difference sequences of the normalized first sequence and the difference sequences of the normalized second sequence are used as inputs. The inner product of the two input sequences is calculated, and the inner product is divided by the sequence dimension. The value of the inner product ranges from -1 to 1. The inner product reflects the relationship between the fluctuation of photovoltaic output and the fluctuation of the current distortion rate after a fault occurs: when the current distortion rate fluctuates with the photovoltaic output, the value of the inner product approaches 1, indicating that the fault has a weak impact on the phase circuit; otherwise, the value of the inner product approaches -1, indicating that the fault has a strong impact on the phase circuit. The calculation of the inner product is a well-known technique and will not be elaborated further in this application.

[0068] Considering that the linear relationship between photovoltaic output and current distortion rate becomes more significant when the fault impact coincides with the photovoltaic output frequency, the inner product will approach 1. However, this does not necessarily mean the fault had no impact. Therefore, the first distortion degree is measured eight cycles after the fault, and the rate of change between this first distortion degree and the reference distortion degree is measured, i.e.:

[0069]

[0070] This represents the rate of change between the first distortion degree of the i-th node and the baseline distortion degree eight cycles after the fault. This represents the first distortion degree of the i-th node eight cycles after the fault. This represents the baseline distortion degree of the i-th node.

[0071] Simultaneously, the inner product is converted into a normalized parameter that is positively correlated with the degree of fault impact using the following formula, thereby adjusting the relationship between the inner product and the rate of change to a positive relationship and normalizing the inner product:

[0072]

[0073] This represents the normalized parameter obtained by transforming the inner product corresponding to the i-th node. This represents the inner product between the difference sequence of the normalized first sequence corresponding to the i-th node and the difference sequence of the normalized second sequence.

[0074] The larger the normalized parameter or the larger the rate of change, the stronger the impact of the fault on the phase circuit; the smaller the normalized parameter or the smaller the rate of change, the weaker the impact of the fault on the phase circuit. Therefore, the maximum value of the normalized parameter and the rate of change is taken as the second distortion degree of the node; the larger the second distortion degree, the stronger the impact of the fault on the phase, and its value range is greater than or equal to 0 and less than or equal to 1. It should be noted that when the first distortion degree after the fault is less than the reference distortion degree, the rate of change is positive; when the first distortion degree after the fault is greater than the reference distortion degree, the rate of change is negative, which does not affect the value range of the second distortion degree.

[0075] At this point, the second distortion degree of each node under different fault types has been obtained.

[0076] Step S004: Obtain the distortion attenuation ratio based on the attenuation degree of the current distortion change.

[0077] For each phase, when a fault occurs, the transient zero-mode current of the actual fault will decay, while the low-order harmonics of the edge-side terminal equipment still exist during the fault period, and the current will not decay. If the current decays, the intensity of the current distortion will decrease, and the influence of photovoltaic output fluctuations on the rate of change of current distortion will indirectly increase. Then, the relationship between the current distortion rate and photovoltaic output will be more similar to that of normal fault-free operation. At this time, the average value of the rate of change of current distortion in the latter four cycles will be less than the average value of the rate of change of current distortion in the first four cycles. The greater the difference between the two relative to the average value of the rate of change of current distortion in the first four cycles, the faster the current decays. The more obvious the decay of the fault-induced current distortion after the fault, the more obvious it is. This proves that the distortion is not a continuous source distortion, but a transient distortion caused by a real fault. Therefore, the impact of the fault on this phase is stronger.

[0078] Specifically, the mean of the rate of change of current distortion for the first four cycles after the fault occurs and the mean of the rate of change of current distortion for the last four cycles are measured in chronological order, and the two are combined using the following formula:

[0079]

[0080] This represents the percentage of distortion attenuation at the i-th node when a fault occurs. This represents the average rate of change of current distortion for the first four cycles after the fault occurs. This represents the average rate of change of current distortion over the four cycles following the fault occurrence. Where, when When the value is 0, the distortion attenuation ratio is set to 0.

[0081] The distortion attenuation ratio reflects the degree of attenuation of the current distortion rate of the node after the fault occurs. The more obvious the attenuation, the larger the distortion attenuation ratio, and the stronger the impact of the fault on that phase.

[0082] Thus, the distortion attenuation ratio of each node under different fault types has been obtained.

[0083] Step S005: Obtain the fault impact offset characteristics based on the offset of the directional relative forward and reverse feed references.

