Low-voltage flexible direct-current power quality optimization method and system based on big data

Abstract

The application relates to a low-voltage flexible DC power quality optimization method and system based on big data, and relates to the technical field of power control, and comprises the following steps: acquiring a DC bus voltage fluctuation sequence and an output current distortion rate sequence; calculating a light-load oscillation risk coefficient according to the distortion rate and the load rate, and calculating a heavy-load voltage drop depth according to the integral of the deviation of the voltage from the rated voltage; when the light-load risk exceeds a first threshold value, extracting a damping ratio decay trajectory of the same frequency from historical big data, and calculating a dynamic compensation coefficient; when the heavy-load drop exceeds a second threshold value, matching a recovery time constant according to the ratio of the current change rate to the voltage recovery speed, and generating a correction amount; nonlinearly fusing the compensation coefficient and the correction amount to obtain an adaptive virtual impedance, replacing a fixed impedance value and issuing the adaptive virtual impedance to a pulse width modulator. The application solves the technical problems that the traditional method cannot simultaneously consider light-load oscillation and heavy-load drop, cannot adapt to working condition changes in real time, and is prone to causing system instability.

Description

Technical Field

[0001] This application relates to the field of power control technology, and in particular to a method and system for optimizing the quality of low-voltage flexible DC power based on big data. Background Technology

[0002] Low-voltage flexible DC distribution technology is a crucial supporting technology for the integration of distributed photovoltaic, energy storage systems, DC charging piles, and other new energy sources into the power distribution network, and it is widely used in industrial parks, residential communities, and data centers. Flexible DC systems achieve voltage transformation and power control through converters, which operate under two typical conditions: light load and heavy load, depending on load changes. Under light load, the converter output current is small and prone to distortion, and coupling with the system impedance may cause high-frequency oscillations. Under heavy load, the current increases sharply, potentially causing a momentary drop in DC bus voltage, which can trigger protection shutdown in severe cases. Existing control methods typically employ a fixed impedance droop control strategy, using a preset set of control parameters to handle all operating conditions. However, fixed parameters cannot simultaneously address both oscillation suppression under light load and voltage support requirements under heavy load.

[0003] Existing fixed impedance droop control methods have the following shortcomings: First, under light load conditions, the fixed virtual impedance is difficult to adaptively match the oscillation frequency, resulting in limited suppression effects; in severe cases, continuous system oscillation can affect equipment lifespan. Second, when heavy load drops occur, the fixed impedance cannot be dynamically adjusted, leading to slow voltage recovery and potential relay protection malfunctions. Third, there is a lack of operating condition identification and historical data utilization mechanisms; converter parameters vary significantly across different distribution areas, and uniform fixed parameters cannot adapt to individual characteristics. Therefore, there is an urgent need for an intelligent power quality optimization method capable of real-time identification of light and heavy load conditions and dynamic adjustment of control parameters. Summary of the Invention

[0004] This application provides a method and system for optimizing the power quality of low-voltage flexible DC power based on big data. It improves the technical problem in the prior art that fixed impedance droop control is difficult to suppress light load oscillations and compensate for heavy load voltage drops at the same time, which leads to the deterioration of power quality and the reduction of equipment operation stability under the two contradictory operating conditions.

[0005] This application discloses the following technical solution: In a first aspect, this application provides a method for optimizing the quality of low-voltage flexible DC power based on big data, the method comprising: Obtain the DC bus voltage fluctuation sequence and output current distortion rate sequence of each transformer area converter within a continuous time window; The light-load oscillation risk coefficient is calculated based on the output current distortion rate sequence and the load rate sequence, and the heavy-load voltage drop depth is calculated based on the integral value of the deviation between the DC bus voltage fluctuation sequence and the rated voltage. When the light load oscillation risk coefficient exceeds the first threshold, the damping ratio attenuation trajectory under the same oscillation frequency mode is extracted from historical big data, and the dynamic compensation coefficient of the virtual impedance is calculated based on the damping ratio attenuation trajectory. When the heavy load voltage drop depth exceeds the second threshold, the recovery time constant under similar drop conditions is matched from the historical big data based on the ratio of the current change rate to the voltage recovery speed at the start of the drop, and the correction amount of the voltage support coefficient is generated. The dynamic compensation coefficient and the correction amount are fused nonlinearly to obtain the adaptive virtual impedance adjustment value; The fixed impedance value in the droop control parameters of each transformer area converter is replaced with the adaptive virtual impedance adjustment value to generate an optimized control command and send it to the pulse width modulator of each transformer area converter.

[0006] Secondly, this application provides a low-voltage flexible DC power quality optimization system based on big data, the system comprising: The data acquisition module is deployed on the DC and AC sides of the converters in each distribution area to acquire the DC bus voltage fluctuation sequence and the output current distortion rate sequence within a continuous time window. The risk analysis module is used to calculate the light load oscillation risk coefficient based on the output current distortion rate sequence and the load rate sequence, and to calculate the heavy load voltage drop depth based on the integral value of the deviation between the DC bus voltage fluctuation sequence and the rated voltage. The compensation calculation module is used to extract the damping ratio decay trajectory under the same oscillation frequency mode from historical big data when the light load oscillation risk coefficient exceeds the first threshold, and calculate the dynamic compensation coefficient of the virtual impedance based on the damping ratio decay trajectory; and when the heavy load voltage drop depth exceeds the second threshold, it matches the recovery time constant under similar drop modes from the historical big data according to the ratio of the current change rate to the voltage recovery speed at the start of the drop, and generates the correction amount of the voltage support coefficient. A nonlinear fusion module is used to obtain an adaptive virtual impedance adjustment value by nonlinearly fusing the dynamic compensation coefficient and the correction amount. The instruction issuing module is used to replace the fixed impedance value in the droop control parameters of each transformer area converter with the adaptive virtual impedance adjustment value, generate optimized control instructions, and issue them to the pulse width modulator of each transformer area converter.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: The technical solution of this application provides a low-voltage flexible DC power quality optimization method based on big data. First, by acquiring the DC bus voltage fluctuation sequence and output current distortion rate sequence of each transformer area converter within a continuous time window, it realizes real-time and continuous monitoring of voltage dynamic changes and current distortion degree. This provides a synchronous and complete time-series data foundation for subsequent dual-condition risk identification, and solves the problems that traditional single-point acquisition cannot reflect voltage fluctuation trajectory and lacks quantitative calculation of distortion rate.

[0008] Furthermore, by calculating the risk coefficient of light-load oscillation based on the output current distortion rate sequence and the load rate sequence, and calculating the depth of heavy-load voltage drop based on the integral value of the deviation between the DC bus voltage fluctuation sequence and the rated voltage, independent quantitative assessment of two typical operating conditions, light-load oscillation and heavy-load voltage drop, is achieved. This solves the problems that traditional methods cannot distinguish operating conditions and that a single indicator cannot simultaneously warn of two fault modes.

[0009] Furthermore, when the risk coefficient of light-load oscillation exceeds the first threshold, the damping ratio attenuation trajectory of the same oscillation frequency mode is extracted from historical big data and the dynamic compensation coefficient of virtual impedance is calculated, realizing adaptive damping compensation based on historical similar oscillation events, and solving the problem that fixed parameters are difficult to match different oscillation frequencies and the suppression effect is limited.

[0010] Furthermore, when the voltage drop depth exceeds the second threshold, the recovery time constant under similar drop conditions is matched from historical big data based on the ratio of the current change rate to the voltage recovery speed at the start of the drop, and a correction value is generated. This realizes personalized learning and compensation of voltage drop recovery characteristics, and solves the problems of traditional methods being unable to dynamically adjust impedance and slow voltage recovery.

[0011] Finally, by nonlinearly fusing the dynamic compensation coefficient and the correction amount to obtain the adaptive virtual impedance adjustment value, the fixed impedance value in the droop control parameters is replaced and sent to the pulse width modulator. This achieves the synergistic optimization of light load oscillation suppression and heavy load voltage support, solves the technical problem that the fixed impedance cannot take into account the two contradictory operating conditions and the system stability is poor, and significantly improves the power quality and operational reliability of the low-voltage flexible DC system in the full load range.

[0012] In summary, the technical solution of this application realizes real-time assessment, adaptive compensation, and collaborative optimization of light-load oscillation and heavy-load voltage drop in low-voltage flexible DC systems. Through multi-source data acquisition, dual-condition risk quantification, historical big data matching, nonlinear fusion, and adaptive virtual impedance adjustment, it effectively improves the technical problems in the prior art, such as the difficulty in balancing the two contradictory conditions with fixed droop control parameters, the tendency for high-frequency oscillation under light load, the slow recovery of voltage drop under heavy load, and the poor system stability. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating the low-voltage flexible DC power quality optimization method based on big data provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a low-voltage flexible DC power quality optimization system based on big data, provided in an embodiment of this application.

[0015] The components represented by each number in the attached diagram are described as follows: data acquisition module 11, risk analysis module 12, compensation calculation module 13, nonlinear fusion module 14, and instruction issuance module 15. Detailed Implementation

[0016] This application provides a low-voltage flexible DC power quality optimization method and system based on big data, which is used to solve the technical problems that existing technologies cannot suppress light load oscillations and compensate for heavy load voltage drops in real time and adaptively, resulting in system instability, equipment life loss, and relay protection malfunction.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. It should be noted that the numerical values ​​in the embodiments are for illustrative purposes only and do not constitute a limitation on this application.

[0018] Example 1, as shown in the appendix Figure 1 As shown, this application provides a low-voltage flexible DC power quality optimization method based on big data, the method comprising the following steps: S100: Obtain the DC bus voltage fluctuation sequence and output current distortion rate sequence of each transformer area converter within a continuous time window.

