A primary and secondary fusion ring network box for distribution network fault on-site research and isolation

Abstract

The present application relates to distribution network fault technical field, especially in one kind distribution network fault on-the-spot research and judge isolation primary and secondary fusion ring network box, including memory, processor and the computer program stored in memory and running on the processor, when the processor executes the computer program realizes the following steps: real-time acquisition multi-dimensional electrical quantity, combine environment working condition matching history normal feature cluster, calculate the Mahalanobis distance of real-time feature vector and mean vector to carry out abnormal deviation detection, power mutation ratio is constructed by using active and reactive power change rate ratio, discriminates fault crossing behavior, through the first order inertia link constructs the persistence confidence attenuation factor, carries out time dimension check to initial fault characteristic quantity, effectively restrains transient disturbance interference, thereby overcomes the defect that fixed threshold method is easy to misjudge under the distributed photovoltaic access scene, realizes the progressive identification and continuous check of fault feature, without relying on the main station communication can complete the high-precision, high-reliable on-the-spot research and judge and isolation of fault.

Description

Technical Field

[0001] This invention relates to the field of distribution network fault technology, and in particular to a primary and secondary integrated ring network box for on-site analysis and isolation of distribution network faults. Background Technology

[0002] In scenarios where distributed photovoltaic (PV) systems are connected to the distribution network on a large scale, the integrated primary and secondary ring network box collects electrical quantities such as three-phase current, zero-sequence current, and voltage through the distribution terminal unit. Based on preset fixed thresholds or linear weighted summation of multiple electrical quantities, it makes on-site judgment of short-circuit faults. In traditional distribution networks with extremely low distributed PV penetration, this method can effectively identify short-circuit faults and trigger isolation actions, and has the advantages of logical intuitiveness and fast response speed.

[0003] As the scale of distributed photovoltaic (PV) grid connection continues to expand, the power grid operating environment becomes more complex. Fault points simultaneously bear bidirectional fault current feeds from both the system side and the PV side. During normal operation, the output power of the PV grid-connected system will naturally change drastically with weather changes such as cloud cover. Similarly, real short-circuit faults will also cause sudden changes in current and voltage values. The current and voltage value changes caused by the two are highly similar and mixed. The fixed threshold comparison method cannot accurately distinguish whether the change in current value comes from normal PV output fluctuations or real line short circuits, and is prone to false triggering during normal fluctuations. At the same time, the existing linear weighting method lacks the ability to suppress transient current disturbances such as lightning strikes and tree branches touching the wires, and is prone to producing incorrect judgment results under transient disturbances.

[0004] In the case of a real short-circuit fault, the existing integrated primary and secondary ring network box is difficult to accurately identify and disconnect the fault section from various complex operating conditions, which limits the reliability of on-site fault assessment and the continuity of power supply. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a primary and secondary integrated ring network box for on-site analysis and isolation of distribution network faults, in order to solve the problem that existing ring network boxes are unable to distinguish between photovoltaic fluctuations and actual short circuits, which leads to easy misjudgment and missed judgment in fault analysis.

[0006] This invention provides a primary and secondary integrated ring network box for local analysis and isolation of distribution network faults, 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 performs the following steps: collecting instantaneous values ​​of three-phase current, zero-sequence current, and three-phase voltage to obtain normalized values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage.

[0007] Based on time period labels, weather type labels, and photovoltaic power output interval labels, a historical normal feature cluster is constructed from the historical feature database, and the mean vector, covariance matrix, and standard deviation of reactive power change rate are obtained statistically.

[0008] Based on the normalized values ​​of positive sequence current, zero sequence current, and positive sequence voltage, a real-time three-dimensional feature vector is constructed and its Mahalanobis distance with the mean vector is calculated. When the Mahalanobis distance is greater than the dynamic deviation threshold, the power mutation ratio is calculated based on the instantaneous active power change rate, instantaneous reactive power change rate, and standard deviation of reactive power change rate. The initial fault characteristic quantity is obtained by combining the normalized value of positive sequence voltage.

[0009] The initial fault characteristic quantity is accumulated and integrated by combining the confidence time constant to calculate the persistence confidence decay factor. The initial fault characteristic quantity is multiplied by the persistence confidence decay factor to obtain the effective fault characteristic quantity. When the effective fault characteristic quantity is greater than the confirmation threshold, a tripping command is issued to the spring operating mechanism.

