A distributed photovoltaic grid-connected adaptability evaluation system

Abstract

The application discloses a kind of distributed photovoltaic grid-connected adaptability evaluation system, it is related to electric power system technical field, including benchmark modeling module, time-frequency analysis module, coupling identification module, capacity evaluation module, risk suppression module and closed-loop control module: benchmark modeling module, establish unified time baseline and unified phase reference field, meteorological data, user power generation curve data and power grid operation data are fused, extract local inherent frequency distribution, and determine zero-phase anchor point in phase reference field.The application constructs unified time baseline and phase reference, realizes multi-source data fusion and dynamic evaluation, identifies oscillation risk, accurately locks coupling node and frequency, fuses second-order cone programming and anti-resonance control, actively weakens coupling energy, finally through phase conjugate injection and feedback calibration, form high-precision closed-loop grid-connected adaptability regulation mechanism.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and more specifically to a distributed photovoltaic grid-connected adaptability assessment system. Background Technology

[0002] The Distributed Photovoltaic Grid-Connected Adaptability Assessment System is a comprehensive intelligent assessment platform for distributed photovoltaic (PV) power plant grid connection scenarios. Its core function is to determine the degree of compatibility between PV output and grid safety operation. The system integrates meteorological data, user-side power generation curves, and real-time grid operation data to construct a high-precision PV output prediction model, dynamically reflecting the fluctuation characteristics and trends of PV output. Based on this, an improved second-order cone programming (SOCP) algorithm is used to calculate the N-1 safety margin of distribution transformers and lines, accurately assessing the maximum connectable capacity at different connection points and avoiding operational risks such as grid voltage exceeding limits and line overload. When the system identifies a high-risk connection point, it automatically triggers an early warning and recommends an optimized "PV + energy storage" scheme based on actual load and output conditions to achieve peak shaving and capacity balance, thus forming a closed-loop grid-connected adaptability assessment mechanism covering prediction, assessment, early warning, and optimization.

[0003] The existing technology has the following shortcomings:

[0004] In existing technologies, distributed photovoltaic (PV) inverters typically rely on dynamic reactive power regulation to maintain stable grid-connected voltage. However, when the distributed PV capacity in a region is excessively high, multiple inverters may experience intermittent coupling with the local resonant frequency of the grid during reactive power regulation. This phenomenon has an extremely low probability of occurrence under normal operating conditions and is therefore often not fully considered in conventional evaluation methods. But once this coupling is triggered, it will cause voltage oscillations to amplify exponentially, and voltage disturbances will spread rapidly across the local grid, causing protection devices to operate frequently and triggering widespread equipment tripping, ultimately resulting in a large-scale concentrated disconnection of distributed power sources from the grid. Such sudden instability events not only disrupt the safe operation of the power grid but may also trigger cascading failures.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a distributed photovoltaic grid-connected adaptability assessment system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a distributed photovoltaic grid-connected adaptability assessment system, comprising a benchmark construction module, a time-frequency analysis module, a coupling identification module, a capacity assessment module, a risk suppression module, and a closed-loop control module:

[0008] The benchmark construction module constructs a unified time baseline and phase reference field based on the time synchronization device, integrates meteorological data, user power generation data and power grid operation data, extracts the regional natural frequency distribution, and determines the zero phase anchor point.

[0009] The time-frequency analysis module, under the constraint of zero-phase anchor point, collects the reactive power trajectory and voltage phasor timing of the inverter, performs reversible time-frequency rearrangement processing, extracts micro-amplitude oscillation characteristics, and constructs the resonance suspect spectrum;

[0010] The coupling identification module injects a micro-amplitude phase traction signal into the inverter based on the resonance suspicion spectrum, monitors the voltage response under a unified time baseline, and identifies the key coupling nodes and their corresponding natural frequencies.

[0011] The capacity assessment module, in conjunction with the identification results, embeds coupling constraints into an improved second-order cone programming model to calculate the N-1 safety margin of distribution transformers and lines, assess the upper limit of photovoltaic grid connection capacity, and generate high-risk time windows.

[0012] The risk suppression module performs virtual admittance shaping within high-risk time windows, constructs energy channel diversion paths, and generates anti-resonance feedforward commands in conjunction with energy storage control to establish a dynamic power transmission path with anti-phase compensation capabilities.

[0013] The closed-loop control module, based on the stability of the reverse-phase compensation path, performs time-inverse phase gating according to the inherent frequency, synchronously injects phase conjugate reactive power envelope signals into multiple inverters, forms an anti-channel with the energy storage energy, suppresses voltage oscillations in real time, and writes the regulation calibration factor into a unified time baseline to complete the closed-loop regulation process.

[0014] Preferably, the zero-phase anchor point determination process is as follows:

[0015] Deploy a time synchronization device to unify the clocks of multiple data acquisition channels at the meteorological acquisition end, photovoltaic power generation end, and power grid operation measurement end, and achieve time synchronization through a time synchronization module and time synchronization protocol;

[0016] Based on the time synchronization results, spatial location matching and time point alignment are performed on all time series data, a two-dimensional matrix is ​​constructed, and standardization and filtering are performed to complete the construction of a unified time baseline.