[0019] In this embodiment of the application, in a scenario where the converters in each area of ​​a low-voltage flexible DC system need to monitor the DC bus voltage and AC side current distortion rate in real time to assess power quality, in order to obtain the voltage fluctuation sequence and current distortion rate sequence within a continuous time window and provide a data basis for subsequent light-load oscillation risk identification and heavy-load voltage drop analysis, it is necessary to synchronously collect voltage and current data with a fixed sampling period by using parallel voltage Hall sensors and series current transformers. After RMS calculation, distortion rate calculation, and time series arrangement, the data can solve the technical problems of traditional single-point acquisition failing to reflect the dynamic characteristics of voltage and current evolution over time, lacking quantitative calculation of distortion rate, and resulting in a lack of reliable time-series data input for oscillation and voltage drop detection.

[0020] Step S100 in the method provided in this application embodiment includes: Voltage Hall sensors connected in parallel between the positive and negative terminals of the DC side of each transformer converter are used to continuously collect the instantaneous DC bus voltage value at a fixed sampling period. The instantaneous voltage values ​​collected at each sampling time are arranged in chronological order to generate the DC bus voltage fluctuation sequence. Current transformers connected in series at the AC side output terminals of each transformer converter are used to synchronously collect the instantaneous three-phase output current values ​​within each sampling period. For each sampling period, the root mean square value of the instantaneous output current values ​​of phases A, B, and C are calculated as the effective value of each phase current, and the effective value of the fundamental component in the waveform of each phase current is extracted. Based on the effective value of each phase current and the effective value of the corresponding phase fundamental component, the instantaneous distortion rate of each phase is calculated, and the arithmetic mean of the instantaneous distortion rates of each phase is taken as the output current distortion rate of the current sampling period. The output current distortion rates calculated for all sampling periods within the continuous time window are arranged in chronological order to generate the output current distortion rate sequence. Detailed explanation is as follows: In this embodiment, the voltage Hall sensor refers to a non-contact sensor that measures voltage based on the Hall effect. It outputs a voltage signal proportional to the voltage by sensing the electric field between the positive and negative poles of the DC bus, used for real-time acquisition of instantaneous voltage values. The current transformer refers to a measuring device connected in series in an AC line, using the principle of electromagnetic induction to convert a large current into a small current, used to acquire the instantaneous values ​​of the three-phase output current. The effective value of each phase current refers to the value obtained by averaging the squares of the instantaneous values ​​of multiple sampling points of each phase current within one cycle, and then taking the square root. The formula is as follows: Where N is the number of sampling points, and j is the corresponding sampling point. This represents the instantaneous current value corresponding to N sampling points within one period, in amperes (A). The effective value of the fundamental component refers to the effective value of the sinusoidal component with the same frequency as the power grid frequency, obtained by Fourier decomposition of the non-sinusoidal periodic current waveform. The formula is: ,in This represents the peak amplitude of the fundamental component, expressed in amperes (A). and These are the complex coefficients obtained by Fourier transforming the instantaneous current value. The output current distortion rate refers to the ratio of the effective values ​​of all harmonic components (excluding the fundamental frequency) in the current waveform to the effective value of the fundamental frequency, expressed by the formula: ×100%, of which This is the effective value of phase A current. The effective value of the fundamental component is given by A, which represents the phase. There are three phases: A, B, and C. The result is expressed as a percentage. A fixed sampling period means that the time interval between two adjacent samples remains constant, for example, data is collected every 0.1 milliseconds.

[0021] In this step, firstly, in order to achieve synchronous recording of voltage and current in the time dimension, a voltage Hall sensor connected in parallel to the DC side of the converter is used to continuously collect the instantaneous voltage value of the DC bus at a fixed sampling period. The collected instantaneous voltage values ​​are then arranged in chronological order to generate a voltage fluctuation sequence, which yields a time-ordered voltage change trajectory. This provides time-aligned voltage data for solving subsequent oscillation frequency analysis and voltage drop depth calculation. For example, with a sampling period of 0.1ms, 10,000 points are collected within 1 second and arranged in time to form a sequence.

[0022] Furthermore, in order to obtain waveform details of the three-phase current to support distortion rate calculation, the instantaneous values ​​of the output currents of phases A, B, and C need to be synchronously collected within the same sampling period by a current transformer connected in series at the AC output end of the converter. This yields a synchronous instantaneous value matrix of the three-phase current, solving the problem that single-phase sampling cannot represent the whole when the three-phase distortion rates are inconsistent. For example, at 10:00:00.000, the instantaneous values ​​of the currents of phases A, B, and C are collected as 150A, -120A, and -30A, respectively.

[0023] Furthermore, in order to convert the instantaneous sampled values ​​into the effective values ​​commonly used in engineering, the root mean square values ​​of phases A, B, and C in each sampling period need to be calculated as the effective values ​​of the current in each phase, and the effective values ​​of the fundamental component in the current waveform of each phase are extracted. This yields the total effective value and fundamental component of the current in each phase, providing two key parameters for the distortion rate calculation. For example, in a certain sampling period, the effective value of the current in phase A is 100A, and the effective value of the fundamental component is 98A.

[0024] Furthermore, in order to quantitatively assess the degree of current waveform distortion, the instantaneous distortion rate of each phase needs to be calculated based on the effective value of the current in each phase and the effective value of the fundamental component according to the distortion rate formula. Then, the arithmetic mean of the three-phase distortion rates is taken as the output current distortion rate of the current cycle, which can yield a distortion rate index between 0 and 100%. This solves the problem of single-phase index distortion when the three-phase distortion is unbalanced. For example, if the distortion rate of phase A is 5%, phase B is 6%, and phase C is 4%, then the distortion rate of the current cycle is 5%.

[0025] Finally, in order to obtain a continuous record of the distortion rate changing over time, the distortion rates calculated from all sampling periods within the continuous time window need to be arranged in chronological order to generate a distortion rate sequence. This will give us distortion rate evolution data in time series form, which will provide input for the subsequent calculation of the light load oscillation risk coefficient. For example, if the time window is 1 second and contains 10,000 sampling periods, then the distortion rate sequence will contain 10,000 values.

[0026] It should be added that, in terms of data synchronization, the voltage Hall sensor and the current transformer have the same sampling period and their timestamps need to be aligned. For example, microsecond-level synchronization can be achieved by using the same clock source or by using the nearest neighbor interpolation method to ensure that the voltage and current data can be paired at every sampling moment.

[0027] For example, taking continuous monitoring data of a single converter unit as an example, the sampling period is 0.1ms, and the continuous time window is 1 second. The voltage Hall sensor collects the instantaneous DC bus voltage sequence, starting at 750.2V, followed by 750.1V, 749.8V, etc.; the current transformer synchronously collects the instantaneous A-phase current values ​​of 150A, -120A, -30A, etc., calculating the A-phase RMS value as 100A, the fundamental component RMS value as 98A, and the distortion rate as 5%; the B-phase RMS value as 98A, the fundamental component RMS value as 95.5A, and the distortion rate as 4.9%; the C-phase RMS value as 102A, the fundamental component RMS value as 99.5A, and the distortion rate as 5.1%; the average of the three phases gives a distortion rate of 5% for the current sampling period. A total of 10,000 sampling periods are collected within 1 second, arranged chronologically to generate a distortion rate sequence, with the values ​​in the sequence remaining stable around 5%.

[0028] In summary, this step synchronously acquires the instantaneous DC bus voltage and three-phase instantaneous current at a fixed sampling period using a voltage Hall sensor and a current transformer. After RMS calculation, fundamental frequency extraction, distortion rate calculation, and time series arrangement, a voltage fluctuation sequence and a current distortion rate sequence are generated. Compared with existing technologies, this step has the following advantages: First, it achieves microsecond-level synchronous acquisition of voltage and current data, solving the problem of data time misalignment; second, by averaging the three-phase distortion rate, it avoids misjudgments caused by single-phase distortion rate deviations; and third, it converts discrete sampled values ​​into a continuous time series, providing a complete time series data foundation for subsequent oscillation and drop analysis.

[0029] S200: Calculate the light load oscillation risk coefficient based on the output current distortion rate sequence and the load rate sequence, and calculate the heavy load voltage drop depth based on the integral value of the deviation between the DC bus voltage fluctuation sequence and the rated voltage.

[0030] In this embodiment of the application, based on the obtained DC bus voltage fluctuation sequence and output current distortion rate sequence, in order to quantify the degree of risk of light load oscillation and the severity of heavy load voltage drop, it is necessary to calculate the risk coefficient by combining the load rate sequence and calculate the drop depth by integrating the deviation between the voltage and the rated voltage. This solves the technical problem that the trend cannot be assessed by relying solely on single-point data and the lack of risk quantification indicators makes it impossible to trigger subsequent optimization control.

[0031] Step S200 in the method provided in this application embodiment includes: The instantaneous load rate corresponding to each sampling period within a continuous time window is read from the local controller of each transformer area converter. The instantaneous load rates are arranged in chronological order to generate the load rate sequence. Based on the distortion rate values ​​in the output current distortion rate sequence and the instantaneous load rate at the same sampling moment, the initial risk component for each sampling moment is determined. Sampling moments with distortion rate values ​​exceeding a reference distortion rate threshold are selected from all sampling moments, and the initial risk component of the corresponding sampling moment is taken as the oscillation risk component. The number of selected sampling moments is counted. The reference distortion rate threshold is determined based on the statistical distribution of historical output current distortion rate values ​​of the target transformer area converter under stable operating conditions. The oscillation risk components corresponding to all selected sampling moments are accumulated to obtain the cumulative risk value. Based on the cumulative risk value and the number of selected sampling moments, the light-load oscillation risk coefficient is calculated. Detailed explanation follows: In this embodiment, the instantaneous load rate refers to the ratio of the current output power of the converter to its rated power, reflecting the degree of load intensity. Its value ranges from 0 to 1, approaching 0 under light load. The initial risk component is an intermediate value obtained by multiplying the output current distortion rate by the instantaneous load rate, calculated as: Initial Risk Component = Distortion Rate × Load Rate. The reference distortion rate threshold is a threshold value determined based on the historical statistical distribution of the output current distortion rate of the target transformer converter under stable operating conditions. Its calculation formula is: arithmetic mean + 2 × standard deviation, used to determine whether the distortion rate is abnormal. Stable operating conditions refer to an output current distortion rate less than 2% for 10 consecutive minutes, a load rate between 30% and 70%, and DC bus voltage fluctuation less than 3% of the rated voltage. The oscillation risk component refers to the initial risk component at the corresponding sampling time when the distortion rate exceeds the reference distortion rate threshold, used for subsequent accumulation. The cumulative risk value is the sum of the oscillation risk components corresponding to all selected sampling times. The light load oscillation risk coefficient is the average value obtained by dividing the cumulative risk value by the number of sampled times selected, and is used to quantify the degree of light load oscillation risk.