[0010] Preferably, obtaining the normalized values ​​of the positive-sequence current, the zero-sequence current, and the positive-sequence voltage includes:

[0011] The instantaneous values ​​of three-phase current, zero-sequence current, and three-phase voltage are filtered using a sliding window mean filtering algorithm. Discrete sampling sequences within a complete power frequency cycle are extracted, and the effective values ​​of three-phase current and zero-sequence current are calculated respectively. Discrete Fourier transform is performed on the discrete sampling sequences of three-phase voltage instantaneous values ​​to extract the fundamental voltage phasor. The fundamental voltage phasor is transformed using the symmetrical component method to obtain the effective value of positive-sequence voltage. The effective value of positive-sequence current is extracted from the effective values ​​of three-phase current using the symmetrical component method. The effective values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage are divided by the set rated current reference value and rated voltage reference value respectively to obtain the normalized values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage.

[0012] Preferably, the method for determining the time period label, weather type label, and photovoltaic output range label is as follows:

[0013] The day is divided into multiple fixed time periods according to the 24-hour system, and the fixed time period in which the current time information is located is determined as the time period label. According to the set meteorological mapping rules, the meteorological cloud map and precipitation data are mapped as weather type labels. The instantaneous active power is divided by the rated active power of the photovoltaic grid-connected system to obtain the real-time output percentage. According to the set threshold nodes, the total output range is divided into progressive intervals, and the interval in which the real-time output percentage is located is determined as the photovoltaic output interval label.

[0014] Preferably, the formula for calculating the Mahalanobis distance is:

[0015] ;

[0016] In the formula, The Mahalanobis distance, For real-time 3D feature vectors and , This is the normalized value of the positive sequence current. This is the normalized value of the zero-sequence current. This is the normalized value of the positive sequence voltage. This is the matrix transpose symbol. It is a mean vector and , This represents the historical average of the normalized positive sequence current. The normalized value of the zero-sequence current is the historical average. The normalized historical average of the positive sequence voltage is given. Let covariance matrix be the variance matrix. is the inverse of the covariance matrix.

[0017] Preferably, the calculation of the power mutation ratio based on the instantaneous active power change rate, the instantaneous reactive power change rate, and the standard deviation of the reactive power change rate includes:

[0018] ;

[0019] In the formula, The power mutation ratio is mentioned. The instantaneous rate of change of active power, The instantaneous reactive power change rate is... This is the minimum value operator. This is the absolute value operator. This is the maximum value operator. The standard deviation of the reactive power change rate is given.

[0020] Preferably, the initial fault characteristic quantity is obtained through piecewise linear mapping operation. When the normalized value of the positive sequence voltage is less than the voltage drop threshold and the power mutation ratio is greater than or equal to the lower limit of the effective value and less than or equal to the upper limit of the effective value, the calculation formula for the initial fault characteristic quantity is: When this condition is not met, the initial fault characteristic quantity is equal to 0, where, These are the initial fault characteristic quantities. The power mutation ratio, This is the normalized value of the positive sequence voltage. This is the voltage drop threshold.

[0021] Preferably, the formula for calculating the effective fault characteristic quantity is:

[0022] ;

[0023] In the formula, For effective fault characteristic quantities, These are the initial fault characteristic quantities. It is a natural exponential function. The confidence level time constant. For the current moment, The initial time when the initial fault characteristic quantity first becomes greater than zero. For integration variables, This is the time-dimension integration operator. For step function, and in When greater than zero It equals 1 under other conditions. It equals 0.

[0024] Preferably, the method for setting the confidence time constant is as follows:

[0025] Extract all historical transient disturbance event records from the historical feature database to statistically obtain the maximum transient disturbance duration. Extract the duration of all historical short-circuit fault event records before the protection device operates to statistically obtain the minimum fault tolerance duration. Calculate the average of the maximum transient disturbance duration and the minimum fault tolerance duration as the confidence time constant.

