Fault identification and network efficiency improvement method for intelligent micro-grid

Abstract

The invention discloses a fault identification and network efficiency improvement method for an intelligent micro-grid, particularly relates to the technical field of cross-domain cooperative avoidance, and is used for solving a resonance frequency band problem caused by coupling of an existing SDN flow table updating strategy and micro-grid power fluctuation. The method comprises the following steps: acquiring micro-grid power fluctuation data and SDN flow table updating parameters in real time, and performing timestamp calibration; extracting a power fluctuation frequency domain characteristic and a flow table updating time domain correlation coefficient; fusing frequency domain integration and time domain convolution through a cross-domain coupling model to generate a dynamic resonance risk index; predicting a future fluctuation main frequency band based on the phase change and generating a constraint condition in combination with a historical avoidance frequency band; dynamically generating a frequency avoidance weight according to the real-time network load and the frequency band effectiveness; triggering the period adjustment of the asymmetric flow table to realize the frequency domain avoidance of the update frequency and the fluctuation main frequency band; resonance risk identification and active avoidance of power and communication cross-domain data fusion are realized, and the transient stability and the network resource utilization rate of the micro-grid are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of cross-domain collaborative avoidance technology, and more specifically, to a method for fault identification and network efficiency improvement for a smart microgrid. Background Art

[0002] As the scale of smart microgrids expands, software-defined networking (SDN) technology has been widely introduced to achieve dynamic scheduling and centralized control of communication resources. In existing technologies, SDN controllers usually use fixed cycles or simple adaptive algorithms to adjust flow table rules according to network traffic status or device load changes to optimize communication bandwidth allocation and transmission delay. For example, by real-time monitoring of packet loss rate or link utilization, the flow table update cycle can be dynamically shortened or extended to improve network resource utilization. This method has achieved certain results in general industrial networks, and its core logic is to focus on the status feedback of the communication network itself.

[0003] However, in the smart microgrid scenario, there will be a cross-domain coupling effect between the SDN flow table update strategy and the power fluctuation of the power system: when the power fluctuation frequency of the microgrid (such as the second-level fluctuation caused by the randomness of wind, solar and storage output) is close to or coincides with the flow table update cycle, the two form a resonant frequency band, causing the communication channel bandwidth to be periodically occupied, causing group delay jitter and timing inaccuracy of key control instructions (such as energy storage charging and discharging instructions, fault isolation signals). This hidden resonance effect not only aggravates network congestion, but may also cause asynchronous execution of multi-node control instructions, seriously threatening the transient stability of the microgrid. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method for fault identification and network performance improvement for a smart microgrid to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A method for fault identification and network performance improvement for a smart microgrid comprises the following steps: S1, real-time collection of microgrid power fluctuation data and software-defined network flow table update parameters, and timestamp calibration; S2. Perform short-time Fourier transform on the calibrated power fluctuation data to extract the power fluctuation frequency domain characteristics, and calculate the time domain correlation coefficient based on the periodic distribution of the flow table update parameters; S3, inputting the power fluctuation frequency domain characteristics and time domain correlation coefficient into the cross-domain coupling model, and generating a dynamic resonance risk index by multiplying the frequency domain amplitude integral and the time domain sliding window convolution; S4. When the dynamic resonance risk index exceeds the preset threshold, the future main frequency band of fluctuation is predicted based on the phase change of the power fluctuation frequency domain characteristics, and the constraint condition is generated by superimposing it with the historical avoidance frequency band; S5. Based on the constraints and the real-time network bandwidth occupancy rate, a sliding window is used to count the effectiveness of the historical avoidance frequency bands, and the frequency avoidance weights are generated in combination with the frequency band priority sorting; S6. Asymmetrically adjust the flow table update period according to the frequency avoidance weight so that the frequency domain distribution of the update frequency has no overlapping interval with the main frequency band of the fluctuation.

[0006] In a preferred embodiment, the power fluctuation data of the microgrid and the flow table update parameters of the software-defined network are collected in real time, and timestamp calibration is performed, including: The voltage fluctuation data at the inverter output end in the microgrid and the charge state change data of the energy storage system are collected in real time as power fluctuation data, and the update time interval of the flow table items of each node recorded in the software-defined network controller is collected as the flow table update parameter; Receive the timestamp of power fluctuation data from the inverter and energy storage system through the microgrid communication gateway, and synchronize and calibrate it with the system clock of the software-defined network controller; The calibrated power fluctuation data and flow table update parameters are stored in the cache queue according to a unified time base.

[0007] In a preferred embodiment, a short-time Fourier transform is performed on the calibrated power fluctuation data to extract the power fluctuation frequency domain characteristics, and a time domain correlation coefficient is calculated based on the periodic distribution of the flow table update parameters, including: The calibrated power fluctuation data is subjected to short-time Fourier transform using a predefined window function to extract the power fluctuation frequency domain features, where the power fluctuation frequency domain features include the amplitude spectral density within the selected frequency band; Based on the periodic distribution of flow table update parameters, the time domain correlation coefficient between the flow table update period and the power fluctuation time domain sequence is calculated; The power fluctuation frequency domain features and time domain correlation coefficients are aligned by timestamps to generate a joint feature dataset.

[0008] In a preferred embodiment, the amplitude spectral density within the selected frequency band is integrated, and the formula is: ;in, It represents the integral result of the amplitude spectrum density of the selected frequency band in the frequency domain characteristics of power fluctuation. It represents the short-time Fourier transform The complex amplitude corresponding to the frequency domain index, Indicates the minimum frequency domain index corresponding to the selected frequency band, Indicates the maximum frequency domain index corresponding to the selected frequency band, represents the frequency resolution, Represents the discrete frequency domain index after short-time Fourier transform.