[0032] In this step, firstly, in order to obtain load data that is time-aligned with the distortion rate sequence, the instantaneous load rate corresponding to each sampling period within a continuous time window needs to be read from the local controller of each transformer area converter, and the load rate sequence is generated by arranging them in chronological order. This yields a load rate sequence with the same length as the distortion rate sequence and corresponding one-to-one at the time, providing paired input for subsequent risk component calculation. For example, in S100, there are 10,000 sampling periods within a continuous second, and the instantaneous load rate of 0.2 is read in each period, forming a load rate sequence of 10,000 values.

[0033] Furthermore, in order to integrate the information from the two dimensions of distortion rate and load rate into a single-point risk indicator, it is necessary to multiply the distortion rate value at the same sampling time with the instantaneous load rate to determine the initial risk component at each sampling time. This yields the risk level value at each time, solving the problem of ignoring the influence of load rate when using distortion rate alone or ignoring the influence of distortion rate when using load rate alone. For example, a distortion rate of 5% multiplied by a load rate of 0.2 yields an initial risk component of 0.01.

[0034] Furthermore, in order to focus on the high distortion rate periods that truly require attention, it is necessary to screen out the sampling times from each sampling time where the distortion rate value exceeds the baseline distortion rate threshold, and take the initial risk component of the corresponding time as the oscillation risk component. At the same time, the number of the selected sampling times is counted to obtain the set of risk components and their number for high-risk periods, thus solving the problem of the accumulation of interference risk during low distortion rate periods. For example, if the baseline distortion rate threshold is 4% and the distortion rate sequence in S100 is around 5%, then all 10,000 sampling times are screened out.

[0035] Furthermore, in order to obtain the total risk energy within the time window, it is necessary to sum up the oscillation risk components corresponding to all the selected sampling times to obtain the cumulative risk value, which is the sum of the risk components at each time. For example, if each oscillation risk component is 0.01, the cumulative risk value of 100 is obtained by summing up 10,000 times.

[0036] Finally, to obtain the normalized average risk level, the light load oscillation risk coefficient needs to be calculated by dividing the cumulative risk value by the number of selected sampling points. This yields a risk index independent of the number of sampling points, facilitating comparisons across time windows. For example, dividing the cumulative risk value of 100 by 10000 gives a light load oscillation risk coefficient of 0.01. It should be noted that the baseline distortion rate threshold needs to be pre-calculated. Taking the distortion rate data under historical stable operating conditions of the target transformer area converter as an example, if the historical distortion rate arithmetic mean is 3% and the standard deviation is 0.5%, then the baseline distortion rate threshold = 3% + 2 × 0.5% = 4%.

[0037] Step S200 in the method provided in this application embodiment further includes: Based on the rated DC-side voltage specified in the factory nameplate parameters of each transformer area converter, the rated DC bus voltage is obtained. The instantaneous DC bus voltage value at each sampling moment in the DC bus voltage fluctuation sequence is compared with the rated voltage, and the instantaneous voltage deviation value at each sampling moment is calculated. When the instantaneous DC bus voltage value is lower than the rated voltage, the instantaneous voltage deviation value is the difference between the rated voltage and the instantaneous DC bus voltage value; otherwise, the instantaneous voltage deviation value is zero. Time intervals in the DC bus voltage fluctuation sequence where the instantaneous voltage deviation value is greater than zero for multiple consecutive sampling moments are identified as voltage sag intervals. Within each voltage sag interval, the instantaneous voltage deviation value is multiplied by the sampling time interval between adjacent sampling points and then summed to obtain the sag integral area corresponding to each voltage sag interval. The maximum value of the sag integral area among all voltage sag intervals is selected as the heavy-load voltage sag depth. Detailed explanation follows: In this embodiment, the instantaneous voltage deviation value refers to the difference when the instantaneous voltage value of the DC bus is lower than the rated voltage; otherwise, it is 0. The formula is: instantaneous voltage deviation value = max(rated voltage - instantaneous voltage, 0). The voltage sag interval refers to the time interval in the voltage fluctuation sequence where the instantaneous voltage deviation value is greater than 0 at multiple consecutive sampling times, with at least 3 consecutive sampling points. The sag integral area refers to the integral value obtained by multiplying the instantaneous voltage deviation value at each sampling time by the sampling time interval between adjacent sampling points within the voltage sag interval, expressed in volt-seconds (V·s). The heavy-load voltage sag depth refers to the maximum value of the sag integral area across all voltage sag intervals, used as a quantitative indicator to measure the severity of the voltage sag, expressed in volt-seconds (V·s).

[0038] In this step, firstly, in order to obtain a reference benchmark for voltage deviation, the rated voltage value needs to be obtained based on the DC side rated voltage specified in the factory nameplate parameters of each transformer area converter. For example, the DC side rated voltage of transformer area #1 in S100 is specified as 750V on the factory nameplate.

[0039] Furthermore, in order to calculate the degree to which the voltage is lower than the rated value at each sampling moment, the instantaneous voltage value at each sampling moment in the DC bus voltage fluctuation sequence needs to be compared with the rated voltage. When the instantaneous voltage is lower than the rated voltage, the difference is calculated as the instantaneous voltage deviation value; otherwise, the deviation value is 0, and a non-negative voltage deviation sequence can be obtained. For example, if the instantaneous voltage is 740V and the rated voltage is 750V, then the deviation value is 10V; if the instantaneous voltage is 755V, then the deviation value is 0.

[0040] Furthermore, in order to identify periods when the voltage is consistently below the rated value, it is necessary to find time intervals in the voltage fluctuation sequence where the deviation values ​​of multiple consecutive sampling times are greater than 0 as voltage drop intervals. Here, "multiple consecutive" means at least 3 consecutive sampling points, which can be used to obtain the start and end time range of each drop. For example, if the deviation values ​​of 5 consecutive sampling points (0.5ms) are all greater than 0, then these 5 points constitute a voltage drop interval.

[0041] Furthermore, in order to quantify the severity of each drop interval, the instantaneous voltage deviation value needs to be multiplied by the sampling time interval within each drop interval and then summed to obtain the drop integral area, which yields the volt-second product for each interval. For example, with a sampling interval of 0.1 ms and deviation values ​​of 10V, 12V, and 15V within the interval, the integral area = (10 + 12 + 15) × 0.0001 = 0.0037 V·s.

[0042] Finally, to obtain the most severe voltage drop event, the maximum value of the integral area of ​​the voltage drop across all voltage drop intervals should be selected as the heavy-load voltage drop depth. For example, if three drops occur within a window, with integral areas of 0.0037, 0.0085, and 0.0021 respectively, then the heavy-load voltage drop depth is taken as 0.0085 V·s. It should be noted that at least three consecutive sampling points should be used for the voltage drop interval, with a sampling period of 0.1 ms corresponding to a duration of 0.3 ms. This threshold can be adjusted according to the actual system noise level.

[0043] For example, taking a single-unit converter under heavy load conditions as an example, the rated voltage is 750V. In a voltage fluctuation sequence, five consecutive sampling points show instantaneous voltages of 740V, 735V, 730V, 732V, and 738V, all lower than 750V. The sampling interval is 0.1ms, and the instantaneous deviations are 10V, 15V, 20V, 18V, and 12V respectively. The voltage drop integral area = (10+15+20+18+12)×0.0001 = 0.0075V·s. Since there are no other larger integral areas within this window, the heavy load voltage drop depth is 0.0075V·s. Simultaneously, from the aforementioned steps, with a load factor of 0.2, a distortion rate of 5%, a baseline distortion rate threshold of 4%, an initial risk component of 0.01, a cumulative risk value of 100, and a sampling time count of 10000, the light load oscillation risk coefficient is 0.01.

[0044] In summary, this step achieves a quantitative assessment of two typical power quality problems by integrating distortion rate and load rate, screening out excessive moments, averaging the results to calculate the light-load oscillation risk coefficient, calculating instantaneous deviation, identifying voltage drop intervals, integrating and taking the maximum value to obtain the heavy-load voltage drop depth. Compared with existing technologies, this step has the following advantages: First, the light-load risk coefficient comprehensively considers the effects of distortion rate and load rate, and adopts an adaptive benchmark distortion rate threshold to avoid misjudgment or omission due to fixed thresholds; second, the heavy-load voltage drop depth uses the integrated area rather than the single-point minimum voltage, more realistically reflecting the cumulative effect of the voltage drop over time; third, the two indicators are independent of each other, providing clear triggering conditions for subsequent branch processing.

[0045] S300: When the light load oscillation risk coefficient exceeds the first threshold, the damping ratio attenuation trajectory under the same oscillation frequency mode is extracted from historical big data, and the dynamic compensation coefficient of the virtual impedance is calculated based on the damping ratio attenuation trajectory. In this embodiment of the application, when the light-load oscillation risk coefficient exceeds the first threshold, indicating that the converter is in a light-load oscillation state, in order to extract the damping ratio attenuation trajectory that matches the current oscillation frequency mode from historical big data and calculate the dynamic compensation coefficient of the virtual impedance based on the trajectory, it is necessary to first determine the current oscillation frequency according to the voltage fluctuation sequence, then retrieve historical events to extract the envelope, then calculate the logarithmic attenuation and the reference damping ratio, and finally combine the expected damping ratio and the rated virtual impedance value to obtain the dynamic compensation coefficient, so as to solve the technical problem of how to use the historical oscillation attenuation characteristics to quantify the compensation amount and avoid blind adjustment leading to system instability.