[0026] Preferably, the method for setting the confirmation threshold is as follows:

[0027] Extract all historical short-circuit fault event records from the historical feature database, calculate the maximum value of the effective fault feature quantity corresponding to each historical short-circuit fault event record during the fault occurrence period to construct a fault feature peak distribution set, and select the 5th percentile of the fault feature peak distribution set as the confirmation threshold.

[0028] Preferably, the steps further include:

[0029] After a short-circuit fault is determined, all calculation sequences, event timestamps, and switch action status are packaged to generate a time-stamped event record and reported to the distribution master station.

[0030] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0031] In this invention, Mahalanobis distance is calculated by combining the normalized values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage with the mean vector and covariance matrix of historical normal feature clusters. The inverse of the covariance matrix is ​​used to eliminate dimensional differences in multidimensional parameters and incorporate inherent correlations, significantly distinguishing photovoltaic natural fluctuations from short-circuit faults in terms of distance magnitude. Furthermore, a power mutation ratio is constructed by calculating the ratio of the negative portion of the instantaneous active power change rate to the maximum value associated with the instantaneous reactive power change rate. This ratio is then combined with the normalized value of positive-sequence voltage for mapping operations to obtain initial fault characteristic quantities, accurately capturing the photovoltaic inverter's electrical... The fault-crossing physical phenomenon of active power reduction and reactive power injection during network short circuits eliminates active power fluctuation interference caused solely by weather changes. Finally, the duration of the initial fault characteristic quantity continuously greater than zero is accumulated and integrated, and the persistence confidence attenuation factor is calculated in combination with the confidence time constant. This causes the spikes generated by transient interferences such as lightning strikes or tree branches touching the wire to be exponentially suppressed and attenuated by the first-order inertial element. The actual short-circuit fault can be completely output as an effective fault characteristic quantity due to the accumulation of time. Thus, without relying on master station communication and fixed delay confirmation, high-precision and high-reliability on-site assessment and physical electrical isolation of short-circuit fault sections are achieved. Detailed Implementation

[0032] The embodiments of this disclosure are described in detail below. The embodiments described below are exemplary and intended to explain this disclosure, and should not be construed as limiting this disclosure.

[0033] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0034] This invention provides a primary and secondary integrated ring network box for local fault diagnosis and isolation in distribution networks, including a processor and a memory. The processor executes a computer program stored in the memory to implement a fault self-diagnosis method for single-phase smart meters. The method includes the following steps:

[0035] Step S101: Obtain normalized electrical parameter features and match them with historical normal feature clusters to obtain statistics.

[0036] In distribution network scenarios where distributed photovoltaic (PV) power is extensively integrated into primary and secondary ring network boxes, the bidirectional fault current feed into the system and PV sides, along with normal fluctuations in PV output, can couple and mask the electrical characteristics of short-circuit faults. This leads to misjudgments and omissions in existing on-site assessment methods based on fixed thresholds or linear weighted criteria for single electrical quantities. To address this, this invention synchronously collects multidimensional instantaneous electrical quantities at high frequency and performs normalization processing to eliminate dimensional differences. It then matches corresponding historical benchmark data with environmental conditions and output ranges, laying a physical and statistical data foundation for distinguishing between normal fluctuations and abnormal deviations.

[0037] Specifically, based on hardware anti-interference requirements and signal sampling rate, the sliding window length is set to 5 sampling points; according to the design specifications of the distribution network and the nameplate parameters of the primary and secondary integrated ring main unit, the rated current reference value and rated voltage reference value are set respectively; the real-time processing module of the edge computing unit is connected to the external electronic instrument transformer through the internally integrated analog-to-digital converter, and synchronously collects the instantaneous values ​​of three-phase current, zero-sequence current, and three-phase voltage of each incoming and outgoing line interval of the primary and secondary integrated ring main unit at a high-frequency sampling rate of not less than 64 points per cycle; the obtained instantaneous values ​​of three-phase current, zero-sequence current, and three-phase voltage are filtered using a sliding window mean filtering algorithm with a sliding window length of 5 sampling points to suppress high-frequency noise; while filtering, the integrity of the instrument transformer channel and the validity of the data amplitude are checked to eliminate abnormal sampling points caused by hardware transmission errors.