[0009] In a preferred embodiment, the power fluctuation frequency domain characteristics and the time domain correlation coefficient are input into the cross-domain coupling model, and the dynamic resonance risk index is generated by the product operation of the frequency domain amplitude integral and the time domain sliding window convolution, including: Normalizing the integral results of the amplitude spectrum density of each selected frequency band in the power fluctuation frequency domain characteristics to obtain the normalized frequency domain amplitude weight; Perform sliding window convolution operation on the time domain correlation coefficient sequence, with the window width consistent with the statistical interval of the flow table update cycle, and output the time domain dynamic correlation strength; The normalized frequency domain amplitude weight and the time domain dynamic correlation strength are aligned with the timestamp and multiplied point by point to generate a dynamic resonance risk index; Among them, the convolution kernel coefficient of the sliding window convolution operation is dynamically adjusted according to the historical distribution of dynamic resonance risk indicators, and the convolution kernel length is positively correlated with the mean of the flow table update cycle.

[0010] In a preferred embodiment, when the dynamic resonance risk index exceeds a preset threshold, the future main frequency band of fluctuation is predicted based on the phase change of the power fluctuation frequency domain characteristics, and the constraint conditions are superimposed with the historical avoidance frequency band, including: When the dynamic resonance risk index exceeds the preset threshold, the target frequency band with the fastest phase change rate in the power fluctuation frequency domain characteristics is extracted, and the frequency offset of the target frequency band in the future time window is predicted by calculating the second-order difference of the phase change; The center frequency and bandwidth of the future fluctuation main frequency band are calculated based on the frequency offset. The center frequency is the algebraic sum of the center frequency of the current target frequency band and the frequency offset. The bandwidth is the linear combination of the bandwidth of the current target frequency band and the absolute value of the frequency offset. The linear combination coefficient is dynamically adjusted based on the effectiveness of the historical avoidance frequency band. The future fluctuation main frequency band and the pre-stored historical avoidance frequency band are superimposed in the frequency domain. If there is an overlapping area between the two, the upper and lower limits of the overlapping area are taken as the frequency band boundaries of the constraint conditions. If there is no overlap, the future fluctuation main frequency band and the historical avoidance frequency band are merged into a discrete constraint interval. The constraints are dynamically modified based on the current network load status. If the real-time bandwidth occupancy rate exceeds the load threshold, the bandwidth of the constrained frequency band is compressed year-on-year according to the bandwidth occupancy rate.

[0011] In a preferred embodiment, based on the constraint conditions and the real-time network bandwidth occupancy rate, a sliding window is used to count the effectiveness of the historical avoidance frequency bands, and a frequency avoidance weight is generated in combination with the frequency band priority sorting, including: Extract the avoidance success rate and average avoidance duration of each frequency band within a predefined time window from the pre-stored historical avoidance frequency bands as effectiveness indicators; According to the real-time network bandwidth occupancy rate and effectiveness indicators, the priority score is obtained by weighted calculation of each frequency band through the dynamic allocation of bandwidth occupancy rate weight coefficient and average avoidance time weight coefficient. The allocation ratio of bandwidth occupancy rate weight coefficient and time weight coefficient changes with the current network load status. The priority scores of each frequency band are arranged in descending order, and a preset number of high frequency bands are selected as candidate avoidance frequency bands; The priority scores of the candidate avoidance frequency bands are normalized to generate frequency avoidance weights.

[0012] In a preferred embodiment, the flow table update period is adjusted asymmetrically according to the frequency avoidance weight so that the frequency domain distribution of the update frequency has no overlapping interval with the main frequency band of the fluctuation, including: The base adjustment amount of the flow table update period is determined based on the frequency avoidance weight, and the base adjustment amount is nonlinearly negatively correlated with the frequency avoidance weight; The baseline adjustment amount is dynamically modified according to the real-time network bandwidth utilization rate. When the bandwidth utilization rate is higher than the preset load threshold, the adjustment amount is additionally reduced according to the excess utilization rate ratio. The software-defined network controller is triggered to perform asymmetric adjustment on the flow table update period of the target node. The adjustment direction is: the update period of the node corresponding to the candidate avoidance frequency band with high frequency avoidance weight is extended, and the update period of the node corresponding to the frequency band with low frequency avoidance weight is shortened; The frequency domain distribution histogram matching is used to verify whether the adjusted flow table update frequency has no overlapping interval with the main frequency band of the fluctuation. If there is residual overlap, the update cycle offset is adjusted secondary based on the bandwidth of the overlapping area.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention collects power fluctuations and flow table update parameters in real time, constructs a cross-domain coupling model, integrates power frequency domain characteristics with flow table time domain parameters to calculate dynamic resonance risk indicators, breaks through the limitations of traditional single-domain optimization, and realizes accurate identification of power-communication coupling risks; further predicts future main frequency bands of fluctuations based on phase changes, and generates dynamic constraints in combination with historical avoidance frequency bands, actively avoids resonance frequency bands through frequency domain weights and asymmetric period adjustments, and eliminates timing inaccuracies and network congestion caused by cross-domain coupling from the root.