[0017] The local intrinsic frequency distribution of each spatial region is extracted based on the time baseline, and a short-time Fourier transform is performed using a sliding window. The dominant frequency set is confirmed based on the consistency of frequency peaks.

[0018] The voltage channel with the strongest steady-state performance was selected, and phase extraction and cross-validation were performed. The phase inflection point was calibrated as the zero-phase anchor point, which served as a unified reference frame for subsequent analysis.

[0019] Preferably, the steps for generating the resonance suspect spectrum are as follows:

[0020] Under the zero-phase anchor point constraint, the reactive power regulation trajectory and three-phase voltage phasor timing of the inverter are extracted, and the uniform interpolation and resampling are performed to a fixed sampling period and aligned with a unified time baseline.

[0021] A sliding window Fourier transform is performed based on the resampled data. The Blackman-Harris window function is used to suppress frequency leakage. The frequency domain energy is focused in the time domain through a reversible rearrangement method, and the waveform is reconstructed by performing an inverse Fourier transform.

[0022] The instantaneous envelope and oscillation segments are extracted from the reconstructed waveform to identify high-confidence time periods with oscillation amplitude characteristics and synchronous response characteristics.

[0023] The frequency components corresponding to the high-confidence time period are quantified and statistically analyzed, and a two-dimensional resonance suspected spectrum is plotted by combining energy density and reactive power trajectory to construct a linked expression of frequency and response characteristics.

[0024] Preferably, the process for identifying the coupling key node and its corresponding inherent frequency is as follows:

[0025] The target frequency is selected based on the frequency energy peak in the resonance suspect spectrum, and a micro-amplitude phase traction test signal with the same frequency as the target frequency and an amplitude less than one percent of the reactive power output capacity is injected into a single photovoltaic inverter.

[0026] Under a unified time baseline, voltage response data of all nodes are collected, and a three-dimensional response matrix is ​​constructed across nodes and phases. The energy response of the target frequency band is analyzed by short-time Fourier transform to identify the strong response node with the earliest, highest intensity and longest duration of energy change.

[0027] Frequency response feature comparison and path tracking analysis are performed on adjacent nodes in the topology path around the strong response node to screen out coupling nodes with frequency consistency higher than a preset threshold, and finally the key coupling trigger point and corresponding resonance frequency are locked.

[0028] Preferably, in the process of identifying strong response nodes, phase change curves within the same frequency range as the injected test signal are used for matching analysis. The matching index is the phase correlation coefficient, and the matching degree exceeding a set threshold is used as the criterion for determining the causal response relationship.

[0029] The preferred steps for generating a high-risk time window are as follows:

[0030] The inherent frequencies corresponding to key nodes are used as frequency coupling sensitive parameters and embedded into the dynamic operating parameters of lines and distribution transformers on the electrical topology path. They are then aligned with a unified time baseline to construct a complete time-space electrical parameter data body.

[0031] Based on this data volume, a second-order cone optimization model with frequency constraints is constructed. Current, voltage and frequency-related influencing factors are set as capacity constraints. The maximum connectable capacity is dynamically calculated and a safety margin of N minus one is retained.

[0032] The obtained capacity limit value is extended over time. Combined with the statistical trends of load rate, voltage level and reactive power regulation capability, the time period with significant capacity reduction is identified, marked as a high-risk time window, and a pre-control strategy template containing reactive power regulation trajectory is output.

[0033] Preferably, the steps for constructing a dynamic power transmission path are as follows:

[0034] Before the high-risk time window, a pre-adjustment period is set. Based on the frequency coupling path and key response nodes, the electrical parameters of the backup power channel are obtained. The reactive current phase angle of the inverter is adjusted to achieve admittance structure shaping and break the original frequency conduction channel.

[0035] After admittance shaping is completed, a path with energy storage and load regulation capabilities is selected as the shunt path based on electrical characteristics, and the main power is naturally transferred through phase shift and current guiding mechanisms.

[0036] The reserved capacity of the energy storage device is used to generate a feedforward control command with a reverse phase, which limits the frequency response range and output rate, forming a reverse energy wave at the target frequency to weaken frequency coupling.

[0037] Based on the energy storage output, the main inverter constructs an anti-phase compensation channel, precisely adjusts the phase angle of the reactive power output trajectory, forms a conjugate energy cancellation structure, monitors the electrical response of key nodes in real time and dynamically switches the compensation state, and completes the frequency stability control closed loop.

[0038] Preferably, based on the stability of the anti-phase compensation path, time-inverted phase gating is performed according to the natural frequency, and the phase conjugate reactive power envelope signal is injected synchronously to construct an anti-channel with the stored energy, suppressing voltage oscillations, and the calibration factor is written into the unified time baseline. The steps are as follows:

[0039] Retrieve the list of identified natural frequencies, select the frequency with the highest coupling strength and the longest historical oscillation duration, extract the corresponding phasor trajectory and perform time inversion, and generate the reverse control trajectory as the conjugate injection reference.