[0038] Furthermore, the real-time processing module of the edge computing unit extracts discrete sampling sequences within a complete power frequency cycle from the filtered instantaneous values ​​of the three-phase current, zero-sequence current, and three-phase voltage. For the discrete sampling sequences of the three-phase current and zero-sequence current within a complete power frequency cycle, the square root of the average of the sums of squares is taken to calculate the effective values ​​of the three-phase current and zero-sequence current. A discrete Fourier transform is performed on the discrete sampling sequence of the three-phase voltage within a complete power frequency cycle to extract the fundamental voltage phasor, and the fundamental voltage phasor is transformed using the symmetrical component method. The positive-sequence component is extracted to obtain the effective value of the positive-sequence voltage. The symmetrical component method is a well-known technique and will not be elaborated here. Based on the three-phase instantaneous voltage and three-phase instantaneous current, the instantaneous active power and instantaneous reactive power are calculated and obtained. This is a well-known technique and will not be elaborated here. The effective value of the positive-sequence current is extracted from the effective value of the three-phase current using the symmetrical component method. The effective values ​​of the positive-sequence current, zero-sequence current, and positive-sequence voltage are normalized by dividing them by the set rated current reference value and rated voltage reference value, respectively, thereby obtaining the dimensionless normalized values ​​of the positive-sequence current, zero-sequence current, and positive-sequence voltage.

[0039] Furthermore, the management module of the edge computing unit extracts the current time information based on the system clock, divides the entire day into multiple fixed time periods of equal duration according to the 24-hour system, and determines the fixed time period in which the current time information is located as the time period label; it obtains meteorological cloud maps and precipitation data by combining external meteorological data interfaces, and maps the meteorological cloud maps and precipitation data into weather type labels according to the set meteorological mapping rules, which include four classification standards: sunny, cloudy, overcast, and rainy / snowy weather; it divides the obtained instantaneous active power by the rated active power of the photovoltaic grid-connected system to obtain the real-time output percentage, and divides the full output range into four progressive intervals according to the threshold nodes of 20%, 50%, and 80%, and determines the interval in which the real-time output percentage is located as the photovoltaic output interval label.

[0040] Furthermore, the historical feature database stores historical steady-state operation records accumulated by the distribution network under fault-free operation conditions. Each historical steady-state operation record includes a time period label, weather type label, photovoltaic output range label, and corresponding normalized values ​​for positive-sequence current, zero-sequence current, positive-sequence voltage, and reactive power change rate. The management module of the edge computing unit uses the currently obtained time period label, weather type label, and photovoltaic output range label as a joint index to filter out the set of all historical steady-state operation records with completely matching label combinations from the historical feature database. This set constitutes a historical normal feature cluster. Mathematical statistical operations are performed on the normalized values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage recorded in the historical normal feature cluster to obtain the mean vector and covariance matrix characterizing the three-dimensional spatial distribution. All reactive power change rates recorded in the historical normal feature cluster are extracted to construct a numerical sequence. The variance of the numerical sequence is calculated and square rooted to obtain the standard deviation of the reactive power change rate. The mean vector, covariance matrix, and standard deviation of the reactive power change rate are pushed to the shared memory area of ​​the real-time processing module for real-time computation.

[0041] Step S102: Based on the Mahalanobis distance between the real-time three-dimensional feature vector and the mean vector, anomalies are determined, and initial fault feature quantities are calculated. The effective fault feature quantities are obtained by using the persistence confidence decay factor for verification.

[0042] Directly comparing electrical quantities in the original dimension cannot identify hidden anomalies, and fixed threshold logic cannot distinguish between parameter changes caused by fluctuations in photovoltaic output and parameter mutations caused by short circuits. In addition, extremely short-term transient disturbances such as lightning strikes or tree branches touching the wires, which are common in the power grid, can easily cause sudden false triggering. Based on this, this invention constructs an operating condition adaptive Mahalanobis distance anomaly deviation detection mechanism based on dimensionless parameters, and integrates the power mutation ratio describing the inverter's unique ride-through behavior and the first-order inertial cumulative verification for the time dimension, so as to achieve step-by-step calculation and differentiation of different types of fluctuations and short-circuit faults.