[0046] Step S300 in the method provided in this application embodiment includes: The system collects the light-load oscillation risk coefficients corresponding to multiple consecutive time windows during the historical steady-state operation of the target transformer area converter. Based on the statistical distribution of these light-load oscillation risk coefficients, a first threshold is determined. The currently calculated light-load oscillation risk coefficient is compared with the first threshold. When the light-load oscillation risk coefficient is greater than the first threshold, the operation of extracting the damping ratio attenuation trajectory and calculating the dynamic compensation coefficient is triggered. The current oscillation frequency is determined based on the zero-crossing characteristics of the continuous oscillation waveform in the DC bus voltage fluctuation sequence. Based on the current oscillation frequency, historical oscillation events with matching oscillation frequencies are retrieved from historical big data, and the oscillation amplitude of each historical oscillation event is extracted at any given time. The envelope formed by the inter-wave decay is used as the damping ratio decay trajectory. The peak amplitudes corresponding to two adjacent oscillation periods are extracted from the damping ratio decay trajectory, and the logarithmic decay is determined based on the ratio of the peak amplitudes. A reference damping ratio is calculated by combining the logarithmic decay with the time interval between adjacent peaks. The dynamic compensation coefficient of the virtual impedance is calculated based on the ratio of the reference damping ratio to the pre-stored desired damping ratio, combined with the rated virtual impedance value. The pre-stored desired damping ratio is determined based on the historical optimal damping ratio of the target transformer converter under rated load and no-oscillation conditions, and the rated virtual impedance value is taken from the fixed impedance value in the droop control parameters. Detailed explanation follows: In this embodiment, the first threshold refers to the threshold value used to determine whether the light load oscillation risk coefficient triggers compensation. The calculation formula is: First threshold = Arithmetic mean of light load oscillation risk coefficient + 3 × Standard deviation of light load oscillation risk coefficient. The damping ratio decay trajectory refers to the envelope of the oscillation amplitude extracted from historical oscillation events, which decays with the number of cycles, forming a curve with the cycle number as the abscissa and the peak amplitude as the ordinate. The dynamic compensation coefficient refers to the adjustment amount calculated based on the ratio of the reference damping ratio to the desired damping ratio. The formula is: Dynamic compensation coefficient = Rated virtual impedance value × [(Desired damping ratio / Reference damping ratio) - 1]. The logarithmic decay amount refers to the value obtained by extracting the peak amplitudes of two adjacent oscillation cycles from the damping ratio decay trajectory, taking the ratio, and then taking the natural logarithm. The formula is: Logarithmic decay amount = .

[0047] The rated virtual impedance value refers to the fixed impedance value taken from the droop control parameters, and the unit is Ω.

[0048] In this step, firstly, to obtain the threshold value for triggering light load compensation, it is necessary to collect the light load oscillation risk coefficients corresponding to multiple consecutive time windows during the historical steady-state operation of the target transformer area converter, and determine the first threshold based on their statistical distribution. For example, if the light load oscillation risk coefficients of 100 time windows in historical data have an arithmetic mean of 0.008 and a standard deviation of 0.001, then the first threshold = 0.008 + 3 × 0.001 = 0.011. The currently calculated light load oscillation risk coefficient of 0.012 is compared with the first threshold of 0.011. Since 0.012 > 0.011, subsequent operations are triggered.

[0049] Furthermore, to extract the damping characteristics from the damping ratio decay trajectory, it is necessary to extract the peak amplitudes of two adjacent oscillation cycles from the trajectory and determine the logarithmic decay based on the ratio. For example, taking the peak amplitude of the first cycle as 710V and the peak amplitude of the second cycle as 690V, the ratio 710 / 690≈1.029, and the logarithmic decay = ln(1.029)≈0.0286. Combining this with the time interval of 6.7ms between adjacent peaks, the reference damping ratio is calculated using the formula: Reference damping ratio = Logarithmic decay / (2×π) = 0.0286 / (2×3.1416)≈0.00455.

[0050] Finally, to obtain the dynamic compensation coefficient, it needs to be calculated based on the ratio of the reference damping ratio to the pre-stored expected damping ratio, combined with the rated virtual impedance value. The expected damping ratio is determined based on the historical best damping ratio. For example, if we take 0.01 and the rated virtual impedance value is 0.5Ω, then the dynamic compensation coefficient = 0.5 × [(0.01 / 0.00455) - 1] = 0.5 × (2.198 - 1) = 0.599Ω. It should be noted that the arithmetic mean and standard deviation of the light load oscillation risk coefficient in the first threshold calculation are derived from multiple time windows of data within the historical steady-state operating period. The expected damping ratio is determined based on the historical best damping ratio of the target transformer converter under rated load and no oscillation conditions.

[0051] Step S300 in the method provided in this application embodiment further includes: The oscillation start time is located from the DC bus voltage fluctuation sequence. Starting from the oscillation start time, the instantaneous voltage values ​​are scanned sequentially. The sampling time corresponding to when the instantaneous DC bus voltage changes from below the rated voltage to above the rated voltage is recorded as a positive zero-crossing point. At least three positive zero-crossing points are recorded consecutively. The number of sampling points between two adjacent positive zero-crossing points is calculated, and the number of sampling points is multiplied by the sampling time interval between adjacent sampling points to obtain the oscillation period value. The reciprocal of each oscillation period value is used as the instantaneous frequency value of the corresponding period to form an instantaneous frequency value sequence. Abnormal fluctuation values ​​in the instantaneous frequency value sequence are removed, and the remaining instantaneous frequency values ​​are smoothed. The processing result is used as the current oscillation frequency. Detailed explanation is as follows: In this embodiment, the oscillation start time refers to the first sampling time at which the oscillation begins in the DC bus voltage fluctuation sequence. The positive zero-crossing point refers to the sampling time corresponding to when the instantaneous DC bus voltage changes from below the rated voltage to above the rated voltage. The instantaneous frequency value refers to the frequency value calculated based on the time interval between adjacent positive zero-crossing points, with the formula: instantaneous frequency = 1 / oscillation period, and the unit is Hz.

[0052] In this step, firstly, in order to locate the start time of oscillation, it is necessary to scan and identify the oscillation start time from the voltage fluctuation sequence. For example, when the peak-to-peak voltage of three consecutive cycles exceeds 5% of the rated voltage of 750V, i.e., 37.5V, the first zero-crossing point is taken as the oscillation start time.

[0053] Furthermore, to obtain stable periodic measurements, at least three positive zero-crossing points need to be recorded. For example, after the start of oscillation, positive zero-crossing points appear at t1=1.0000ms, t2=1.0067ms, and t3=1.0134ms, with an adjacent interval of 0.0067ms. With a sampling interval of 0.1ms, the number of sampling points between adjacent zero-crossing points is 67, and the oscillation period = 67 × 0.1ms = 6.7ms.

[0054] Furthermore, to calculate the instantaneous frequency, the reciprocal of each oscillation period value needs to be taken. For example, a period of 6.7 ms corresponds to an instantaneous frequency of 1 / 0.0067 ≈ 149 Hz. A sequence of instantaneous frequency values ​​is obtained by calculating multiple periods sequentially, such as 149 Hz, 150 Hz, and 151 Hz.

[0055] Finally, to eliminate noise interference, abnormal fluctuation values ​​in the sequence (such as points deviating from the arithmetic mean by more than 3 standard deviations) need to be removed, and the remaining values ​​need to be smoothed (e.g., by taking the arithmetic mean) to obtain the current oscillation frequency of 150Hz. It should be noted that the removal of abnormal fluctuation values ​​follows the 3σ principle, and the smoothing process can use moving average or median filtering.

[0056] Step S300 in the method provided in this application embodiment further includes: Centered on the current oscillation frequency, a frequency retrieval interval is defined by extending half the interquartile range of historical oscillation frequencies to both sides. The interquartile range is the difference between the upper and lower quartiles of the historical oscillation frequency. Oscillation events in the historical big data are traversed, and the oscillation frequency of each event is extracted. Oscillation events whose frequencies fall within the frequency retrieval interval are marked as matching events. For each matching event, the corresponding historical sequence of DC bus voltage fluctuations is read. The oscillation start point is identified from the historical sequence of DC bus voltage fluctuations, and the peak amplitude of each oscillation cycle is extracted sequentially. The extracted peak amplitudes are connected according to the time sequence of the oscillation cycles, with the oscillation cycle number as the horizontal axis and the peak amplitude as the vertical axis, forming an envelope that decreases with the number of cycles. This envelope is used as the damping ratio decay trajectory of the corresponding matching event. Detailed explanation follows: In this embodiment, the frequency retrieval interval refers to a frequency range centered on the current oscillation frequency and with a width equal to the interquartile range of historical oscillation frequencies. The interquartile range refers to the difference between the upper and lower quartiles of the historical oscillation frequencies. The envelope refers to the curve formed by connecting the peak amplitudes of each oscillation cycle in the oscillation event according to the cycle number, i.e., the damping ratio decay trajectory. The peak amplitude refers to the maximum value reached by the DC bus voltage waveform in each oscillation cycle.

[0057] In this step, firstly, to set a reasonable frequency search range, the current oscillation frequency of 150Hz should be used as the center, and the interquartile range of the historical oscillation frequency should be used as the width. Let the lower quartile of the historical oscillation frequency be 145Hz and the upper quartile be 155Hz. Then the interquartile range = 10Hz, the lower limit of the search range = 150 - 10 / 2 = 145Hz, and the upper limit = 150 + 10 / 2 = 155Hz.

[0058] Furthermore, to filter out historical events with similar frequencies, it is necessary to traverse all oscillation events in the historical big data and mark events whose oscillation frequencies fall within the range of [145Hz, 155Hz] as matching events. For example, 5 historical events were matched.

[0059] Furthermore, in order to extract the damping ratio decay trajectory for each matched event, it is necessary to read the corresponding historical voltage fluctuation sequence, identify the oscillation start point, and extract the peak amplitude of each oscillation cycle in sequence. For example, in a certain matched event, the peak amplitude is 710V in the first cycle, 690V in the second cycle, 675V in the third cycle, and 665V in the fourth cycle.