[0040] Based on the inversion trajectory, the injection tasks of multiple inverters are constructed, the injection frequency and phase parameters are set, and a unified time baseline is used as the start-up reference so that each inverter can synchronously complete the phase conjugate reactive power injection.

[0041] Monitor the frequency energy spectrum and phase angle change trends of key coupling nodes, compare the energy density decrease and waveform changes before and after injection, and judge the oscillation suppression effect of injection behavior;

[0042] Key response factors in the control process are extracted and written into the response data structure in a unified time baseline to form a frequency response profile for subsequent calls.

[0043] By combining data from six stages—prediction, identification, constraint, compensation, feedback, and calibration—regression analysis is performed on the entire frequency coupling process to construct a closed-loop control path with self-learning capabilities.

[0044] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0045] This invention achieves deep alignment and fusion analysis of multi-source heterogeneous data by constructing a unified time baseline and phase reference field, breaking through the time consistency limitations of traditional static evaluation methods. It introduces reversible time-frequency rearrangement and resonance suspect spectrum extraction techniques, enabling the first-ever early identification of latent oscillation signs. Through phase-driven probing and full-network voltage synchronous response analysis, it constructs a dynamic coupling evidence chain, accurately pinpointing key nodes and frequencies that trigger system resonance. Combined with an improved second-order cone programming model, the system can dynamically evaluate the N-1 safety margin and predict the capacity limit under frequency constraints, forming a quantified risk warning window. Furthermore, it introduces virtual admittance shaping, anti-resonance feedforward control, and dynamic energy channel construction techniques to actively reduce and reverse-cancel frequency coupling energy, effectively preventing instability propagation caused by coupling accumulation. Finally, through a time-inversion phase gating mechanism and a conjugate reactive power injection strategy, it achieves real-time and precise suppression of voltage oscillations, dynamically writing the response factor into the unified time baseline, forming a prediction-identification-evaluation-suppression-feedback closed-loop control chain. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0047] Figure 1 This is a schematic diagram of a distributed photovoltaic grid-connected adaptability assessment system according to the present invention. Detailed Implementation

[0048] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0049] This invention provides, for example Figure 1 The distributed photovoltaic grid-connected adaptability assessment system shown includes a benchmark construction module, a time-frequency analysis module, a coupling identification module, a capacity assessment module, a risk mitigation module, and a closed-loop control module.

[0050] The benchmark construction module establishes a unified time baseline and a unified phase reference field, integrates meteorological data, user-side power generation curve data and power grid operation data, extracts local natural frequency distributions based on the integrated data, and determines the zero-phase anchor point in the unified phase reference field as the time and phase reference benchmark for subsequent analysis.

[0051] To accurately identify potential oscillation risks in distributed photovoltaic grid-connected environments and prepare for subsequent dynamic intervention, it is necessary to integrate multiple types of raw sampling data under a unified time reference and a unified phase benchmark, extract the inherent frequency distribution characteristics of the region, and establish a precise zero-phase anchor point as the starting reference to achieve consistency between time-phase synchronization and frequency domain benchmark. The specific steps are as follows:

[0052] Time synchronization devices are deployed in the target power distribution area to unify the clocks of multiple data acquisition channels distributed across meteorological acquisition points, photovoltaic power generation points, and power grid operation measurement points. Meteorological data, including total solar radiation, direct radiation, wind speed, wind direction, atmospheric pressure, temperature, and humidity, is transmitted via RS485 bus, collected once per second, with timestamps provided by a GPS timing module. User-side power generation data, including voltage, current, active power, and reactive power, is collected from the AC / DC interfaces of each inverter, recorded at a sampling period of 500 milliseconds, and synchronized using the NTP protocol. Power grid-side operation data is provided by voltage sampling units and current transformers installed at feeder outlets and main distribution lines, with a sampling frequency set to 10kHz, and uploaded to a local data concentrator via the IEC 61850 message structure. After initial acquisition, the above data enters a time correction stage, using a distributed timing model to compensate for network latency, eliminate sampling errors and communication jitter, and ensure that all data achieves a time synchronization accuracy of no more than 2 milliseconds under an absolute clock reference, thus establishing a unified time baseline.

[0053] After establishing the time baseline, spatial location matching and time point alignment were performed on time-series data from different data sources. First, the distribution network was divided into multiple analysis regions based on geographical location, using substation feeder numbers, transformer numbers, and access node coordinates as grouping criteria. Time-series data within each region were mapped to spatial correspondences based on their corresponding numbers. Subsequently, all data streams were resampled according to a unified time baseline to ensure synchronization of data from different sources at millisecond intervals. To prevent imbalances in feature extraction caused by unit differences, variables in the meteorological data were range-scaled to map their range to between 0 and 1; Z-score standardization was applied to variables such as grid voltage, current, and power generation to ensure comparability across different units in subsequent analyses. The sampled data was organized into a two-dimensional matrix of "time step × number of data channels," with each matrix corresponding to a spatial region. Data channels included multiple dimensions such as temperature, irradiance, power generation, voltage, current, and frequency. After matrix construction, in order to reduce the impact of high-frequency noise, a third-order filter bank based on median filter is used to perform preliminary denoising on each data channel, filtering out spikes and jump values, while retaining edge slope change information to ensure the stability and accuracy of subsequent frequency domain extraction.