[0043] Specifically, the real-time processing module of the edge computing unit reads the pushed mean vector and covariance matrix from the shared memory area; combines the normalized positive-sequence current value, the normalized zero-sequence current value, and the normalized positive-sequence voltage value obtained at the current sampling time to form a real-time three-dimensional feature vector; and calculates the Mahalanobis distance of the current electrical state based on the real-time three-dimensional feature vector, the mean vector, and the covariance matrix.

[0044] The formula for calculating the Mahalanobis distance is as follows:

[0045]

[0046] In the formula, The Mahalanobis distance; It is a real-time 3D feature vector, and ; This is the normalized value of the positive sequence current; This is the normalized value of the zero-sequence current; This is the normalized value of the positive sequence voltage; This is the matrix transpose symbol; It is a mean vector, and ; This represents the historical average of the normalized value of the positive sequence current. The normalized value of the zero-sequence current is the historical average. This represents the historical average of the normalized positive sequence voltage. It is the covariance matrix; is the inverse of the covariance matrix.

[0047] The inverse of the covariance matrix performs decorrelation and scale normalization on the feature space, eliminating the multidimensional dimensional differences and linear correlations between the normalized values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage, thus incorporating the inherent correlations between the dimensions under normal operating conditions into the measurement benchmark.

[0048] Among them, if the photovoltaic slow ramp-up leads to the normalized value of the positive sequence current Increased and accompanied by positive sequence voltage normalization value A slight decrease, because the correlation change pattern has been changed by the mean vector. The inverse of the covariance matrix The recorded, calculated Mahalanobis distance is output. Maintain at a low level; if a short circuit occurs, the normalized value of the positive sequence current will be affected. Surge and positive sequence voltage normalization value A disproportionate and significant decrease disrupts the inherent multidimensional correlation, thus affecting the Mahalanobis distance. The increase was rapid.

[0049] Furthermore, the edge computing unit calculates the Mahalanobis distance from each sample point in the historical normal feature cluster to the mean vector and forms a distance distribution. The 99th percentile of this distance distribution is selected as the dynamic deviation threshold. The real-time processing module compares the calculated Mahalanobis distance of the current electrical state with the set dynamic deviation threshold. When the Mahalanobis distance of the current electrical state is greater than the dynamic deviation threshold, the real-time processing module determines that the current state is abnormal.

[0050] On a discrete time axis, the instantaneous active power change rate is obtained by dividing the difference between the instantaneous active power at the current moment and the historical moment of the previous cycle by the time interval. Similarly, the instantaneous reactive power change rate is obtained by dividing the difference between the instantaneous reactive power at the current moment and the historical moment of the previous cycle by the time interval. The maximum value between the absolute value of the instantaneous reactive power change rate and the standard deviation of the reactive power change rate is extracted. The ratio of the absolute value of the negative portion of the instantaneous active power change rate to this maximum value is then calculated to construct the power mutation ratio. The specific calculation formula is as follows:

[0051]

[0052] In the formula, Power mutation ratio; The instantaneous rate of change of active power; This represents the instantaneous rate of change of reactive power. This is the minimum value operator; This is the absolute value operator; This is the maximum value operator; This represents the standard deviation of the reactive power change rate.

[0053] By combining the normalized positive-sequence voltage value, a piecewise linear mapping operation is performed on the constructed power mutation ratio to obtain the initial fault characteristic quantity; the calculation formula for the initial fault characteristic quantity is as follows:

[0054]

[0055] In the formula, These are the initial fault characteristic quantities; Power mutation ratio; This is the normalized value of the positive sequence voltage; Voltage sag threshold; This is the lower limit of the valid values; This represents the upper limit of valid values.

[0056] The method for setting the effective lower limit is as follows: extract all historical capacitor switching event records from the historical feature database, calculate the ratio of the absolute value of the negative part of the instantaneous active power change rate to the absolute value of the instantaneous reactive power change rate for each historical capacitor switching event record, construct a capacitor switching ratio distribution set, and select the 95th percentile of the capacitor switching ratio distribution set as the effective lower limit; the method for setting the effective upper limit is as follows: extract all historical cloud shading event records from the historical feature database, calculate the ratio of the absolute value of the negative part of the instantaneous active power change rate to the absolute value of the instantaneous reactive power change rate for each historical cloud shading event record, construct a cloud shading ratio distribution set, and select the 5th percentile of the cloud shading ratio distribution set as the effective upper limit; the fault ride-through mode threshold of 0.9 volts specified in the current "Technical Requirements for Photovoltaic Power Generation Grid-Connected Inverters" (standard number GB / T 37408-2019) is used as the voltage drop threshold. .