[0060] Finally, to form the envelope, the peak amplitudes need to be connected sequentially according to the period number, with the period number as the x-axis and the peak amplitude as the y-axis, resulting in a decreasing curve, which is the damping ratio decay trajectory of the event. For example, the curve formed by connecting points (1,710), (2,690), (3,675), and (4,665). It should be noted that if no matching event is found within the search interval, the interval width is gradually increased by a factor of 1.2 until at least one event is matched, or the nearest neighbor frequency event is used as a substitute.

[0061] For example, taking a single-unit converter under light load conditions, the light load oscillation risk coefficient of 0.012 exceeds the first threshold of 0.011. The oscillation start time is located from the voltage fluctuation sequence, and the positive zero-crossing points t1=1.0000ms, t2=1.0067ms, and t3=1.0134ms are recorded. The oscillation period is calculated to be 6.7ms, the instantaneous frequency is 149Hz, and the smoothed current oscillation frequency is 150Hz. Using 150Hz as the center and a historical frequency interquartile range of 10Hz, a search interval [145Hz, 155Hz] is set, matching three historical events. One of these events is selected, with peak amplitude sequences of 710V, 690V, 675V, and 665V, and a logarithmic decay... =0.0286, reference damping ratio 0.0286 / (2×3.1416)=0.00455, expected damping ratio 0.01, rated virtual impedance 0.5Ω, dynamic compensation coefficient =0.5×[(0.01 / 0.00455)-1]=0.599Ω.

[0062] In summary, this step involves triggering the light-load oscillation risk coefficient, measuring frequency using zero-crossing features, retrieving historical similar events, extracting the damping ratio decay trajectory, calculating the logarithmic decay and reference damping ratio, and finally obtaining the dynamic compensation coefficient, thus achieving adaptive compensation for light-load oscillations based on historical big data. Compared with existing technologies, this step has the following advantages: First, it uses similar oscillation events from historical big data to guide current compensation, avoiding trial-and-error based on experience; second, it quantifies the oscillation decay characteristics through the damping ratio decay trajectory, giving the compensation coefficient a clear physical meaning; and third, it improves the accuracy and robustness of matching based on statistical adaptive thresholds and retrieval intervals.

[0063] S400: When the heavy load voltage drop depth exceeds the second threshold, based on the ratio of the current change rate to the voltage recovery speed at the start of the drop, the recovery time constant under similar drop patterns is matched from the historical big data, and a correction amount for the voltage support coefficient is generated. In this embodiment, when the heavy load voltage drop depth exceeds the second threshold, it indicates that the converter is in a heavy load voltage drop state. It is necessary to match the recovery time constant under similar drop conditions from historical big data based on the ratio of the current change rate to the voltage recovery speed at the start of the drop, and generate the correction amount of the voltage support coefficient. This is to solve the technical problem of how to use the recovery characteristics of similar historical drop events to quantify the current required voltage support strength and avoid blind adjustment leading to overcompensation or undercompensation.

[0064] Step S400 in the method provided in this application embodiment includes: The system collects the heavy-load voltage sag depth corresponding to multiple consecutive time windows during the historical steady-state operation of the target transformer area converter. A second threshold is determined based on the statistical distribution of the heavy-load voltage sag depth. The currently calculated heavy-load voltage sag depth is compared with the second threshold. When the heavy-load voltage sag depth is greater than the second threshold, the voltage sag initiation time is identified from the DC bus voltage fluctuation sequence, and the instantaneous output current value within one sampling period before and after the sag initiation time is simultaneously acquired. The current change rate is calculated based on the ratio of the difference between the instantaneous output current values ​​before and after the sag initiation time to the corresponding time interval. The DC bus instantaneous voltage after the sag initiation time is then calculated. The rate of change of the current value is used to calculate the voltage recovery speed; the ratio of the current change rate to the voltage recovery speed is calculated as the current voltage drop characteristic value; based on the current voltage drop characteristic value, the historical voltage drop event with the smallest characteristic value deviation is retrieved from the historical big data, and the time required for the voltage to recover from the lowest point of the voltage drop to a specified proportion of the rated voltage in the historical voltage drop event is extracted as the recovery time constant; the ratio of the reference recovery time constant to the recovery time constant is used as the correction amount of the voltage support coefficient, wherein the reference recovery time constant is the statistical average value of the time required for the voltage of the target transformer area converter to naturally recover from the lowest point of the voltage drop to the rated voltage under rated load conditions. Detailed explanation follows: In this embodiment, the second threshold refers to the threshold value used to determine whether the heavy-load voltage drop depth triggers support compensation. The calculation formula is: Second threshold = Arithmetic mean of heavy-load voltage drop depth + 3 × Standard deviation of heavy-load voltage drop depth. The current change rate refers to the rate of change of current before and after the start of the drop. The calculation formula is: Current change rate = (Instantaneous current value in the sampling period after the start of the drop - Instantaneous current value in the sampling period before the start of the drop) / (2 × Sampling period). The voltage recovery speed refers to the rate of voltage rise after the lowest point of the voltage drop. The calculation formula is: Voltage recovery speed = (Rated voltage × Specified proportion - Voltage at the lowest point of the drop) / Recovery time. The drop morphology characteristic value refers to the ratio of the current change rate to the voltage recovery speed. The formula is: Drop morphology characteristic value = Current change rate / Voltage recovery speed. The recovery time constant refers to the time required for the voltage to recover from the lowest point of the drop to a specified proportion of the rated voltage in historical voltage drop events. The specified proportion is 90%, and the unit is seconds (s). The reference recovery time constant refers to the statistical average of the time required for the voltage of the target transformer area converter to naturally recover from the lowest point of voltage drop to the rated voltage under rated load conditions, measured in seconds. The correction amount is the ratio of the reference recovery time constant to the recovery time constant, expressed as: Correction amount = Reference recovery time constant / Recovery time constant.

[0065] In this step, firstly, to obtain the threshold value for triggering overload compensation, it is necessary to collect the overload voltage sag depth corresponding to multiple consecutive time windows during the historical steady-state operation of the target transformer area converter, and determine the second threshold based on its statistical distribution. For example, if the overload voltage sag depth of 100 time windows in the historical data is taken, with an arithmetic mean of 0.005 V·s and a standard deviation of 0.001 V·s, then the second threshold = 0.005 + 3 × 0.001 = 0.008 V·s. Comparing the currently calculated overload voltage sag depth of 0.0085 V·s with the second threshold of 0.008 V·s, since 0.0085 > 0.008, subsequent operations are triggered.

[0066] Furthermore, to extract the voltage drop pattern characteristics, it is necessary to identify the starting time of the voltage drop from the voltage fluctuation sequence and simultaneously acquire the instantaneous output current value within one sampling period before and after that time. For example, if the starting time of the voltage drop is t0 = 10.0000 ms, the sampling period is 0.1 ms, the instantaneous current value at time t-1 of the previous sampling period is 100 A, and the instantaneous current value at time t+1 of the next sampling period is 150 A, then the current change rate = (150 - 100) / (2 × 0.0001) = 50 / 0.0002 = 250000 A / s.

[0067] Furthermore, to calculate the voltage recovery rate, the instantaneous change rate of the DC bus voltage after the initial voltage drop needs to be considered. For example, with a sampling period of 0.1 ms, taking the next 10 sampling periods starting from the initial voltage drop, the measured voltage values ​​are: 750.000V, 749.999V, 749.998V, 749.997V, 749.996V, 749.995V, 749.994V, 749.993V, 749.992V, and 749.991V. Through linear fitting, the initial voltage change rate is calculated to be approximately -10V / s, and the absolute value is taken as the voltage recovery rate = 10V / s.

[0068] Furthermore, to match similar historical voltage drop events, based on the current voltage drop pattern characteristic value of 25000, it is necessary to retrieve the historical voltage drop event with the smallest characteristic value deviation from the historical big data. For example, the characteristic values ​​of historical events are 24800, 25200, and 26000, among which the smallest deviation is 24800, with a deviation of 200. The time required for the voltage to recover from the lowest point of the drop to 90% of the rated voltage in this event is extracted as the recovery time constant, for example, 0.5s.

[0069] Finally, to generate the correction, the ratio of the baseline recovery time constant to the recovery time constant needs to be taken. The baseline recovery time constant is taken as the statistical average value under rated load conditions, for example, 0.6s, then the correction = 0.6 / 0.5 = 1.2.

[0070] It should be added that the arithmetic mean and standard deviation of the heavy-load voltage drop depth in the second threshold calculation are derived from data from multiple time windows within the historical steady-state operating period. The specified percentage can also be 95%, adjusted according to system requirements. The absolute difference is used to minimize the eigenvalue deviation; if multiple events have the same deviation, the median recovery time is used.

[0071] For example, taking a single-unit converter under heavy load conditions, the voltage drop depth of 0.0085V·s exceeds the second threshold of 0.008V·s. The instantaneous current values ​​before and after the drop initiation are 100A and 150A respectively, with a sampling period of 0.1ms and a current change rate of 250,000A / s. The lowest voltage point of the drop is 670V, and it takes 0.5s to recover to 90% of the rated voltage (675V), with a voltage recovery speed of 10V / s and a drop morphology characteristic value of 25,000. An event with characteristic value 24,800 is matched in the historical database, with a recovery time constant of 0.5s, a baseline recovery time constant of 0.6s, and a correction amount of 0.6 / 0.5 = 1.2.

[0072] In summary, this step uses the ratio of current change rate to voltage recovery speed as a sag morphology characteristic value, matches the recovery time constant of the most similar sag event from historical big data, and obtains the correction amount by comparing it with the benchmark value, thus achieving adaptive support compensation for heavy-load voltage sags. Compared with existing technologies, this step has the following advantages: First, it quantifies the relationship between current surge and voltage recovery using sag morphology characteristic values, making the matching more accurate; second, it guides current compensation based on the recovery time of similar historical events, avoiding the limitations of fixed time constants; and third, the specified ratio is adjustable to adapt to the recovery requirements of different systems.