[0054] Based on a standardized time-series data matrix, a natural frequency extraction process is performed within each spatial analysis region. First, each data channel is truncated into a sliding window of 4 seconds with a 1-second step, and a short-time Fourier transform is performed within each window. The Fourier transform results form a time-frequency energy spectrum, where the horizontal axis represents the time step, the vertical axis represents the frequency components, and the amplitude represents the energy density. A peak detection algorithm is applied to each spectrum, marking frequency points with amplitudes exceeding three times the background mean as candidate frequency peaks. Then, a lateral frequency consistency check is performed on all candidate points; only those frequency positions that recur across multiple windows and whose power intensity fluctuations are no greater than 10% are considered valid natural frequencies. This process is repeated for each data channel, and frequency statistics and dominant frequency sorting are performed by region, ultimately obtaining a set of long-term stable dominant frequencies for each region. This set constitutes the local natural frequency distribution of the region, accurately reflecting the electrical dynamic characteristics of the region under actual operating conditions.

[0055] Based on the dominant frequency set for each spatial region, a zero-phase anchor point location operation is performed. First, the channel with the strongest steady-state performance is selected from the voltage phasor data within that region. Specific selection criteria include: no abrupt changes in the data, a standard deviation of fluctuation less than 2%, and an average frequency fluctuation less than 0.1 Hz. Within the selected voltage channel, the waveform signal at the dominant frequency is subjected to a Hilbert transform to extract the instantaneous phase sequence, constructing a phase-time curve. From this phase-time sequence, the moment when the first phase transitions from negative to positive is found and marked as a candidate zero-phase point. To ensure consistency of this candidate point across different channels, and by performing the same phase detection in both power generation and current data, a cross-validation process is constructed. The standard deviation of the phase difference at this point across channels is calculated; if it is less than 5 degrees, it is considered a valid zero-phase anchor point. Finally, this time point is recorded as the unified starting reference frame for all subsequent trajectory analysis, frequency identification, traction signal injection, and other operations, achieving absolute phase consistency and precise temporal alignment.

[0056] The time-frequency analysis module, under the constraint of the zero-phase anchor point, obtains the reactive power regulation trajectory and voltage phasor timing of the distributed photovoltaic inverter, performs a reversible time-frequency rearrangement operation based on a unified time baseline, extracts micro-amplitude oscillation energy from the rearrangement result and reconstructs waveform details to generate a resonance suspect spectrum characterizing potential oscillation risks.

[0057] To accurately identify minute power disturbances in distributed photovoltaic inverters during grid-connected operation, it is necessary to extract the inverter's reactive power regulation trajectory and three-phase voltage phasor timing under the constraints of an established zero-phase anchor point and a unified time baseline. High-resolution reversible time-frequency rearrangement processing is then used to complete detailed reconstruction and oscillation risk feature extraction. The specific steps are as follows:

[0058] After establishing the zero-phase anchor point, a continuous data segment with this anchor point as the starting reference is selected, with a time length of 15 seconds to ensure coverage of multiple typical inverter control cycles. Within this time period, the three-phase reactive power adjustment trajectory is acquired from the inverter's AC output terminal with an accuracy of 0.01 kvar and a sampling period not exceeding 5 milliseconds. Simultaneously, the three-phase voltage waveform is extracted from the synchronous measurement unit installed at the bus voltage acquisition location, recording the amplitude and phase angle for each phase separately. The amplitude unit is volts with an accuracy requirement of 0.1 volts, and the phase angle unit is degrees with an accuracy controlled within 0.1 degrees. All sampled data undergoes interpolation and resampling immediately after recording, uniformly converting to a fixed sampling period of 1 millisecond. The interpolation method uses a third-order interpolation algorithm based on spline curves to improve transition smoothness on short time scales. Subsequently, by calibrating the timestamps and aligning them with the previously established unified time baseline, all time-series data are strictly synchronized to the zero-phase anchor point, constructing a dual-channel synchronous dataset with a fixed time length and completely consistent data points, providing a unified time reference for subsequent waveform analysis.