[0014] 2. The present invention uses a sliding window to count the effectiveness of historical avoidance frequency bands, and dynamically generates frequency avoidance weights in combination with real-time network load status, to achieve a multi-objective balance of bandwidth occupancy, avoidance success rate, and timing accuracy; further, through an asymmetric flow table period adjustment mechanism, the update cycle is extended or shortened for frequency bands with different weights, so that communication resource scheduling closely matches the characteristics of power fluctuations, maximizing network utilization while avoiding channel preemption. In addition, the frequency domain distribution verification and secondary adjustment mechanism are introduced to ensure the global optimality of the avoidance strategy, significantly improving the reliability of microgrid operation and the execution synchronization rate of control instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flow chart of a method for fault identification and network performance improvement for a smart microgrid. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Example: Figure 1 The present invention provides a method for fault identification and network performance improvement for a smart microgrid, which comprises the following steps: S1, real-time collection of microgrid power fluctuation data and software-defined network flow table update parameters, and timestamp calibration; S2. Perform short-time Fourier transform on the calibrated power fluctuation data to extract the power fluctuation frequency domain characteristics, and calculate the time domain correlation coefficient based on the periodic distribution of the flow table update parameters; S3, inputting the power fluctuation frequency domain characteristics and time domain correlation coefficient into the cross-domain coupling model, and generating a dynamic resonance risk index by multiplying the frequency domain amplitude integral and the time domain sliding window convolution; S4. When the dynamic resonance risk index exceeds the preset threshold, the future main frequency band of fluctuation is predicted based on the phase change of the power fluctuation frequency domain characteristics, and the constraint condition is generated by superimposing it with the historical avoidance frequency band; S5. Based on the constraints and the real-time network bandwidth occupancy rate, a sliding window is used to count the effectiveness of the historical avoidance frequency bands, and the frequency avoidance weights are generated in combination with the frequency band priority sorting; S6. Asymmetrically adjust the flow table update period according to the frequency avoidance weight so that the frequency domain distribution of the update frequency has no overlapping interval with the main frequency band of the fluctuation.

[0018] S1. Real-time collection of microgrid power fluctuation data and software-defined network flow table update parameters, and timestamp calibration, including: The voltage fluctuation data at the output end of the inverter in the microgrid is collected in real time. For example, the voltage fluctuation data is obtained through a three-phase voltage sensor connected to the AC side of the inverter at a sampling frequency of 100 times per second. At the same time, the charge state change data of the energy storage system is collected. The charge state data is obtained through the built-in battery management system of the energy storage system at a sampling frequency of 10 times per second.

[0019] The flow table update parameters of the software-defined network controller are obtained by reading the update time interval of the flow table items of each node recorded in the controller. The time interval data includes but is not limited to the flow table item addition time, deletion time and modification time, and the time interval accuracy is millisecond level.

[0020] The timestamp calibration of power fluctuation data and flow table update parameters is performed in the following way: the timestamp of power fluctuation data from three-phase voltage sensors and battery management system is received through the microgrid communication gateway, and the timestamp format is Unix millisecond timestamp; at the same time, the system clock of the software-defined network controller is obtained, and the system clock is synchronized to the Beidou satellite time through the network time protocol; the timestamp of power fluctuation data is compared with the system clock of the software-defined network controller. If the time deviation exceeds 1 millisecond, the data is discarded and re-collection is triggered. If the time deviation is within 1 millisecond, the timestamp of power fluctuation data is corrected to the system clock time of the software-defined network controller.

[0021] The calibrated power fluctuation data and flow table update parameters are stored in the cache queue according to a unified time reference, and the cache queue adopts a first-in-first-out storage structure.

[0022] S2. Perform short-time Fourier transform on the calibrated power fluctuation data to extract the power fluctuation frequency domain characteristics, and calculate the time domain correlation coefficient based on the periodic distribution of the flow table update parameters, including: When performing short-time Fourier transform on the calibrated power fluctuation data, a predefined window function is used to perform windowing on the power fluctuation time domain series. The predefined window function includes but is not limited to a Hanning window or a rectangular window, and the window length is dynamically adjusted according to the power fluctuation characteristics. The extraction range of the power fluctuation frequency domain characteristics is a selected frequency band, which is determined based on the typical fluctuation frequency of the distributed power source in the microgrid. For example, for photovoltaic output fluctuations, the selected frequency band is 0.5Hz to 5Hz; for wind power generation fluctuations, the selected frequency band is 0.2Hz to 2Hz; for energy storage charging and discharging fluctuations, the selected frequency band is 0.1Hz to 1Hz. The calculation of frequency domain features includes integrating the amplitude spectral density within the selected frequency band, and the integral formula is: ;in, Represents the integral result of the amplitude spectrum density of the selected frequency band in the frequency domain characteristics of power fluctuation; It represents the short-time Fourier transform The complex amplitude corresponding to the frequency domain index is calculated by discrete Fourier transform; Indicates the minimum frequency domain index corresponding to the selected frequency band; Indicates the maximum frequency domain index corresponding to the selected frequency band; represents the frequency resolution, which is determined by the window length and the sampling interval; Represents the discrete frequency domain index after short-time Fourier transform, and the value range is an integer.

[0023] When calculating the time domain correlation coefficient based on the periodic distribution of the flow table update parameters, the time series of the flow table update period is first counted, and the statistical interval of the time series is consistent with the sampling interval of the power fluctuation data.