[0073] S500: The dynamic compensation coefficient and the correction amount are fused nonlinearly to obtain an adaptive virtual impedance adjustment value; In this embodiment of the application, based on the obtained dynamic compensation coefficient and correction amount, it is necessary to obtain an adaptive virtual impedance adjustment value by nonlinear fusion of the two to solve the technical problem of how to coordinate when the two compensation coefficients may be physically different and avoid adjustment conflicts that lead to system performance degradation.

[0074] Step S500 in the method provided in this application embodiment includes: The dynamic compensation coefficient is added to the reference compensation offset to obtain the normalized compensation coefficient, where the reference compensation offset is 1. The smaller value between the correction amount and the reference correction offset is taken as the normalized correction amount, where the reference correction offset is 1. The square root of the normalized compensation coefficient is calculated to determine the nonlinear compensation factor. The nonlinear compensation factor is multiplied by the normalized correction amount to obtain the fusion factor. The fixed impedance value in the droop control parameters is adjusted according to the fusion factor to obtain the adaptive virtual impedance adjustment value. Detailed explanation is as follows: In this embodiment, the normalized compensation coefficient refers to the value of the dynamic compensation coefficient plus 1, and the formula is: Normalized compensation coefficient = Dynamic compensation coefficient + 1. The normalized correction amount refers to the smaller value between the correction amount and 1, and the formula is: Normalized correction amount = min(Correction amount, 1). The nonlinear compensation factor refers to the square root of the normalized compensation coefficient, and the formula is: Nonlinear compensation factor = The fusion factor is the product of the nonlinear compensation factor and the normalization correction, expressed as: Fusion Factor = Nonlinear Compensation Factor × Normalization Correction. The adaptive virtual impedance adjustment value is the fusion factor multiplied by the fixed impedance value in the droop control parameters, expressed as: Adaptive Virtual Impedance = Fixed Impedance × Fusion Factor.

[0075] In this step, firstly, in order to convert the dynamic compensation coefficient to a non-negative value and facilitate nonlinear processing, the dynamic compensation coefficient needs to be added to the reference compensation offset of 1 to obtain the normalized compensation coefficient. For example, if the dynamic compensation coefficient is 0.599, then the normalized compensation coefficient = 0.599 + 1 = 1.599.

[0076] Furthermore, to ensure that the overload correction does not lead to an increase in virtual impedance, the smaller value between the correction and 1 should be used as the normalized correction. For example, if the correction is 1.2, then the normalized correction is min(1.2,1) = 1.0. If the correction is 0.8, then the normalized correction is 0.8.

[0077] Furthermore, in order to compress the light-load compensation coefficient and reduce its weight in the fusion result, the square root of the normalized compensation coefficient needs to be calculated as a nonlinear compensation factor. For example, if the normalized compensation coefficient is 1.599, the square root ≈ 1.264.

[0078] Furthermore, to integrate the two compensation requirements, the nonlinear compensation factor needs to be multiplied by the normalization correction to obtain the fusion factor. For example, if the nonlinear compensation factor is 1.264 and the normalization correction is 1.0, then the fusion factor = 1.264 × 1.0 = 1.264. If the normalization correction is 0.8, then the fusion factor = 1.264 × 0.8 = 1.0112.

[0079] Finally, to obtain the final control parameters, the fixed impedance value in the droop control parameters needs to be adjusted according to the fusion factor. Assuming the fixed impedance value is 0.5Ω, then the adaptive virtual impedance = 0.5 × 1.264 = 0.632Ω. If the fusion factor is less than 1, the impedance decreases.

[0080] It should be added that taking the smaller value between the correction amount and 1 ensures that the normalized correction amount does not exceed 1 under heavy load drop conditions, thus avoiding an unexpected increase in impedance due to an excessively large correction amount. The square root operation compresses the variation range of the dynamic compensation coefficient, preventing excessive compensation under light load from dominating the fusion result.

[0081] For example, taking a single-unit converter as an example, the dynamic compensation coefficient is 0.599, and the correction amount is 1.2. The normalized compensation coefficient = 1.599, square root ≈ 1.264. The normalized correction amount = min(1.2,1) = 1.0. The fusion factor = 1.264 × 1.0 = 1.264. The fixed impedance is 0.5Ω, and the adaptive virtual impedance = 0.5 × 1.264 = 0.632Ω. If the correction amount is 0.8, then the normalized correction amount = 0.8, the fusion factor = 1.264 × 0.8 = 1.0112, and the adaptive virtual impedance = 0.5056Ω, slightly larger than the original impedance. If the dynamic compensation coefficient is 0, then the normalized compensation coefficient = 1, square root = 1, fusion factor = 1 × 0.8 = 0.8, and the adaptive virtual impedance = 0.4Ω, reducing the impedance to support the voltage.

[0082] In summary, this step organically combines the compensation requirements for light-load oscillations and heavy-load drops into a final adaptive virtual impedance adjustment value through normalization by adding 1, taking a small clamping value, square root nonlinearity, product fusion, and multiplication by a fixed impedance. Compared with existing technologies, this step has the following advantages: First, taking the smaller value between the correction amount and 1 ensures the correct physical direction of prioritizing impedance reduction under heavy load; second, the square root of the dynamic compensation coefficient achieves nonlinear compression, avoiding excessive interference of heavy-load response with light-load compensation; third, the fusion factor has a simple form, is easy to implement in engineering, and has a clear physical meaning.

[0083] S600: Replace the fixed impedance value in the droop control parameters of each transformer area converter with the adaptive virtual impedance adjustment value, generate an optimized control command, and send it to the pulse width modulator of each transformer area converter.

[0084] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects: This application proposes a low-voltage flexible DC power quality optimization method based on big data. First, the DC bus voltage fluctuation sequence and output current distortion rate sequence of each transformer area's converter are obtained. Then, the light-load oscillation risk coefficient and the heavy-load voltage drop depth are calculated. When the light-load risk exceeds a first threshold, the damping ratio attenuation trajectory of the same frequency is extracted from historical big data to calculate the virtual impedance dynamic compensation coefficient. When the heavy-load drop exceeds a second threshold, the recovery time constant of the historical drop pattern is matched according to the ratio of current change rate to voltage recovery speed to generate a voltage support correction. Finally, the dynamic compensation coefficient and the correction are nonlinearly fused to obtain an adaptive virtual impedance adjustment value, which replaces the fixed impedance value and is sent to the pulse width modulator. This method, through dual-condition divide-and-conquer and historical big data matching, solves the problem that fixed control parameters are difficult to balance under contradictory operating conditions of light-load oscillation and heavy-load drop.

[0085] Example 2, as shown in the appendix Figure 2As shown, based on the inventive concept of the low-voltage flexible DC power quality optimization method based on big data provided in Embodiment 1, this application also provides a low-voltage flexible DC power quality optimization system based on big data, specifically including: Data acquisition module 11 is deployed on the DC and AC sides of the converters in each distribution area to acquire the DC bus voltage fluctuation sequence and output current distortion rate sequence within a continuous time window; Risk analysis module 12 is used to calculate the light load oscillation risk coefficient based on the output current distortion rate sequence and the load rate sequence, and to calculate the heavy load voltage drop depth based on the integral value of the deviation between the DC bus voltage fluctuation sequence and the rated voltage. The compensation calculation module 13 is used to extract the damping ratio decay trajectory under the same oscillation frequency mode from historical big data when the light load oscillation risk coefficient exceeds the first threshold, and calculate the dynamic compensation coefficient of the virtual impedance based on the damping ratio decay trajectory; and when the heavy load voltage drop depth exceeds the second threshold, match the recovery time constant under similar drop modes from the historical big data according to the ratio of the current change rate to the voltage recovery speed at the start of the drop, and generate the correction amount of the voltage support coefficient. The nonlinear fusion module 14 is used to obtain an adaptive virtual impedance adjustment value by nonlinearly fusing the dynamic compensation coefficient and the correction amount. The instruction issuing module 15 is used to replace the fixed impedance value in the droop control parameters of each transformer area converter with the adaptive virtual impedance adjustment value, generate an optimized control instruction, and issue it to the pulse width modulator of each transformer area converter.

[0086] In one embodiment, the data acquisition module 11 is further configured to: continuously acquire the instantaneous DC bus voltage value at a fixed sampling period using voltage Hall sensors connected in parallel between the positive and negative poles of the DC side of each transformer converter, and arrange the acquired instantaneous voltage values ​​at each sampling time in chronological order to generate the DC bus voltage fluctuation sequence; synchronously acquire the instantaneous three-phase output current values ​​in each sampling period using current transformers connected in series at the AC side output terminals of each transformer converter; calculate the root mean square values ​​of the instantaneous output current values ​​of phase A, phase B, and phase C for each sampling period as the effective value of each phase current, and extract the effective value of the fundamental component in the waveform of each phase current; calculate the instantaneous distortion rate of each phase based on the effective value of each phase current and the effective value of the corresponding phase fundamental component, and take the arithmetic mean of the instantaneous distortion rates of each phase as the output current distortion rate of the current sampling period; and arrange the output current distortion rates calculated in all sampling periods within the continuous time window in chronological order to generate the output current distortion rate sequence.

[0087] In one embodiment, the risk analysis module 12 is further configured to: read the instantaneous load rate corresponding to each sampling period within a continuous time window from the local controller of each transformer converter; arrange the instantaneous load rates in chronological order to generate the load rate sequence; determine the initial risk component of each sampling moment based on the distortion rate value in the output current distortion rate sequence and the instantaneous load rate at the same sampling moment; select sampling moments from each sampling moment whose distortion rate value exceeds a reference distortion rate threshold, take the initial risk component of the corresponding sampling moment as the oscillation risk component, and count the number of selected sampling moments; wherein, the reference distortion rate threshold is determined based on the statistical distribution of the historical values ​​of the output current distortion rate of the target transformer converter under stable operating conditions; accumulate the oscillation risk components corresponding to all selected sampling moments to obtain a cumulative risk value; and calculate the light load oscillation risk coefficient based on the cumulative risk value and the number of selected sampling moments.