[0059] Sliding window Fourier transforms were performed on the dual-channel dataset to extract the frequency domain structure. Each data sequence was segmented into windows of 1024 data points each, with a window sliding step size of 128 points. The Blackman-Harris window function was used to suppress edge frequency leakage. After the transform, the amplitude and phase spectra of each time slice were obtained, forming a two-dimensional time-frequency graph. To enhance the expressive power of the small-amplitude oscillation components in the signal, a reversible rearrangement method based on energy density weighting was introduced to move high-frequency, high-energy regions to both ends of the time axis, allowing small-energy frequency bands that were originally suppressed in the background to be concentrated in the time domain. This rearrangement process ensures that the amplitude and phase information maintain a reversible structure throughout, and the original signal can be completely restored through the reverse rearrangement operation. The processed frequency domain data was then returned to the time domain by an inverse Fourier transform, resulting in two sets of rearranged waveform sequences. The energy features in these sequences have been refocused into their respective significant regions according to the frequency distribution, effectively enhancing the sensitivity of subsequent small perturbation identification.

[0060] The Blackman-Harris window function is a windowing function with excellent sidelobe suppression performance, commonly used in spectral analysis in signal processing, particularly suitable for extracting weak frequency components from time series. In this invention, the Blackman-Harris window function is used for weighting operations when performing sliding window segmentation on reactive power trajectories and voltage phasor time series data. Its core function is to effectively reduce spectral leakage caused by window edge effects by assigning smaller weights to edge sampling points and larger weights to center points in each data segment. Spectral leakage refers to the phenomenon where a single frequency component in the original signal appears at multiple frequency positions after transformation. Especially when the target signal contains weak periodic disturbances, the leakage effect causes these weak frequencies to be submerged in background noise and difficult to distinguish. Compared to common rectangular windows, Hamming windows, or Hanning windows, the Blackman-Harris window function achieves a better balance between main lobe width and sidelobe attenuation, with a typical sidelobe suppression capability of -92 dB, effectively improving the resolution of the main spectral peak and the ability to identify small oscillation components. Therefore, the use of the Blackman-Harris window function in the frequency domain analysis process of this invention helps to enhance the positioning accuracy of the oscillation frequency and provides a cleaner and clearer initial spectral structure for subsequent reversible time-frequency rearrangement and resonance suspect spectrum construction, which is one of the key basic steps to achieve high-sensitivity frequency extraction.

[0061] In the reconstructed waveform, amplitude fluctuation analysis is performed on the voltage phasor signal, and synchronous response verification is conducted on the reactive power trajectory to identify time periods with potential resonance characteristics. Specifically, the envelope of the rearranged voltage waveform is extracted to identify local oscillations in its amplitude. The instantaneous amplitude is obtained through a Fast Hilbert Transform, and a complete envelope curve is constructed. In this curve, slowly changing segments with continuous amplitude changes less than 0.5% are removed, retaining only periodic fluctuation segments with amplitude changes exceeding 1% and continuous occurrences exceeding 300 milliseconds as preliminary suspect segments. Subsequently, these preliminary suspect segments are mapped one-to-one with the reactive power trajectory on the time axis to analyze whether there is a significant phase correlation, i.e., whether a synchronous response with the same direction occurs in the reactive power within 30 milliseconds after the voltage change. If the response is valid, the time period is further marked as a high-confidence oscillation segment. After all channels are marked, all time periods in the reconstructed data that meet the above conditions are summarized and their corresponding frequency, oscillation period, phase start point, response lag time, and energy density are recorded.

[0062] Based on high-confidence oscillation segments, a suspected resonance spectrum was constructed within a unified time baseline framework. The spectrum construction process is as follows: First, the frequency components of all marked segments were extracted and uniformly quantized into 0.1Hz frequency intervals. The frequency occurrence and average energy density within each frequency interval were statistically analyzed as evaluation indicators of frequency confidence. Next, on a two-dimensional image coordinate system with time as the horizontal axis and frequency as the vertical axis, the energy density of each frequency interval was plotted point by point as a color intensity, forming a complete suspected resonance spectrum. This spectrum not only intuitively reflects the occurrence time, oscillation intensity, and duration of each frequency component under a unified time series, but also stratifies their degree of danger through color depth. In addition, a trend line of reactive power response trajectory change was simultaneously superimposed on the spectrum, constituting a visual verification result of the time-series linkage relationship, significantly enhancing the application value of the spectrum in subsequent dynamic coupling analysis and trigger point location.

[0063] The coupling identification module constructs a dynamic coupling evidence chain based on the resonance suspect spectrum. It injects a micro-amplitude phase traction test signal into a single distributed photovoltaic inverter and synchronously monitors the voltage response of each node under a unified time baseline. Based on the response results, it identifies the key nodes that cause grid coupling and their corresponding natural frequencies.