[0024] The time domain correlation coefficient is obtained by calculating the ratio of the covariance of the flow table update cycle sequence and the power fluctuation time domain sequence to the standard deviation of the two. The calculation formula is: ;in, Represents the time domain correlation coefficient between the flow table update period and the power fluctuation time domain series; Indicates the flow table update cycle sequence, which indicates the set of update time intervals of flow table entries of each node, in seconds; represents the power fluctuation time domain sequence, which represents the time domain sampling value set of power fluctuation data, in kilowatts; Representation sequence and The covariance of Indicates the standard deviation of the flow table update cycle sequence; Represents the standard deviation of the power fluctuation time domain series.

[0025] When aligning the power fluctuation frequency domain features and time domain correlation coefficients by timestamp to generate a joint feature data set, the alignment rule is: if there are multiple power fluctuation frequency domain feature values ​​at the same timestamp (such as data from multiple inverters), the average is taken as the feature value of the timestamp; if the timestamp of the flow table update cycle sequence does not completely match the timestamp of the power fluctuation data, linear interpolation is used to fill in the missing values. The storage format of the joint feature data set is a two-dimensional matrix, where the rows of the matrix correspond to the timestamps, and the columns correspond to the amplitude spectrum density integration results and time domain correlation coefficients of different frequency bands.

[0026] S3. Input the power fluctuation frequency domain characteristics and time domain correlation coefficient into the cross-domain coupling model, and generate the dynamic resonance risk index through the product operation of the frequency domain amplitude integral and the time domain sliding window convolution, including: When normalizing the amplitude spectral density integration results of each selected frequency band in the power fluctuation frequency domain characteristics, the normalization method is: for the integration results of each frequency band at each timestamp, respectively calculate the maximum and minimum values ​​in the historical data of the frequency band, and map the current integration value to the interval of 0 to 1 according to the formula "(current value - minimum value) / (maximum value - minimum value)" to obtain the normalized frequency domain amplitude weight.

[0027] For example, if the historical integral value range of a frequency band is 100 to 500 and the current integral value is 300, the normalized result is (300-100) / (500-100)=0.5.

[0028] For newly added frequency bands with insufficient historical data, normalized parameters of adjacent frequency bands are used for interpolation. Interpolation methods include but are not limited to linear interpolation or nearest neighbor interpolation.

[0029] When performing a sliding window convolution operation on the time domain correlation coefficient sequence, the window width is set to an integer multiple of the flow table update cycle statistical interval. For example, if the flow table update cycle statistical interval is 1 second, the window width can be set to 5 seconds or 10 seconds.

[0030] The convolution kernel coefficient is dynamically adjusted according to the distribution of historical resonance risk indicators. The specific method is: statistically calculate the distribution of time domain correlation coefficients corresponding to the moment when the dynamic resonance risk indicator in the historical data exceeds the preset threshold, and extract its mean and variance as the basis for adjusting the convolution kernel coefficient.

[0031] For example, if the time domain correlation coefficient corresponding to the historical high-risk period has a mean of 0.8 and a variance of 0.1, the convolution kernel coefficient is generated according to the Gaussian distribution, with a central coefficient of 0.8 and an edge coefficient attenuated by a variance of 0.1. The convolution kernel length is positively correlated with the mean of the flow table update cycle. For example, when the mean of the flow table update cycle is 2 seconds, the convolution kernel length is set to 5 sampling points; when the mean is 5 seconds, the length is set to 10 sampling points. The positive correlation ratio coefficient is determined by fitting historical data.

[0032] When the normalized frequency domain amplitude weight and the time domain dynamic correlation strength are aligned according to the timestamp and then multiplied point by point, the alignment rules include: if only a single data in the frequency domain weight or time domain strength exists at a certain timestamp, the timestamp data is discarded; if there are normalized frequency domain amplitude weights of multiple frequency bands at the same timestamp, the sum of the weights of each frequency band is taken as the frequency domain comprehensive weight of the timestamp.

[0033] The calculation method of point-by-point multiplication is: dynamic resonance risk index = normalized frequency domain amplitude weight × time domain dynamic correlation strength. For example, if the frequency domain comprehensive weight of a timestamp is 0.6 and the time domain dynamic correlation strength is 0.9, then the dynamic resonance risk index is 0.6×0.9=0.54. For data with timestamp deviation (such as the frequency domain data timestamp is t1, the time domain data timestamp is t2, and |t1-t2|≤10 milliseconds), time tolerance matching is allowed, and the two are treated as the same timestamp.

[0034] After the dynamic resonance risk indicator is generated, the specific implementation method of dynamically adjusting the convolution kernel coefficient according to the distribution of historical resonance risk indicators includes: regularly (such as every hour) counting the frequency and duration of the dynamic resonance risk indicator exceeding the preset threshold. If the frequency exceeds the historical average by 20% or the single duration exceeds the historical longest value, the convolution kernel coefficient update is triggered. The update logic is: extract the statistical characteristics (such as mean, variance, extreme value) of the time domain correlation coefficient sequence in the current period, regenerate the convolution kernel coefficient, and smoothly transition to the new coefficient through the sliding average algorithm to avoid sudden changes causing shocks. For example, if the original convolution kernel coefficient is [0.5, 0.3, 0.2] and the new coefficient is [0.6, 0.25, 0.15], it will be gradually replaced at a rate of 10% per day during the transition period.

[0035] The implementation method of the convolution kernel length being positively correlated with the mean of the flow table update cycle further includes: determining the convolution kernel length using a piecewise function according to the dynamic change of the mean of the flow table update cycle. For example, when the mean of the flow table update cycle is less than 3 seconds, the convolution kernel length = mean × 2.5; when the mean is between 3 seconds and 6 seconds, the length = mean × 2; when the mean is greater than 6 seconds, the length = mean × 1.5.