[0088] Furthermore, the risk analysis module 12 is also used to: obtain the rated voltage of the DC bus based on the DC side rated voltage specified in the factory nameplate parameters of each transformer area converter; compare the instantaneous DC bus voltage value at each sampling moment in the DC bus voltage fluctuation sequence with the rated voltage, and calculate the instantaneous voltage deviation value at each sampling moment, wherein when the instantaneous DC bus voltage value is lower than the rated voltage, the instantaneous voltage deviation value is the difference between the rated voltage and the instantaneous DC bus voltage value, otherwise the instantaneous voltage deviation value is zero; identify the time intervals in the DC bus voltage fluctuation sequence where the instantaneous voltage deviation value corresponding to multiple consecutive sampling moments is greater than zero, as voltage drop intervals; within each voltage drop interval, multiply the instantaneous voltage deviation value by the sampling time interval between adjacent sampling points and accumulate them to obtain the drop integral area corresponding to each voltage drop interval; select the maximum value of the drop integral area among all voltage drop intervals as the heavy load voltage drop depth.

[0089] In one embodiment, the compensation calculation module 13 is further configured to: collect the light load oscillation risk coefficients corresponding to multiple consecutive time windows during the historical steady-state operation period of the target transformer area converter; determine the first threshold based on the statistical distribution of the light load oscillation risk coefficients; compare the currently calculated light load oscillation risk coefficients with the first threshold; and when the light load oscillation risk coefficients are greater than the first threshold, trigger the operation of extracting the damping ratio attenuation trajectory and calculating the dynamic compensation coefficient; determine the current oscillation frequency based on the zero-crossing characteristics of the continuous oscillation waveform in the DC bus voltage fluctuation sequence; and, based on the current oscillation frequency, retrieve historical oscillation events with matching oscillation frequencies from historical big data and extract each historical oscillation event. The envelope formed by the decay of the oscillation amplitude over time during the oscillation event is used as the damping ratio decay trajectory. The peak amplitudes corresponding to two adjacent oscillation cycles are extracted from the damping ratio decay trajectory, and the logarithmic decay is determined based on the ratio of the peak amplitudes. The reference damping ratio is calculated by combining the logarithmic decay with the time interval between adjacent peaks. The dynamic compensation coefficient of the virtual impedance is calculated based on the ratio of the reference damping ratio to the pre-stored expected damping ratio and the rated virtual impedance value. The pre-stored expected damping ratio is determined based on the historical optimal damping ratio of the target transformer converter under rated load and no oscillation conditions, and the rated virtual impedance value is taken from the fixed impedance value in the droop control parameters.

[0090] Furthermore, the compensation calculation module 13 is also used to: locate the oscillation start time from the DC bus voltage fluctuation sequence; scan the instantaneous voltage value sequentially from the oscillation start time; record the sampling time corresponding to when the DC bus instantaneous voltage value changes from below the rated voltage to above the rated voltage, as a positive zero-crossing point, and continuously record at least three positive zero-crossing points; calculate the number of sampling points between two adjacent positive zero-crossing points, multiply the number of sampling points by the sampling time interval between adjacent sampling points to obtain the oscillation period value; take the reciprocal of each oscillation period value as the instantaneous frequency value of the corresponding period to form an instantaneous frequency value sequence; remove abnormal fluctuation values ​​in the instantaneous frequency value sequence, smooth the remaining instantaneous frequency values, and take the processing result as the current oscillation frequency.

[0091] Furthermore, the compensation calculation module 13 is also used to: extend half of the interquartile range of historical oscillation frequencies to both sides of the current oscillation frequency as the center, and set a frequency retrieval interval, wherein the interquartile range is the difference between the upper and lower quartiles of the historical oscillation frequency; traverse the oscillation events in the historical big data, extract the oscillation frequency of each oscillation event, and mark the oscillation events whose oscillation frequencies fall within the frequency retrieval interval as matching events; for each matching event, read the corresponding DC bus voltage fluctuation historical sequence; identify the oscillation start point from the DC bus voltage fluctuation historical sequence, and extract the peak amplitude of each oscillation cycle in sequence; connect the extracted peak amplitudes according to the time order of the oscillation cycle, and form an envelope that decreases with the number of cycles with the oscillation cycle number as the horizontal axis and the peak amplitude as the vertical axis, and use the envelope as the damping ratio decay trajectory of the corresponding matching event.

[0092] Furthermore, the compensation calculation module 13 is also used for: collecting the heavy-load voltage sag depth corresponding to multiple consecutive time windows during the historical steady-state operation period of the target transformer area converter; determining the second threshold based on the statistical distribution of the heavy-load voltage sag depth; comparing the currently calculated heavy-load voltage sag depth with the second threshold; when the heavy-load voltage sag depth is greater than the second threshold, identifying the voltage sag initiation time from the DC bus voltage fluctuation sequence, and simultaneously acquiring the instantaneous output current value within one sampling period before and after the sag initiation time; calculating the current change rate based on the ratio of the difference between the instantaneous output current values ​​before and after the sag initiation time and the corresponding time interval; and calculating the current change rate based on the ratio of the difference between the instantaneous output current values ​​before and after the sag initiation time and the corresponding time interval. The rate of change of the instantaneous DC bus voltage is used to calculate the voltage recovery speed; the ratio of the current change rate to the voltage recovery speed is calculated as the current voltage drop characteristic value; based on the current voltage drop characteristic value, the historical voltage drop event with the smallest characteristic value deviation is retrieved from the historical big data, and the time required for the voltage to recover from the lowest point of the voltage drop to a specified proportion of the rated voltage in the historical voltage drop event is extracted as the recovery time constant; the ratio of the reference recovery time constant to the recovery time constant is used as the correction amount of the voltage support coefficient, wherein the reference recovery time constant is the statistical average value of the time required for the voltage of the target transformer area converter to naturally recover from the lowest point of the voltage drop to the rated voltage under rated load conditions.

[0093] In one embodiment, the nonlinear fusion module 14 is further configured to: add the dynamic compensation coefficient to the reference compensation offset to obtain a normalized compensation coefficient, wherein the reference compensation offset is 1; take the smaller value between the correction amount and the reference correction offset as the normalized correction amount; wherein the reference correction offset is 1; calculate the square root of the normalized compensation coefficient to determine the nonlinear compensation factor; multiply the nonlinear compensation factor by the normalized correction amount to obtain a fusion factor; and adjust the fixed impedance value in the droop control parameters according to the fusion factor to obtain the adaptive virtual impedance adjustment value.

[0094] The low-voltage flexible DC power quality optimization system provided in this application, based on big data, can achieve intelligent closed-loop power quality management in scenarios such as daily operation of low-voltage flexible DC distribution networks, suppression of light-load oscillations in transformer substations, and compensation for heavy-load voltage dips. This management process involves multi-source data acquisition, dual-condition risk quantification, and adaptive virtual impedance adjustment. It can be integrated into DC microgrid energy management systems or local converter control devices, effectively improving power quality optimization under contradictory conditions of light-load oscillations and heavy-load voltage dips, reducing system instability risk and voltage dip depth, and providing reliable data support for adaptive adjustment of converter control parameters. For the specific workflow and optimization details of this system, please refer to Example 1.

[0095] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

Claims

1. A method for optimizing the quality of low-voltage flexible DC power based on big data, characterized in that, The method includes: Obtain the DC bus voltage fluctuation sequence and output current distortion rate sequence of each transformer area converter within a continuous time window; The light-load oscillation risk coefficient is calculated based on the output current distortion rate sequence and the load rate sequence, and the heavy-load voltage drop depth is calculated based on the integral value of the deviation between the DC bus voltage fluctuation sequence and the rated voltage. When the light load oscillation risk coefficient exceeds the first threshold, the damping ratio attenuation trajectory under the same oscillation frequency mode is extracted from historical big data, and the dynamic compensation coefficient of the virtual impedance is calculated based on the damping ratio attenuation trajectory. When the heavy load voltage drop depth exceeds the second threshold, the recovery time constant under similar drop conditions is matched from the historical big data based on the ratio of the current change rate to the voltage recovery speed at the start of the drop, and the correction amount of the voltage support coefficient is generated. The dynamic compensation coefficient and the correction amount are fused nonlinearly to obtain the adaptive virtual impedance adjustment value; The fixed impedance value in the droop control parameters of each transformer area converter is replaced with the adaptive virtual impedance adjustment value to generate an optimized control command and send it to the pulse width modulator of each transformer area converter.

2. The low-voltage flexible DC power quality optimization method based on big data according to claim 1, characterized in that, The steps for obtaining the DC bus voltage fluctuation sequence and the output current distortion rate sequence include: By using voltage Hall sensors connected in parallel between the positive and negative poles of the DC side of each transformer converter, the instantaneous voltage value of the DC bus is continuously collected at a fixed sampling period, and the instantaneous voltage values ​​collected at each sampling time are arranged in chronological order to generate the DC bus voltage fluctuation sequence. The instantaneous value of the three-phase output current is synchronously collected in each sampling cycle by a current transformer connected in series at the AC output end of the converter in each area. For each sampling period, the root mean square value of the instantaneous output current of phase A, phase B, and phase C is calculated as the effective value of the current of each phase, and the effective value of the fundamental component in the waveform of each phase current is extracted. Based on the effective value of the current in each phase and the effective value of the fundamental component of the corresponding phase, the instantaneous distortion rate of each phase is calculated, and then the arithmetic mean of the instantaneous distortion rate of each phase is taken as the output current distortion rate of the current sampling period. The output current distortion rate calculated for all sampling periods within a continuous time window is arranged sequentially in chronological order to generate the output current distortion rate sequence.