[0057] The numerator term extracts the negative portion of the instantaneous active power change rate to capture the physical phenomenon of reduced active power output of the photovoltaic inverter during a grid short circuit. The denominator term introduces the standard deviation of the reactive power change rate as a lower limit to prevent the calculated ratio from diverging when the reactive power change rate approaches zero under normal operating conditions. The piecewise mapping function uses the positive sequence voltage normalization value to ensure it is below the voltage drop threshold. The judgment conditions confirm that voltage drop and fault ride-through behavior are triggered; the physical characteristics of the photovoltaic inverter's active power output being reduced when a short circuit occurs in the grid are captured in one direction, preventing the calculated ratio from diverging when the reactive power change rate approaches zero under normal operating conditions.

[0058] When cloud cover obstructs the photovoltaic panels, causing a sudden drop in active power, the numerator increases, but the inverter does not generate reactive power, resulting in a very small denominator. The calculated power surge ratio is... Greater than the upper limit of valid values The natural fluctuation is piecewise mapped to zero-valued initial fault characteristic quantities. When a short-circuit fault occurs, the instantaneous active power decreases and the instantaneous reactive power is injected, causing a power surge ratio. Falling into the lower limit of valid values With the upper limit of valid values Between, and simultaneously the normalized value of the positive sequence voltage voltage drop to voltage drop threshold The model then calculates and outputs a non-zero initial fault characteristic that is proportional to the drop depth. .

[0059] The edge computing unit extracts all historical transient disturbance event records from the historical feature database, calculates the duration of all historical transient disturbance event records, and obtains the maximum transient disturbance duration; it also extracts all historical short-circuit fault event records from the historical feature database, calculates the duration of all historical short-circuit fault event records before the protection device operates, and obtains the minimum fault tolerance duration; it calculates the average of the maximum transient disturbance duration and the minimum fault tolerance duration, and sets the average value as the confidence time constant.

[0060] Furthermore, the real-time processing module of the edge computing unit performs discrete-time cumulative integration on the duration during which the initial fault feature quantity is continuously greater than 0. Starting from the moment when the initial fault feature quantity first becomes greater than 0, the sampling time interval corresponding to each sampling moment when it is determined to be greater than 0 is accumulated. The obtained cumulative integration result and the set confidence time constant are substituted into the step response function of the first-order inertial element to calculate the persistence confidence decay factor asymptotically approaching 1 from 0. The initial fault feature quantity extracted at the current moment is multiplied by the calculated persistence confidence decay factor to obtain the effective fault feature quantity.

[0061] The formula for calculating the effective fault characteristic quantity is as follows:

[0062]

[0063] In the formula, Effective fault characteristic quantities; These are the initial fault characteristic quantities; It is a natural exponential function; The confidence level time constant; The current moment; The initial time when the initial fault characteristic quantity first becomes greater than 0; For integration variables; This is the time-dimension integration operator; For step function, when hour, ,otherwise, The time integral term is embedded in the negative exponential decay structure of the step response function of the first-order inertial element, which makes the confidence of the characteristic intensity related to the duration of the abnormal phenomenon, thus eliminating the system response lag caused by relying on external fixed delay to prevent false triggering.

[0064] Specifically, when the confidence time constant is 50 milliseconds, if a lightning flashover occurs, resulting in an initial fault characteristic quantity greater than zero within 5 milliseconds... Since the integration time of 5 milliseconds is much smaller than the confidence time constant of 50 milliseconds, negative exponent The value is close to 1, resulting in an extremely small calculated persistence confidence decay factor, which reduces the effective fault characteristic quantity of the final multiplied output. The value is significantly attenuated to near zero; if a short-circuit fault occurs and the abnormal state lasts for more than 100 milliseconds, the accumulation of the integration time causes the negative exponential term to decay to 0 and the persistence confidence decay factor to approach 1, thereby reducing the effective fault characteristic quantity. Approximately equal to the initial fault characteristic quantity And output the complete result.