[0036] The proportional coefficient is determined by historical data regression analysis, for example, by using the least squares method to fit the relationship curve between the mean and the optimal convolution kernel length. If the mean of the flow table update cycle changes suddenly (such as the mean change exceeds 3 standard deviations of the historical fluctuation range), the convolution kernel length adjustment is temporarily frozen until the mean stabilizes and then dynamic adaptation is resumed.

[0037] S4. When the dynamic resonance risk index exceeds the preset threshold, the future main frequency band of fluctuation is predicted based on the phase change of the power fluctuation frequency domain characteristics, and the constraints are superimposed with the historical avoidance frequency band to generate constraints, including: When the dynamic resonance risk index exceeds the preset threshold, the target frequency band with the fastest phase change rate is first extracted from the power fluctuation frequency domain characteristics. The phase change rate is calculated as follows: a first-order difference is performed on the phase sequence of the target frequency band to obtain the instantaneous phase change, and then a second-order difference operation is performed on the first-order difference result. The calculation process of the second-order difference is as follows: the first-order difference value of the phase at the current moment is subtracted from the first-order difference value of the phase at the previous moment to obtain the acceleration of the phase change.

[0038] The frequency offset is calculated by multiplying the acceleration of the phase change by the frequency domain resolution and dividing by twice pi, where the frequency domain resolution is determined by the window length and sampling interval of the short-time Fourier transform.

[0039] When calculating the center frequency and bandwidth of the future fluctuation main frequency band based on the frequency offset, the center frequency is calculated by directly adding the center frequency of the current target frequency band to the frequency offset. The sign of the frequency offset is determined by the acceleration direction of the phase change. The frequency offset is positive when the acceleration is positive, and negative when the acceleration is negative. The bandwidth is calculated by multiplying the bandwidth of the current target frequency band and the absolute value of the frequency offset by the linear combination coefficient and then adding them. The value range of the linear combination coefficient is 0.1 to 0.5, and the specific value is dynamically adjusted according to the effectiveness of the historical avoidance frequency band.

[0040] The effectiveness of a historical avoidance frequency band is defined as the ratio of the number of successful avoidance attempts for the frequency band to the total number of avoidance attempts within a preset time window (e.g., 24 hours). For example, if a frequency band has been tried to be avoided 10 times and 8 times were successful within the last 24 hours, the effectiveness is 80%, and the linear combination coefficient is 0.2 (the coefficient is 0.2 when the effectiveness is ≥80%, 0.3 when 60%≤effectiveness<80%, and 0.5 when effectiveness<60%).

[0041] When the future fluctuation main frequency band and the pre-stored historical avoidance frequency band are superimposed in the frequency domain, the superposition rules include: if there is an overlapping area between the future fluctuation main frequency band and any historical avoidance frequency band (that is, the frequency domain ranges of the two have an intersection), the frequency band boundary of the constraint condition takes the minimum frequency lower limit and the maximum frequency upper limit of the overlapping area; if there is no overlap, the future fluctuation main frequency band and all historical avoidance frequency bands are merged into a discrete constraint interval.

[0042] For example, if the main frequency band of future fluctuations is 1.5 Hz to 3 Hz and the historical avoidance frequency band is 2 Hz to 4 Hz, then the overlapping area is 2 Hz to 3 Hz, and the constraint frequency band boundaries are 2 Hz (lower limit) and 3 Hz (upper limit); if the main frequency band of future fluctuations is 1 Hz to 2 Hz and the historical avoidance frequency band is 3 Hz to 5 Hz, then the constraints are the discrete intervals [1 Hz, 2 Hz] and [3 Hz, 5 Hz].

[0043] When dynamically modifying the constraints based on the current network load status, if the real-time bandwidth occupancy rate exceeds the load threshold (for example, 80%), the bandwidth of the constrained frequency band is compressed year-on-year according to the bandwidth occupancy rate. The compression logic is: the original bandwidth multiplied by the compression factor, and the compression factor is equal to 1 minus (real-time bandwidth occupancy rate minus the load threshold) divided by (100 minus the load threshold). For example, when the original bandwidth is 2 Hz, the real-time bandwidth occupancy rate is 90%, and the load threshold is 80%, the compression factor is 1 - (90-80) / (100-80) = 0.5, and the compressed bandwidth is 2 Hz × 0.5 = 1 Hz.

[0044] It is worth noting that the preset threshold corresponding to the dynamic resonance risk index is determined jointly through historical operation data statistics and experimental calibration: the initial threshold range is generated after eliminating outliers based on the distribution characteristics of the indicator values ​​in the historical data, and then the effectiveness of the threshold is verified by artificially injecting typical power disturbance scenarios; the threshold is dynamically adjusted according to the effectiveness of the historical avoidance frequency band during operation. When the effectiveness is high, the threshold is proportionally increased to reduce the false alarm rate, and when the effectiveness is low, the threshold is proportionally reduced to enhance the sensitivity; at the same time, the threshold is regularly calibrated in combination with the microgrid operation cycle to ensure that it is adapted to the frequency band fluctuation characteristics of different seasons or load scenarios.