[0064] To further verify whether the oscillation frequency components identified in the suspected resonance spectrum possess a genuine network coupling effect, it is necessary to actively apply a perturbation signal to a specific photovoltaic inverter based on a unified time baseline and a zero-phase anchor point, and to perform time-domain tracking of the voltage response of all network nodes to establish a causal response relationship, thereby identifying the key nodes that truly constitute the oscillation source and their coupling paths. The specific steps are as follows:

[0065] In the constructed resonance suspect spectrum, the dominant frequency with the highest energy peak was selected as the target frequency for this analysis. This frequency must exhibit energy enhancement in multiple measurement channels and have a stable frequency position and periodic characteristics in the previous time-frequency rearrangement results. For this target frequency, a set of probing signals with a small phase offset was designed. The signal type was a phase-modulated sine wave with a frequency completely consistent with the target frequency. The amplitude was set between 0.5% and 1% of the inverter's current reactive power output capacity, the phase disturbance amplitude was controlled within 1 / 16 of the main period, and the duration was controlled between 300 and 500 milliseconds. This signal was controlled by applying a phase offset command to the reactive power setting channel of the target inverter. It did not change the voltage amplitude or adjust the current peak value; it only generated a small disturbance by adjusting the phase angle of the output current, thus introducing a trackable frequency disturbance source without affecting the overall stable operation of the power grid. The start time of this probing signal was set to an integer second under a unified time baseline to ensure that all response data had a unified time reference in subsequent analyses.

[0066] After the probe signal is injected, a full voltage response monitoring process is initiated, acquiring three-phase voltage waveform data from all nodes and recording their amplitude, frequency, and phase angle parameters in real time. The sampling period is fixed at once per millisecond, and the total data acquisition time is 1000 milliseconds after signal injection, covering the entire process of disturbance occurrence, propagation, and attenuation. The acquired data is immediately aligned, and a three-dimensional response matrix spanning nodes and phases is constructed with the starting time of the probe signal application as the time zero point. Data from each node constitutes a response channel. Subsequently, a sliding window short-time Fourier analysis is performed on each response channel. The sliding window length is set to 64 milliseconds, the step size is 16 milliseconds, and the Kaiser window is used to enhance frequency resolution. For the spectrum within each time window, the energy change trend of the target frequency band is extracted, and the start and peak times of frequency energy abrupt changes are recorded. All nodes are ranked based on three evaluation dimensions: earliest response, highest energy peak, and longest duration. The strongest response node that simultaneously meets all three conditions is selected. Subsequently, bandpass filtering is performed on the original voltage waveform at this node to retain frequency components within ±0.2Hz of the target frequency. The phase change curve of this frequency band is then calculated and compared with the waveform of the test signal to further verify whether a causal relationship exists in the response. If the matching degree exceeds a preset threshold (e.g., a Pearson correlation coefficient greater than 0.8), the node can be determined to be a strongly coupled response point at the target frequency.

[0067] After identifying nodes with high response characteristics, the electrical topology path is traced back from these nodes, and adjacent upstream and downstream nodes are included in the analysis to expand the observation area. Response analysis is performed on each node within this area at the same frequency band, recording its phase response time, energy density change rate, and response direction trend, and the difference is calculated with the main response node. Based on the order of response start time, frequency consistency, and relative magnitude of fluctuations, a propagation chain structure is established between nodes, depicting the energy diffusion path of the disturbance from the application point to the high-response node. For the response frequencies of all nodes in the path, their average value is taken and compared with the previously identified inherent frequency distribution of the region. If the error is within 0.3 Hz, this frequency is designated as the dynamic coupling frequency under this path. Simultaneously, to eliminate randomness, the trial signal application operation is repeated at least three times to ensure that similar response trajectories are observed at the same node each time. If the conditions are met, this frequency is finally recorded as the resonant frequency of the path, and the node with the strongest response is marked as the key coupling trigger point, providing a target location for subsequent risk mitigation and dynamic compensation steps.

[0068] The capacity assessment module, under the condition that the key nodes and inherent frequencies are identified, embeds the coupling constraint parameters into the improved second-order cone programming model, performs distribution network operation status assessment under the condition of unified time baseline, calculates the N-1 safety margin of distribution transformers and lines, outputs the upper limit of the photovoltaic system's access capacity, and generates a high-risk time window containing reactive power trajectory pre-control information.

[0069] To achieve safe operation analysis of the distribution network based on identified key nodes and inherent frequencies, an optimization model incorporating frequency coupling factors needs to be constructed under a unified time baseline constraint. This model dynamically calculates the upper limit of the connected capacity and identifies high-risk operating periods, ensuring the stability and foresight of photovoltaic grid connection. The specific steps are as follows:

[0070] Key nodes and their corresponding natural frequencies are used as strong constraint parameters in the electrical network and embedded into the parameter initialization process of the dynamic optimization model. Specifically, based on the key node identifiers obtained in the previous stage, an absolute position index of the node in the topology is established, and a complete topology path is extracted, centered on the node, extending upwards to the main bus and downwards to the end user access point. Along this path, the rated current, current load current, conductor type, resistance and reactance parameters, thermal stability limits, and current safety margin limits of each line are extracted. All data must come from real-time monitoring results within a continuous operating cycle and be synchronized point-by-point with a unified time baseline. For each distribution transformer associated with this path, parameters such as transformer capacity, active and reactive load rates, cooling status, compensation device operating status, voltage regulation level, and load regulation response delay are extracted, and their behavioral changes are traced back according to a time series. Subsequently, the above line and equipment parameters are uniformly mapped onto a time axis with a time step of 1 second, forming a three-dimensional data volume of time, space, and electrical parameters. Based on this, the frequency response intensity, test signal response time difference, and response energy change rate corresponding to the region where the key node is located are used as dynamic influencing factors and mapped to frequency coupling sensitive parameters within the network, thus setting dynamic boundary constraints for subsequent models.