[0065] Step S103: Trigger an on-site isolation decision based on the comparison result between the effective fault characteristic quantity and the confirmation threshold, and record and report the fault assessment data.

[0066] After multi-dimensional feature calculation and time-dimensional confidence accumulation verification, the output effective fault feature quantity has extracted stable short-circuit physical characteristics. In order to ensure the speed of fault isolation and realize the self-healing of the distribution network, this invention performs local judgment and action decision in the primary and secondary integrated ring network box, cuts off the fault section without relying on the intervention of the distribution master station, and simultaneously retains the judgment process data for global collaborative analysis by the distribution master station.

[0067] First, the edge computing unit extracts all historical short-circuit fault event records from the historical feature database, calculates the maximum value of the effective fault feature quantity corresponding to each historical short-circuit fault event record during the fault occurrence period, constructs a fault feature peak distribution set, selects the 5th percentile of the fault feature peak distribution set, and sets the 5th percentile as the confirmation threshold.

[0068] Furthermore, the real-time processing module of the edge computing unit continuously calculates and compares the output valid fault feature quantity with the set confirmation threshold: when the valid fault feature quantity is greater than the set confirmation threshold, the edge computing unit determines that the currently occurring abnormal over-limit event is a short-circuit fault; based on the specific sampling time that triggers the current valid fault feature quantity and the incoming and outgoing line interval channel sequence to which the valid fault feature quantity belongs, the incoming and outgoing line interval where the short-circuit fault section is located is determined; through the built-in hardware control loop, a tripping command is issued to the spring operating mechanism of the determined incoming and outgoing line interval to drive the corresponding circuit breaker to perform mechanical tripping action, thereby blocking the bidirectional fault current feed in within the set time window and completing the physical electrical isolation of the short-circuit fault section.

[0069] Furthermore, the management module of the edge computing unit extracts the analysis chain data of this short-circuit fault event from the shared memory area of ​​the real-time processing module. This analysis chain data includes the full-process calculation sequence of the Mahalanobis distance that triggered the anomaly, the initial fault characteristic quantity, and the effective fault characteristic quantity. The analysis chain data is then jointly serialized and packaged with the event timestamp of the current power grid time synchronization system, the located fault interval number, and the physical feedback switch action status to generate a time-stamped event record. The time-stamped event record is stored in the local non-volatile storage area. Through the communication module of the distribution terminal unit, the fault analysis result and the isolation execution status containing the time-stamped event record are reported to the distribution master station for dispatch personnel to perform subsequent network reconstruction and power restoration operations.

[0070] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A primary and secondary integrated ring network box for on-site analysis and isolation of distribution network faults, 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 performs the following steps: The instantaneous values ​​of three-phase current, zero-sequence current, and three-phase voltage are collected to obtain the normalized values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage. Based on time period labels, weather type labels, and photovoltaic power output interval labels, a historical normal feature cluster is constructed from the historical feature database, and the mean vector, covariance matrix, and standard deviation of reactive power change rate are obtained statistically. A real-time three-dimensional feature vector is constructed based on the normalized values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage, and its Mahalanobis distance to the mean vector is calculated. When the Mahalanobis distance is greater than the dynamic deviation threshold, the power mutation ratio is calculated based on the instantaneous rate of change of active power, the instantaneous rate of change of reactive power, and the standard deviation of the rate of change of reactive power, including: In the formula, The power mutation ratio is mentioned. The instantaneous rate of change of active power, The instantaneous reactive power change rate is... This is the minimum value operator. This is the absolute value operator. This is the maximum value operator. The standard deviation of the reactive power change rate is used as the initial fault characteristic quantity, which is obtained by combining the normalized value of the positive sequence voltage. This initial fault characteristic quantity is obtained through piecewise linear mapping. When the normalized value of the positive sequence voltage is less than the voltage drop threshold and the power mutation ratio is greater than or equal to the lower limit of the effective value and less than or equal to the upper limit of the effective value, the initial fault characteristic quantity is calculated as follows: When this condition is not met, the initial fault characteristic quantity is equal to 0, where, These are the initial fault characteristic quantities. The power mutation ratio, This is the normalized value of the positive sequence voltage. Voltage sag threshold; The persistence confidence decay factor is calculated by accumulating and integrating the initial fault features using the confidence time constant. The effective fault features are then obtained by multiplying the initial fault features by the persistence confidence decay factor. The formula for calculating the effective fault features is as follows: In the formula, For effective fault characteristic quantities, These are the initial fault characteristic quantities. It is a natural exponential function. The confidence level time constant, For the current moment, The initial time when the initial fault characteristic quantity first becomes greater than zero. For integration variables, This is the time-dimension integration operator. For step function, and in When greater than zero Equal to 1, under other conditions When the value equals 0, a tripping command is sent to the spring operating mechanism when the effective fault characteristic quantity is greater than the confirmation threshold.