[0045] S5. Based on the constraints and real-time network bandwidth occupancy, a sliding window is used to count the effectiveness of historical avoidance frequency bands, and frequency avoidance weights are generated in combination with frequency band priority sorting, including: When extracting the avoidance success rate and average avoidance duration of each frequency band in a predefined time window from the pre-stored historical avoidance frequency bands as effectiveness indicators, the predefined time window is set to the latest 24 hours, and the avoidance success rate is calculated as follows: the number of times a frequency band is successfully avoided in the time window accounts for the percentage of the total number of times the frequency band is attempted to avoid; the average avoidance duration is the average duration of all successful avoidance events in the frequency band. For example, if a frequency band is attempted to avoid 20 times within 24 hours and succeeds 15 times, the avoidance success rate is 75%; the duration of the 15 successful avoidances is 5 seconds, 8 seconds, 6 seconds, etc., respectively, and the average avoidance duration is (5+8+6+...) / 15. If there are multiple time windows in the historical data (such as hourly segment storage), the overall effectiveness index is calculated by time weighting, and the more recent the data, the higher the weight.

[0046] According to the real-time network bandwidth occupancy and effectiveness indicators, when weighted calculation is performed on each frequency band to obtain a priority score through dynamic allocation of the bandwidth occupancy weight coefficient and the average avoidance time weight coefficient, the dynamic allocation rule is: if the real-time bandwidth occupancy is higher than the preset load threshold (for example, 80%), the bandwidth occupancy weight coefficient is set to 0.7, and the average avoidance time weight coefficient is set to 0.3; if it is lower than the load threshold, the bandwidth occupancy weight coefficient is set to 0.3, and the average avoidance time weight coefficient is set to 0.7.

[0047] After the avoidance success rate and average avoidance duration are dimensionless, the priority score is calculated as: ;in, represents the priority score, To avoid the success rate, is the average avoidance time, is the bandwidth occupancy weight coefficient, is the average avoidance time weight coefficient.

[0048] For example, the avoidance success rate of a certain frequency band is 80%, the average avoidance time is 6 seconds, and when the real-time bandwidth occupancy rate is 85% (higher than the threshold), , , then the priority score = 80×0.7 + 6×0.3=56+1.8=57.8.

[0049] Among them, the bandwidth occupancy weight coefficient and the average avoidance duration weight coefficient are set based on the real-time network load status: when the real-time bandwidth occupancy exceeds the preset load threshold, the bandwidth occupancy weight coefficient is allocated as the primary proportion, and the average avoidance duration weight coefficient is the secondary proportion, so as to give priority to the efficient release of bandwidth resources; when the bandwidth occupancy is lower than the threshold, the two proportions are reversely allocated, the average avoidance duration weight coefficient is increased to the primary proportion, and the bandwidth occupancy weight coefficient is reduced to the secondary proportion, so as to optimize the timing accuracy of the control instructions. The sum of the two types of weight coefficients remains constant to ensure the coordination of the calculation ratio of the priority score, and at the same time adapt to the core needs of different load scenarios through dynamic allocation.

[0050] It is worth noting that multiplying the avoidance success rate by the bandwidth occupancy weight coefficient is in line with the logic of dynamic dispatch of microgrids: when the network bandwidth occupancy rate is high (such as > 80%), the calculation weight of the avoidance success rate is increased, and the frequency band with good historical avoidance effect is given priority to quickly release bandwidth resources; when the bandwidth occupancy rate is low, its weight is reduced to take into account other factors (such as avoidance duration). This design dynamically adjusts the influence of historical data through real-time network status, realizes the linkage adaptation of network load and avoidance strategy, solves the problem of over-avoidance or under-avoidance caused by fixed weights in traditional methods, and ensures resource scheduling efficiency and stability.

[0051] The priority scores of each frequency band are arranged in descending order. When selecting the preset number of high frequency bands as candidate avoidance frequency bands, the preset number is dynamically adjusted according to the total number of historical avoidance frequency bands. The adjustment rule is: if the total number of historical avoidance frequency bands is less than or equal to 10, the preset number = total number × 50%; if it is greater than 10, the preset number = 5 + (total number - 10) × 20%. For example, if the total number of historical avoidance frequency bands is 8, the preset number = 8 × 50% = 4; if the total number is 15, the preset number = 5 + (15-10) × 20% = 6.

[0052] When the priority scores of the candidate avoidance frequency bands are normalized to generate frequency avoidance weights, the normalization method is: divide the priority score of each candidate frequency band by the sum of the priority scores of all candidate frequency bands so that the sum of all weights is 1. For example, the priority scores of the three candidate frequency bands are 60, 40, and 20, respectively. The normalized weights are 60 / (60+40+20)=0.5, 40 / 120≈0.33, and 20 / 120≈0.17.

[0053] S6. Asymmetrically adjust the flow table update period according to the frequency avoidance weight so that the frequency domain distribution of the update frequency has no overlapping interval with the main frequency band of the fluctuation, including: When determining the baseline adjustment amount of the flow table update period based on the frequency avoidance weight, the nonlinear negative correlation between the baseline adjustment amount and the frequency avoidance weight is realized through an exponential function. The specific rule is: the higher the frequency avoidance weight, the greater the reduction in the baseline adjustment amount, but the reduction rate gradually slows down as the weight increases. For example, when the frequency avoidance weight is 0.8, the baseline adjustment amount is reduced to 30% of the original period; when the weight is 0.5, it is reduced to 50%; when the weight is 0.2, it is reduced to 80%. The parameters of the nonlinear relationship are determined by fitting historical data to ensure that the avoidance priority of high-weight frequency bands is significantly higher than that of low-weight frequency bands.