[0084] When distributed generation (DG) is connected to the distribution network feeder, the output power of the power source is random and mismatched with the timing of the local load: when the DG output is greater than the local load, the excess power will flow back to the bus, forming a power reverse flow; when the DG output is less than the local load, the power flows from the bus to the line, forming a power forward flow. Therefore, two operating conditions will occur: power forward flow and power reverse flow, and the current direction on the edge side will also change bidirectionally with the active power transmission and reception relationship. If only the absolute values ​​of the current direction angle or voltage phase angle are used as the criteria, the same fault under the two operating conditions will be confused in terms of characteristics.

[0085] like Figure 3 The diagram illustrates the deviation of current direction under power forward and power reverse operation conditions according to an embodiment of this application. U represents the phase direction of the voltage, I+ represents the current direction during power forward, I- represents the current direction during power reverse, and I~ represents the direction of the current after deviation caused by the fault current. During power forward, the current direction angle and voltage phase angle tend to be in the same direction; during power reverse, the current direction angle and voltage phase angle tend to be in opposite directions; and when a fault occurs, the presence of the fault current causes a shift in the phase angle of the current on the edge side.

[0086] Specifically, during normal, fault-free operation, for each cycle, the sign of its fundamental positive-sequence active power is taken as sgn(p). If the power is reversed, the fundamental positive-sequence active power is less than 0, and sgn(p) is negative; otherwise, sgn(p) is positive. The directionality between the fundamental direction angle of the positive-sequence current and the fundamental phase of the positive-sequence voltage within that cycle is calculated using the following formula:

[0087]

[0088] This represents the directionality between the fundamental direction angle of the positive-sequence current and the fundamental phase of the positive-sequence voltage during the k-th cycle of the i-th node under normal, fault-free operation. Indicates sign-taking operation. This represents the direction angle of the fundamental current of the positive sequence at the i-th node in the k-th cycle. This represents the fundamental phase of the positive-sequence voltage at the i-th node in the k-th cycle. Represents a symbolic function.

[0089] Based on the consistency of sgn(p), the directionality of each cycle during normal fault-free operation is divided into a power forward set and a power reverse set. If the power forward set and the power reverse set have zero or unequal numbers during normal fault-free operation, simulation needs to be continued to supplement the number of directions in the set with fewer directions so that the number of directions in the two sets is equal. If the number of directions in the two sets is equal, the average value of all directions in the power forward set is taken as the forward direction reference, and the average value of all directions in the power reverse set is taken as the reverse direction reference.

[0090] After the fault occurs, the directivity between the fundamental direction angle of the positive-sequence current and the fundamental phase of the positive-sequence voltage is measured for eight cycles. The stronger the fault current caused by the fault, the more significant the deviation of the current direction angle will be, and the greater the difference between the directivity and the reference. Therefore, the sign of the fundamental positive-sequence active power sgn(p) of this cycle is first determined. If sgn(p) is positive, the calculation formula for the fault-affected deviation is:

[0091]

[0092] This represents the offset of the fault impact on the i-th node in the j-th cycle after the fault occurs. This represents the directionality between the fundamental direction angle of the positive-sequence current and the phase of the fundamental positive-sequence voltage at the i-th node in the j-th cycle after a fault occurs. This represents the forward feed direction reference for the i-th node; if sgn(p) is negative, the forward feed direction reference is replaced with the reverse feed direction reference.

[0093] The fault influence offset reflects the degree of deviation in the current direction angle caused by the fault current. The larger the fault influence offset, the stronger the impact of the fault on that phase. The largest fault influence offset among the eight cycles is taken as the fault influence offset characteristic of that node.

[0094] Thus, the fault impact offset characteristics of each node under different fault types have been obtained.

[0095] Step S006: After splicing the three-phase features into a fault feature vector, compare it with the intermediate measurement vector of each fault type to determine the fault type.

[0096] A single feature can only reflect one aspect of the fault's behavior: the second distortion degree reflects the extent to which the fault disrupts the linear relationship between distortion rate and photovoltaic output; the distortion attenuation ratio reflects whether the distortion caused by the fault has transient attenuation characteristics; the fault impact offset feature reflects the degree of current phase angle shift caused by the fault current; and the first distortion degree reflects the sensitivity of the node's current distortion rate to changes in photovoltaic output after the fault. Therefore, it is necessary to piece together the features of each phase into a whole to comprehensively characterize the node's state during the fault.

[0097] Specifically, for the data of each node operating normally and for each fault type, the second distortion degree, distortion attenuation ratio, first distortion degree, and fault impact offset feature calculated for the three phases ABC are concatenated to obtain a 3×4 fault feature vector. Among them, the second distortion degree, distortion attenuation ratio, and fault impact offset feature during normal fault-free operation are all set to 0; the first distortion degree under each fault type is calculated by the eight cycles after the fault occurs according to the method described in step S002.