2. The integrated primary and secondary ring network box for on-site fault diagnosis and isolation in distribution networks according to claim 1, characterized in that, Obtaining the normalized values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage includes: The instantaneous values ​​of three-phase current, zero-sequence current, and three-phase voltage are filtered using a sliding window mean filtering algorithm. Discrete sampling sequences within a complete power frequency cycle are extracted, and the effective values ​​of three-phase current and zero-sequence current are calculated respectively. Discrete Fourier transform is performed on the discrete sampling sequences of three-phase voltage instantaneous values ​​to extract the fundamental voltage phasor. The fundamental voltage phasor is transformed using the symmetrical component method to obtain the effective value of positive-sequence voltage. The effective value of positive-sequence current is extracted from the effective values ​​of three-phase current using the symmetrical component method. The effective values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage are divided by the set rated current reference value and rated voltage reference value respectively to obtain the normalized values ​​of positive-sequence current, zero-sequence current, and positive-sequence voltage.

3. The integrated primary and secondary ring network box for on-site fault diagnosis and isolation in distribution networks according to claim 1, characterized in that, The method for determining the time period label, weather type label, and photovoltaic power output range label is as follows: The day is divided into multiple fixed time periods according to the 24-hour system, and the fixed time period in which the current time information is located is determined as the time period label. According to the set meteorological mapping rules, the meteorological cloud map and precipitation data are mapped as weather type labels. The instantaneous active power is divided by the rated active power of the photovoltaic grid-connected system to obtain the real-time output percentage. According to the set threshold nodes, the total output range is divided into progressive intervals, and the interval in which the real-time output percentage is located is determined as the photovoltaic output interval label.

4. The integrated primary and secondary ring network box for on-site fault diagnosis and isolation in distribution networks according to claim 1, characterized in that, The formula for calculating the Mahalanobis distance is: ； In the formula, The Mahalanobis distance, For real-time 3D feature vectors and , This is the normalized value of the positive sequence current. This is the normalized value of the zero-sequence current. This is the normalized value of the positive sequence voltage. This is the matrix transpose symbol. It is a mean vector and , This represents the historical average of the normalized positive sequence current. The normalized value of the zero-sequence current is the historical average. The normalized historical average of the positive sequence voltage is given. Let covariance matrix be the variance matrix. is the inverse of the covariance matrix.

5. The integrated primary and secondary ring network box for on-site fault diagnosis and isolation in distribution networks according to claim 1, characterized in that, The method for setting the confidence level time constant is as follows: Extract all historical transient disturbance event records from the historical feature database to statistically obtain the maximum transient disturbance duration. Extract the duration of all historical short-circuit fault event records before the protection device operates to statistically obtain the minimum fault tolerance duration. Calculate the average of the maximum transient disturbance duration and the minimum fault tolerance duration as the confidence time constant.

6. The integrated primary and secondary ring network box for on-site fault diagnosis and isolation in distribution networks according to claim 1, characterized in that, The method for setting the confirmation threshold is as follows: Extract all historical short-circuit fault event records from the historical feature database, calculate the maximum value of the effective fault feature quantity corresponding to each historical short-circuit fault event record during the fault occurrence period to construct a fault feature peak distribution set, and select the 5th percentile of the fault feature peak distribution set as the confirmation threshold.

7. The integrated primary and secondary ring network box for on-site fault diagnosis and isolation in distribution networks according to claim 1, characterized in that, The steps also include: After a short-circuit fault is determined, all calculation sequences, event timestamps, and switch action status are packaged to generate a time-stamped event record and reported to the distribution master station.

Images