[0054] When dynamically correcting the baseline adjustment amount based on the real-time network bandwidth usage, if the bandwidth usage is higher than the preset load threshold (for example, 80%), the adjustment amount is additionally reduced according to the usage excess ratio. The excess ratio is calculated as: (real-time bandwidth usage - load threshold) / (100 - load threshold). For example, if the preset load threshold is 80% and the real-time bandwidth usage is 90%, the excess ratio is (90-80) / (100-80) = 0.5, and the additional reduction amount = baseline adjustment amount × 0.5. If the bandwidth usage is lower than the threshold, the baseline adjustment amount remains unchanged.

[0055] When the software-defined network controller is triggered to perform asymmetric adjustment on the flow table update period of the target node, the specific rules for the adjustment direction are as follows: for the network nodes corresponding to the candidate avoidance frequency bands with high frequency avoidance weights, their flow table update period is extended by the adjustment amount; for the nodes corresponding to the frequency bands with low weights, the period is shortened by the adjustment amount. For example, if the frequency avoidance weight of a node is 0.8 and the baseline adjustment amount is a 30% reduction, and the bandwidth occupancy rate exceeds the limit ratio by 0.5, then the final adjustment amount = 30% × (1 + 0.5) = 45%, and the update period is extended from the original 1 second to 1 second / (1% ~ 45%) = 1.82 seconds; the weight of another node is 0.2 and the adjustment amount is a 20% reduction, without additional reduction, then the period is shortened from 1 second to 1 second × (1% ~ 20%) = 0.8 seconds.

[0056] When verifying whether the adjusted flow table update frequency has no overlapping interval with the main frequency band of fluctuation through frequency domain distribution histogram matching, the frequency domain distribution histogram is generated in the following way: the adjusted flow table update frequency of all nodes is counted, the frequency band is divided according to the preset resolution (for example, 0.1Hz), and the frequency of occurrence of each frequency band is calculated; the histogram is compared with the range of the main frequency band of fluctuation, and if there is frequency band overlap, it is determined to be residual overlap. For example, if the main frequency band of fluctuation is 1.5Hz to 2.5Hz, and the adjusted flow table update frequency is distributed between 2Hz and 3Hz, then the overlapping area is 2Hz to 2.5Hz.

[0057] If there is residual overlap, the update cycle offset is adjusted again based on the overlap area bandwidth. The adjustment rule is: according to the ratio of the overlap area bandwidth to the total bandwidth of the main frequency band, the update cycle is further reduced or extended. For example, if the overlap area bandwidth is 0.5Hz and the total bandwidth of the main frequency band is 1Hz, the adjustment amount = original adjustment amount × (1 + 0.5 / 1) = 1.5 times the original adjustment amount, and the update cycle is adjusted by an additional 50%.

[0058] The above formulas are all dimensionless and numerical calculations. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0059] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or may be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0060] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0061] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0062] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0063] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0064] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0065] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0066] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for fault identification and network performance improvement for a smart microgrid, characterized in that: The steps include: S1, real-time collection of microgrid power fluctuation data and software-defined network flow table update parameters, and timestamp calibration; S2. Perform short-time Fourier transform on the calibrated power fluctuation data to extract the power fluctuation frequency domain characteristics, and calculate the time domain correlation coefficient based on the periodic distribution of the flow table update parameters; S3, inputting the power fluctuation frequency domain characteristics and time domain correlation coefficient into the cross-domain coupling model, and generating a dynamic resonance risk index by multiplying the frequency domain amplitude integral and the time domain sliding window convolution; S4. When the dynamic resonance risk index exceeds the preset threshold, the future main frequency band of fluctuation is predicted based on the phase change of the power fluctuation frequency domain characteristics, and the constraint condition is generated by superimposing it with the historical avoidance frequency band; S5. Based on the constraints and the real-time network bandwidth occupancy rate, a sliding window is used to count the effectiveness of the historical avoidance frequency bands, and the frequency avoidance weights are generated in combination with the frequency band priority sorting; S6. Asymmetrically adjust the flow table update period according to the frequency avoidance weight so that the frequency domain distribution of the update frequency has no overlapping interval with the main frequency band of the fluctuation.

2. A method for fault identification and network performance improvement for a smart microgrid according to claim 1, characterized in that: Real-time collection of microgrid power fluctuation data and software-defined network flow table update parameters, and timestamp calibration, including: The voltage fluctuation data at the inverter output end in the microgrid and the charge state change data of the energy storage system are collected in real time as power fluctuation data, and the update time interval of the flow table items of each node recorded in the software-defined network controller is collected as the flow table update parameter; Receive the timestamp of power fluctuation data from the inverter and energy storage system through the microgrid communication gateway, and synchronize and calibrate it with the system clock of the software-defined network controller; The calibrated power fluctuation data and flow table update parameters are stored in the cache queue according to a unified time base.

3. The method for fault identification and network performance improvement for a smart microgrid according to claim 1, characterized in that: Perform short-time Fourier transform on the calibrated power fluctuation data to extract the power fluctuation frequency domain features, and calculate the time domain correlation coefficient based on the periodic distribution of the flow table update parameters, including: The calibrated power fluctuation data is subjected to short-time Fourier transform using a predefined window function to extract the power fluctuation frequency domain features, where the power fluctuation frequency domain features include the amplitude spectral density within the selected frequency band; Based on the periodic distribution of flow table update parameters, the time domain correlation coefficient between the flow table update period and the power fluctuation time domain sequence is calculated; The power fluctuation frequency domain features and time domain correlation coefficients are aligned by timestamps to generate a joint feature dataset.