[0098] Since the changes in node characteristics vary under different fault types, it is necessary to establish comparable decision criteria for each fault type and provide decision thresholds that can tolerate simulation and measurement fluctuations in order to give a definite fault type conclusion in actual judgment.

[0099] Specifically, for each fault type, the fault feature vectors obtained from all simulations are taken, and the mean of all fault feature vectors is calculated as the intermediate measurement vector for that fault type. The cosine similarity between all fault feature vectors and the intermediate measurement vector is taken, and all cosine similarity values ​​are processed using the box-and-loop method to obtain a lower threshold and an upper threshold. The lower threshold is used as the fault measurement threshold for each fault type. The calculation of the cosine similarity and the box-and-loop method are well-known techniques and will not be elaborated upon further in this application. In a specific embodiment of this application, the fault type includes single-phase ground fault, two-phase ground fault, and two-phase short-circuit fault, which is determined by the actual application and is not specifically limited in this application.

[0100] In actual fault assessment, the second distortion degree, distortion attenuation ratio, first distortion degree, and fault influence offset characteristics of the node on the three phases A, B, and C are measured respectively, and the results are concatenated to obtain a fault feature vector. This fault feature vector is then compared with the intermediate measurement vectors of each fault type using cosine similarity. If the cosine similarity with the intermediate measurement vector of a certain fault type is not lower than the fault measurement threshold for that type, it indicates that the node has a fault of that type. When multiple fault types meet the above conditions, the fault type with the highest cosine similarity value is taken as the fault type of the node. In addition, based on the single-end traveling wave ranging algorithm, traveling wave ranging is performed upstream and downstream of the fault node to identify the fault location. The single-end traveling wave ranging algorithm is a well-known algorithm and will not be described in detail in this application.

[0101] This completes the assessment of fault types at each node on the edge side of the distributed photovoltaic power distribution network.

[0102] Based on the same inventive concept as the above method, this embodiment of the invention also provides a fault assessment system for the edge side of a distribution network containing distributed photovoltaic power. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the above-described fault assessment method for the edge side of a distribution network containing distributed photovoltaic power. The system includes a data acquisition module, an electrical parameter measurement module, and a fault assessment module: the data acquisition module is used to acquire the electrical parameters of each node and perform preprocessing and harmonic decomposition; the electrical parameter measurement module is used to measure the fault characteristics of the electrical parameters at each node under different operating conditions, namely, the first distortion degree, the second distortion degree, the distortion attenuation ratio, and the fault influence offset characteristics; the fault assessment module identifies different types of faults at each node based on the fault characteristics.

[0103] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power, characterized in that, The method includes the following steps: The three-phase voltage and three-phase current sequences of each node are collected, and the current distortion rate, fundamental positive sequence active power, positive sequence current fundamental direction angle and positive sequence voltage fundamental phase of each node are obtained through preprocessing and harmonic decomposition. The first distortion degree is obtained based on the correlation between the fundamental positive sequence active power and the current distortion rate of each node, and the reference distortion degree and reference distortion rate are determined by the first distortion degree and the mean current distortion rate when there is no fault. After a fault occurs, the second distortion degree of each node is obtained based on the change of the current distortion rate after the fault relative to the reference distortion rate, and the correlation between the fundamental positive sequence active power fluctuation and the change of current distortion after the fault. The distortion attenuation ratio of each node is obtained based on the degree of attenuation of the current distortion change over time after the fault occurs. During normal, fault-free operation, a directional reference is established between the direction angle of the positive sequence current and the phase of the positive sequence voltage under both power forward and power reverse conditions. The fault impact offset characteristics are obtained based on the degree of offset of the directionality of each node relative to the directional reference of its respective operating condition after a fault. The second distortion degree, the distortion attenuation ratio, the first distortion degree, and the fault influence offset feature of each node's three phases are concatenated into a fault feature vector; the fault feature vector is compared with the pre-constructed measurement intermediate vectors of each fault type to determine the fault type of each node.

2. The method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power as described in claim 1, characterized in that, The acquisition of the three-phase voltage and three-phase current sequences of each node is performed at the measurement terminals deployed at the grid connection point of each distributed photovoltaic system, with each measurement terminal serving as a node; the sampling frequency of the three-phase voltage and three-phase current sequences is 12.8 kHz, and the number of sampling points within each power frequency cycle is 256.

3. The method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power as described in claim 1, characterized in that, The preprocessing and harmonic decomposition include: The three-phase voltage sequence and the three-phase current sequence under various operating conditions are low-pass filtered to eliminate noise interference. Then, Fourier transforms are performed to obtain the complex current and voltage numbers of the fundamental and each harmonic. The current distortion rate is calculated based on the complex current numbers. The positive sequence and zero sequence components of the three-phase voltage sequence and the three-phase current sequence are obtained based on the symmetrical component method. Then, the fundamental positive sequence active power, the fundamental direction angle of the positive sequence current, and the fundamental phase of the positive sequence voltage are obtained. The number of sampling points for the Fourier transform is set to 256, the frequency band is set to 13, and the windowing method is Hanning window.