[0071] Based on a complete data structure, a second-order cone optimization structure with frequency constraints is constructed to perform dynamic capacity boundary assessment. This structure uses each distributed photovoltaic inverter connection point as the main variable, aiming to maximize the grid-connectable photovoltaic capacity while ensuring line current safety, voltage compliance, and controllable frequency coupling risks. The specific solution steps are as follows: The current limit of each feeder is set as a hard constraint; the node voltage deviation is controlled within ±5% as the voltage stability target; the frequency sensitivity of each node on the identified coupling path is used as a flexible constraint; and a dynamic adjustment factor is constructed so that the capacity limit curve of this path automatically shrinks with frequency fluctuations. For critical paths with significant frequency response, a dynamic voltage drop mechanism positively correlated with the frequency energy amplitude is introduced into their capacity limit boundary. When the frequency disturbance energy increases, the grid-connected capacity allowable value is automatically reduced. This optimization model adopts an iterative calculation strategy, starting from the current time step, successively increasing the photovoltaic injection power, and calculating line power flow, voltage distribution, and frequency-related influencing factors in real time. Once any constraint is triggered, it is considered a capacity limit point. In each iteration, the model retains the minimum safety margin value of each path. Compared with the traditional N-1 safety check condition, it performs post-fault power flow redistribution calculation for each unit fault scenario to ensure that the required capacity can still maintain electrical stability in the event that any line or transformer is out of service.

[0072] After obtaining the photovoltaic grid-connected capacity limit, this limit is further extended over time to analyze its trend changes with load variations, voltage response, and reactive power regulation capability fluctuations, identifying high-risk operating periods and generating early warning windows. The specific process is as follows: Using a unified time baseline as a reference, sliding statistics are performed on transformer load rate, voltage level, and inverter reactive power output capability. Each calculation covers a 3-minute data window with a step size of 30 seconds. During the statistical process, if the transformer load rate increases by more than 90% of the rated capacity, the node voltage drop exceeds 3%, or the reactive power regulation curve changes by more than ±5% within 1 minute, the window is marked as a potentially high-risk period. Subsequently, capacity backtracking checks are performed on these periods to determine whether the capacity limit has decreased by more than 15% under the identified frequency response conditions. If so, the period is precisely defined as a high-risk time window. For each high-risk time window, the reactive power output trajectory, node voltage change curve, frequency response amplitude, and trend of all inverters within that period are exported to establish a dynamic behavior sequence. Based on this sequence, a pre-control strategy template is constructed. The template clearly indicates the reactive power adjustment range that needs to be intervened in advance at future time points, the allowable response delay duration, the frequency range to be prioritized for suppression, and the expected response target of the energy storage device. This provides a priori control basis for subsequent anti-harmonic control and energy channel shaping operations.

[0073] The risk suppression module performs virtual admittance shaping operations and constructs a diversion path for the grid energy channel during high-risk time windows. It generates corresponding anti-resonance feedforward control commands in conjunction with energy storage devices, actively suppresses the accumulation process of frequency coupling based on the feedforward control commands, and constructs a dynamic power transmission path with anti-phase compensation capability.

[0074] To avoid the exponential amplification effect of frequency-energy coupling within identified high-risk time windows, proactive intervention is necessary. This involves adjusting the grid admittance structure, guiding power diversion, generating feedforward suppression commands, and constructing a dynamically adjustable anti-phase compensation path to form a frequency stability control closed loop. The specific steps are as follows:

[0075] Based on the identified frequency coupling paths and critical response nodes, a pre-adjustment period with a lead time of 20 seconds is set before the start of the high-risk time window to perform admittance shaping operations to break the original frequency conduction channels. The specific process is as follows: First, all backup power transmission channels in the electrical topology that are parallel to the critical path but have not yet formed a frequency amplification response are acquired. Their equivalent resistance, equivalent inductance, and line-to-line capacitive reactance parameters are extracted, and complex admittance values ​​are calculated. The admittances of each path are dynamically reweighted along the time dimension to select admittance combinations that can maintain low reflection characteristics under high-frequency disturbances. Subsequently, the inverter's reactive power output direction is guided to shift slightly in the control command, actively turning towards the target admittance direction. In physical implementation, this operation injects a small-amplitude phase step control command into the inverter's phase-locked controller, adjusting the vector direction of the reactive power output current without changing the active power, thereby reducing the principal admittance value of the original main coupling channel and forcing some current to be diverted through the new channel, thus artificially reshaping the admittance network structure and achieving pre-intervention of the frequency response path.