3. The low-voltage flexible DC power quality optimization method based on big data according to claim 1, characterized in that, The light-load oscillation risk coefficient is calculated based on the output current distortion rate sequence and the load rate sequence, including: The instantaneous load rate corresponding to each sampling period within a continuous time window is read from the local controller of each transformer area converter, and the instantaneous load rates are arranged in chronological order to generate the load rate sequence. Based on the distortion rate values ​​in the output current distortion rate sequence and the instantaneous load rate at the same sampling time, the initial risk component at each sampling time is determined; From each sampling time, the sampling time with a distortion rate value exceeding the benchmark distortion rate threshold is selected, the initial risk component of the corresponding sampling time is taken as the oscillation risk component, and the number of the selected sampling times is counted. The reference distortion rate threshold is determined based on the statistical distribution of the historical values ​​of the output current distortion rate of the target transformer area converter under stable operating conditions. The cumulative risk value is obtained by summing the oscillation risk components corresponding to all selected sampling times. The light load oscillation risk coefficient is calculated based on the cumulative risk value and the number of selected sampling times.

4. The low-voltage flexible DC power quality optimization method based on big data according to claim 1, characterized in that, The heavy-load voltage sag depth is calculated based on the integral value of the deviation between the DC bus voltage fluctuation sequence and the rated voltage, including: The rated voltage of the DC bus is obtained based on the DC side rated voltage specified in the factory nameplate parameters of each transformer area converter. The instantaneous DC bus voltage value at each sampling time in the DC bus voltage fluctuation sequence is compared with the rated voltage, and the instantaneous voltage deviation value at each sampling time is calculated. Wherein, when the instantaneous voltage value of the DC bus is lower than the rated voltage, the instantaneous voltage deviation value is the difference between the rated voltage and the instantaneous voltage value of the DC bus; otherwise, the instantaneous voltage deviation value is zero. The time intervals in which the instantaneous voltage deviation values ​​corresponding to multiple consecutive sampling times are all greater than zero are identified from the DC bus voltage fluctuation sequence and are used as voltage drop intervals. Within each voltage drop interval, the instantaneous voltage deviation value is multiplied by the sampling time interval between adjacent sampling points and then summed to obtain the voltage drop integral area corresponding to each voltage drop interval; The maximum value of the integral area of ​​voltage drop across all voltage drop intervals is selected as the heavy-load voltage drop depth.

5. The low-voltage flexible DC power quality optimization method based on big data according to claim 1, characterized in that, When the light-load oscillation risk coefficient exceeds a first threshold, the damping ratio attenuation trajectory under the same oscillation frequency mode is extracted from historical big data, and the dynamic compensation coefficient of the virtual impedance is calculated based on the damping ratio attenuation trajectory, including: Collect the light load oscillation risk coefficients of the target transformer converter during multiple consecutive time windows in the historical steady-state operation period, and determine the first threshold based on the statistical distribution of the light load oscillation risk coefficients; The currently calculated light load oscillation risk coefficient is compared with the first threshold. When the light load oscillation risk coefficient is greater than the first threshold, the operation of extracting the damping ratio attenuation trajectory and calculating the dynamic compensation coefficient is triggered. The current oscillation frequency is determined based on the zero-crossing characteristics of the continuous oscillation waveform in the DC bus voltage fluctuation sequence; Based on the current oscillation frequency, historical oscillation events with matching oscillation frequencies are retrieved from historical big data, and the envelope formed by the decay of the oscillation amplitude over time in each historical oscillation event is extracted as the damping ratio decay trajectory. Extract the peak amplitude corresponding to two adjacent oscillation periods from the damping ratio decay trajectory, and determine the logarithmic decay amount based on the ratio of the peak amplitudes; The reference damping ratio is calculated by combining the logarithmic attenuation with the time interval between adjacent peaks; Based on the ratio of the reference damping ratio to the pre-stored desired damping ratio, and in conjunction with the rated virtual impedance value, the dynamic compensation coefficient of the virtual impedance is calculated. The pre-stored expected damping ratio is determined based on the historical best damping ratio of the target transformer converter under rated load and no oscillation conditions, and the rated virtual impedance value is taken from the fixed impedance value in the droop control parameters.

6. The low-voltage flexible DC power quality optimization method based on big data according to claim 5, characterized in that, Based on the zero-crossing characteristics of the continuous oscillation waveform in the DC bus voltage fluctuation sequence, the current oscillation frequency is determined, including: Locate the oscillation start time from the DC bus voltage fluctuation sequence, and scan the instantaneous voltage values ​​sequentially starting from the oscillation start time; Record the sampling time when the instantaneous DC bus voltage changes from below the rated voltage to above the rated voltage, and use this as the positive zero-crossing point. Record at least three positive zero-crossing points consecutively. Calculate the number of sampling points between two adjacent positive zero crossings, and multiply the number of sampling points by the sampling time interval between adjacent sampling points to obtain the oscillation period value; The reciprocal of each oscillation period value is used as the instantaneous frequency value of the corresponding period to form a sequence of instantaneous frequency values; Abnormal fluctuation values ​​in the instantaneous frequency value sequence are removed, and the remaining instantaneous frequency values ​​are smoothed. The result of the smoothing is used as the current oscillation frequency.

7. The low-voltage flexible DC power quality optimization method based on big data according to claim 5, characterized in that, Based on the current oscillation frequency, historical oscillation events with matching oscillation frequencies are retrieved from historical big data. The envelope formed by the decay of the oscillation amplitude over time in each historical oscillation event is extracted as the damping ratio decay trajectory, including: With the current oscillation frequency as the center, extend half of the interquartile range of the historical oscillation frequency to both sides to set the frequency retrieval interval, wherein the interquartile range is the difference between the upper quartile and the lower quartile of the historical oscillation frequency. Traverse the oscillation events in historical big data, extract the oscillation frequency of each oscillation event, and mark the oscillation events whose oscillation frequencies fall within the frequency retrieval interval as matching events; For each matching event, read the corresponding historical sequence of DC bus voltage fluctuations; Identify the oscillation start point from the DC bus voltage fluctuation history sequence and extract the peak amplitude of each oscillation cycle in sequence; The extracted peak amplitudes are connected in chronological order according to the oscillation period. An envelope is formed with the oscillation period number as the horizontal axis and the peak amplitude as the vertical axis. This envelope is used as the damping ratio decay trajectory of the corresponding matching event.

8. The low-voltage flexible DC power quality optimization method based on big data according to claim 1, characterized in that, When the heavy-load voltage drop depth exceeds the second threshold, based on the ratio of the current change rate to the voltage recovery rate at the start of the drop, the recovery time constant under similar drop patterns is matched from the historical big data, and a correction amount for the voltage support coefficient is generated, including: The heavy load voltage sag depth of the target transformer converter during the historical steady-state operation period is collected for multiple consecutive time windows, and the second threshold is determined based on the statistical distribution of the heavy load voltage sag depth. The calculated overload voltage drop depth is compared with the second threshold. When the overload voltage drop depth is greater than the second threshold, the voltage drop start time is identified from the DC bus voltage fluctuation sequence, and the instantaneous output current value within one sampling period before and after the drop start time is obtained simultaneously. The rate of change of current is calculated based on the ratio of the difference between the instantaneous values ​​of the output current before and after the start of the drop to the corresponding time interval. The voltage recovery speed is calculated based on the rate of change of the instantaneous DC bus voltage value after the initial moment of the drop. Calculate the ratio of the current change rate to the voltage recovery rate, and use it as the characteristic value of the current drop pattern; Based on the current voltage drop pattern characteristic value, retrieve the historical voltage drop event with the smallest characteristic value deviation from the historical big data, and extract the time required for the voltage to recover from the lowest point of the voltage drop to a specified proportion of the rated voltage in the historical voltage drop event as the recovery time constant; The ratio of the reference recovery time constant to the recovery time constant is used as the correction amount for the voltage support coefficient, wherein the reference recovery time constant is the statistical average value of the time required for the voltage of the target transformer converter to naturally recover from the lowest point of voltage drop to the rated voltage under rated load conditions.

9. The low-voltage flexible DC power quality optimization method based on big data according to claim 1, characterized in that, The adaptive virtual impedance adjustment value is obtained by nonlinearly fusing the dynamic compensation coefficient and the correction amount, including: Adding the dynamic compensation coefficient to the benchmark compensation offset yields the normalized compensation coefficient, wherein the benchmark compensation offset is 1. The smaller value between the correction amount and the reference correction offset is taken as the normalized correction amount; wherein, the reference correction offset is 1. Calculate the square root of the normalized compensation coefficient to determine the nonlinear compensation factor; Multiplying the nonlinear compensation factor by the normalization correction amount yields the fusion factor; The fixed impedance value in the droop control parameters is adjusted according to the fusion factor to obtain the adaptive virtual impedance adjustment value.

10. A low-voltage flexible DC power quality optimization system based on big data, characterized in that, The system is used to execute the low-voltage flexible DC power quality optimization method based on big data as described in any one of claims 1-9, and the system comprises: The data acquisition module is deployed on the DC and AC sides of the converters in each distribution area to acquire the DC bus voltage fluctuation sequence and the output current distortion rate sequence within a continuous time window. The risk analysis module is used to calculate the light load oscillation risk coefficient based on the output current distortion rate sequence and the load rate sequence, and to calculate the heavy load voltage drop depth based on the integral value of the deviation between the DC bus voltage fluctuation sequence and the rated voltage. The compensation calculation module is used to extract the damping ratio decay trajectory under the same oscillation frequency mode from historical big data when the light load oscillation risk coefficient exceeds the first threshold, and calculate the dynamic compensation coefficient of the virtual impedance based on the damping ratio decay trajectory; and when the heavy load voltage drop depth exceeds the second threshold, it matches the recovery time constant under similar drop modes from the historical big data according to the ratio of the current change rate to the voltage recovery speed at the start of the drop, and generates the correction amount of the voltage support coefficient. A nonlinear fusion module is used to obtain an adaptive virtual impedance adjustment value by nonlinearly fusing the dynamic compensation coefficient and the correction amount. The instruction issuing module is used to replace the fixed impedance value in the droop control parameters of each transformer area converter with the adaptive virtual impedance adjustment value, generate optimized control instructions, and issue them to the pulse width modulator of each transformer area converter.

Images