4. The method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power as described in claim 1, characterized in that, The method of obtaining the first distortion degree based on the correlation between the fundamental positive-sequence active power and the current distortion rate at each node includes: During normal, fault-free operation, the fundamental positive-sequence active power and the current distortion rate for all cycles are normalized to their respective maximum values. The first distortion degree is obtained by using the deviation of the normalized value of the fundamental positive-sequence active power from its mean as the independent variable and the deviation of the normalized value of the current distortion rate from its mean as the dependent variable, and by employing a univariate linear regression slope relationship. The larger the first distortion degree, the more sensitive the current distortion rate is to changes in the fundamental positive-sequence active power.

5. The method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power as described in claim 4, characterized in that, The process of obtaining the second distortion degree for each node includes: The current distortion rate after the fault is obtained in chronological order relative to the reference distortion rate for each cycle. The fundamental positive-sequence active power of each cycle after the fault is formed into a first sequence, and the current distortion rate after the fault is formed into a second sequence. The first sequence and the second sequence are then subjected to first-order difference and normalized. The inner product of the difference value sequence of the normalized first sequence and the difference value sequence of the normalized second sequence is calculated. The inner product is converted into a normalized parameter that is positively related to the degree of fault impact according to a preset formula, and the rate of change between the first distortion degree and the reference distortion degree after the fault is obtained. The maximum value between the normalized parameter and the rate of change is taken as the second distortion degree of the node.

6. The method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power as described in claim 5, characterized in that, The method of obtaining the distortion attenuation ratio of each node based on the attenuation degree of current distortion change over time after a fault occurs includes: The average current distortion rate of the first four cycles and the average current distortion rate of the last four cycles after the fault are obtained in chronological order. The ratio of the difference between the two to the average current distortion rate of the first four cycles is taken as the distortion attenuation ratio of the node. The larger the distortion attenuation ratio, the more obvious the current distortion attenuation after the fault and the stronger the influence of the actual fault on the phase where the node is located.

7. The method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power as described in claim 1, characterized in that, The method for obtaining the directional reference includes: During normal, fault-free operation, the sign of the fundamental positive-sequence active power of each cycle is obtained, and the directivity between the fundamental direction angle of the positive-sequence current and the fundamental phase of the positive-sequence voltage within that cycle is calculated. Based on the sign, the directivity of each cycle during normal, fault-free operation is divided into a power forward set and a power reverse set. The mean value of the directivity in the power forward set and the mean value of the directivity in the power reverse set are used as the forward direction reference and the reverse direction reference, respectively. The acquisition of the fault impact offset characteristic includes: after a fault occurs, selecting the corresponding directivity reference based on the sign of the fundamental positive-sequence active power of that cycle, calculating the degree of offset of the directivity of that cycle relative to the selected directivity reference, and taking the largest offset among all cycles after the fault as the fault impact offset characteristic.

8. The method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power as described in claim 1, characterized in that, The construction of the intermediate measurement vector includes: For each fault type, multiple simulations are performed, and the mean of the fault feature vectors of each node obtained from each simulation is used as the measurement intermediate vector for that fault type. The cosine similarity between the fault feature vectors obtained from each simulation and the measurement intermediate vector is obtained. The cosine similarity is processed based on the box curve method to obtain a lower limit threshold and an upper limit threshold. The lower limit threshold is used as the fault measurement threshold for that fault type. During normal fault-free operation, the second distortion degree, the distortion attenuation ratio, and the fault influence offset feature of each node are all set to zero. The first distortion degree of each fault type is calculated from eight cycles after the fault occurs.

9. The method for fault assessment on the edge side of a distribution network containing distributed photovoltaic power as described in claim 8, characterized in that, Determining the fault type of each node includes: Obtain the cosine similarity between the fault feature vector of the node and the measurement intermediate vector of each fault type. The fault type with the largest cosine similarity value and a cosine similarity value that is not lower than the fault measurement threshold of the corresponding fault type is taken as the fault type of the node. The number of simulations for each fault type is not less than 30. The acquisition period during normal fault-free operation contains 300 cycles, and the acquisition period after the fault occurs contains 8 cycles.

10. A fault assessment system for the edge side of a distribution network including distributed photovoltaic power, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the fault assessment method for the edge side of a distribution network containing distributed photovoltaic power as described in any one of claims 1-9.

Images