4. A method for fault identification and network performance improvement for a smart microgrid according to claim 3, characterized in that: The amplitude spectral density within the selected frequency band is integrated, and the formula is: ;in, It represents the integral result of the amplitude spectrum density of the selected frequency band in the frequency domain characteristics of power fluctuation. It represents the short-time Fourier transform The complex amplitude corresponding to the frequency domain index, Indicates the minimum frequency domain index corresponding to the selected frequency band, Indicates the maximum frequency domain index corresponding to the selected frequency band, represents the frequency resolution, Represents the discrete frequency domain index after short-time Fourier transform.

5. The method for fault identification and network performance improvement for a smart microgrid according to claim 1, characterized in that: The power fluctuation frequency domain characteristics and time domain correlation coefficients are input into the cross-domain coupling model, and the dynamic resonance risk index is generated by the product operation of the frequency domain amplitude integral and the time domain sliding window convolution, including: Normalizing the integral results of the amplitude spectrum density of each selected frequency band in the power fluctuation frequency domain characteristics to obtain the normalized frequency domain amplitude weight; Perform sliding window convolution operation on the time domain correlation coefficient sequence, with the window width consistent with the statistical interval of the flow table update cycle, and output the time domain dynamic correlation strength; The normalized frequency domain amplitude weight and the time domain dynamic correlation strength are aligned with the timestamp and multiplied point by point to generate a dynamic resonance risk index; Among them, the convolution kernel coefficient of the sliding window convolution operation is dynamically adjusted according to the historical distribution of dynamic resonance risk indicators, and the convolution kernel length is positively correlated with the mean of the flow table update cycle.

6. A method for fault identification and network performance improvement for a smart microgrid according to claim 1, characterized in that: When the dynamic resonance risk index exceeds the preset threshold, the future main frequency band of fluctuation is predicted based on the phase change of the power fluctuation frequency domain characteristics, and the constraints are superimposed with the historical avoidance frequency band to generate the following conditions: When the dynamic resonance risk index exceeds the preset threshold, the target frequency band with the fastest phase change rate in the power fluctuation frequency domain characteristics is extracted, and the frequency offset of the target frequency band in the future time window is predicted by calculating the second-order difference of the phase change; The center frequency and bandwidth of the future fluctuation main frequency band are calculated based on the frequency offset. The center frequency is the algebraic sum of the center frequency of the current target frequency band and the frequency offset. The bandwidth is the linear combination of the bandwidth of the current target frequency band and the absolute value of the frequency offset. The linear combination coefficient is dynamically adjusted based on the effectiveness of the historical avoidance frequency band. The future fluctuation main frequency band and the pre-stored historical avoidance frequency band are superimposed in the frequency domain. If there is an overlapping area between the two, the upper and lower limits of the overlapping area are taken as the frequency band boundaries of the constraint conditions. If there is no overlap, the future fluctuation main frequency band and the historical avoidance frequency band are merged into a discrete constraint interval. The constraints are dynamically modified based on the current network load status. If the real-time bandwidth occupancy rate exceeds the load threshold, the bandwidth of the constrained frequency band is compressed year-on-year according to the bandwidth occupancy rate.

7. A method for fault identification and network performance improvement for a smart microgrid according to claim 1, characterized in that: Based on the constraints and real-time network bandwidth occupancy, a sliding window is used to count the effectiveness of historical avoidance frequency bands, and frequency avoidance weights are generated in combination with frequency band priority sorting, including: Extract the avoidance success rate and average avoidance duration of each frequency band within a predefined time window from the pre-stored historical avoidance frequency bands as effectiveness indicators; According to the real-time network bandwidth occupancy rate and effectiveness indicators, the priority score is obtained by weighted calculation of each frequency band through the dynamic allocation of bandwidth occupancy rate weight coefficient and average avoidance time weight coefficient. The allocation ratio of bandwidth occupancy rate weight coefficient and time weight coefficient changes with the current network load status. The priority scores of each frequency band are arranged in descending order, and a preset number of high frequency bands are selected as candidate avoidance frequency bands; The priority scores of the candidate avoidance frequency bands are normalized to generate frequency avoidance weights.

8. The method for fault identification and network performance improvement for a smart microgrid according to claim 1, characterized in that: The flow table update period is adjusted asymmetrically according to the frequency avoidance weight so that the frequency domain distribution of the update frequency has no overlap with the main frequency band of the fluctuation, including: The base adjustment amount of the flow table update period is determined based on the frequency avoidance weight, and the base adjustment amount is nonlinearly negatively correlated with the frequency avoidance weight; The baseline adjustment amount is dynamically modified according to the real-time network bandwidth utilization rate. When the bandwidth utilization rate is higher than the preset load threshold, the adjustment amount is additionally reduced according to the excess utilization rate ratio. The software-defined network controller is triggered to perform asymmetric adjustment on the flow table update period of the target node. The adjustment direction is: the update period of the node corresponding to the candidate avoidance frequency band with high frequency avoidance weight is extended, and the update period of the node corresponding to the frequency band with low frequency avoidance weight is shortened; The frequency domain distribution histogram matching is used to verify whether the adjusted flow table update frequency has no overlapping interval with the main frequency band of the fluctuation. If there is residual overlap, the update cycle offset is adjusted secondary based on the bandwidth of the overlapping area.

Images