[0076] After completing the admittance structure shaping, the electrical characteristics of the physical path from the main power node to the adjacent load node are evaluated to determine the appropriate route for constructing the energy transfer channel. Paths with the following characteristics are selected as the shunt paths: at least one bidirectional converter-type energy storage device is installed on the path; the thermal stability limits of all lines have a usable margin of no less than 30% under the current load level; no abnormal frequency amplification events have occurred at any intermediate node on the path in past operating cycles; and the area where the terminal node is located has local load regulation capabilities. After selecting the path, a dynamic energy guiding mechanism is initiated between the main inverter and the energy storage node. A low-amplitude current guiding process is pre-established when the energy storage device is charging, causing the power factor to gradually shift towards the lagging direction, increasing the current-induced voltage level of the energy storage node, and triggering a voltage drop, thereby attracting a portion of the power output from the main inverter to naturally transfer to this path. To ensure that no new frequency amplification effect is formed after the path is turned on, the frequency response gain curve of each node in the path is continuously monitored during the transfer process. If a second-order or higher fast response occurs, the guiding amplitude is immediately reduced or the transfer process is briefly interrupted to ensure path stability.

[0077] Under the premise of stable operation of the shunt path, the reserved capacity in the energy storage device is invoked to generate a reverse-suppression feedforward control command centered on the target coupling frequency. This command has its output frequency range, response phase angle, and output slope parameters pre-set by the power dispatch controller. In the control logic, the energy storage device is specified to form a single-peak frequency response wave within the target frequency ±0.5Hz range at an output rate not exceeding 20% ​​of the maximum discharge power. The phase angle of the command is set to a 180-degree phase reversal with reference to the inverter output waveform, ensuring reverse energy superposition with the main coupling path in the frequency domain. The output duration is set from 5 seconds before the start of the high-risk time window to 2 seconds after the end of the window. To ensure the energy concentration characteristics of the control signal, a bandpass filter is introduced in the energy storage output control loop, allowing only signals in the target frequency band to pass through, preventing the injection of invalid power at non-target frequencies. During control execution, the voltage and current phase trajectories of the energy storage output are acquired in real time and synchronously compared with the frequency response characteristics of the target node to analyze the energy cancellation ratio between the two. If the peak energy of the response signal decreases by more than 30%, the feedforward control is considered effective; otherwise, the phase angle and output curve are finely adjusted in real time according to the feedback adjustment mechanism.

[0078] After the energy storage device completes its feedforward output, the main inverter begins to collaboratively construct an anti-phase compensation channel, forming a highly efficient dynamic energy cancellation structure. This structure precisely adjusts the phase angle path of the main inverter's reactive power output trajectory, causing it to generate a composite conjugate phase characteristic with the energy storage output waveform at the target frequency, forming a pair of opposite energy propagation waves. Specifically, the operation involves: acquiring the current output frequency envelope of the main inverter and its phase difference with the energy storage output; calculating the required phase angle deviation every 10 milliseconds; and inputting this amount to the inverter's internal controller for micro-step adjustment. Each step must not exceed 1.5 degrees to prevent discontinuity in the main waveform. After the compensation channel is formed, the voltage phasor change rate, frequency drift amplitude, and remaining energy density of the coupling path at key nodes are monitored in real time. This data is then compared with the original coupling characteristic data using a sliding comparison analysis to identify the suppression effect. After identifying stable suppression, the compensation channel enters a low-frequency maintenance state, maintaining only a small output for field stability control. If the frequency disturbance increases again, it automatically returns to a high-energy output state, forming a closed-loop operation characteristic.

[0079] Under the condition of stable operation of the dynamic power transmission path, the closed-loop control module performs a time-reverse phase gating process based on the identified inherent frequency parameters. It synchronously injects the reactive power envelope signal of phase conjugate into multiple distributed photovoltaic inverters, so that the envelope signal and the output energy of the energy storage device form an anti-energy channel to achieve real-time suppression of voltage oscillation. The calibration factor obtained in the regulation process is written into the unified time baseline, thereby completing the closed-loop evaluation process of predicting, identifying, constraining, compensating and dynamically regulating grid coupling risks.

[0080] To complete the final closed-loop stage of grid frequency coupling control, based on the stable dynamic power transmission path described above, time inversion and phase conjugate control are performed on the identified inherent frequency characteristics. This enables multi-inverter coordinated injection, real-time cancellation of oscillation waveforms, and, after execution, the key response factors are written into a unified time baseline to form a continuously optimizable control closed loop. The specific steps are as follows:

[0081] Under the premise of stable operation of the frequency response path, the list of inherent frequencies identified in the previous stage is retrieved, and the time-domain response characteristics of each frequency point are classified and filtered. The target frequency with the highest coupling strength and the longest historical oscillation duration is selected as the current master control object. Subsequently, based on a unified time baseline as the time coordinate, a complete voltage phasor sequence is extracted from the historical oscillation data corresponding to this frequency, and the sequence is converted into a complex phase trajectory. Next, the time axis of this trajectory is reversed to construct a theoretically applicable inverse control waveform for oscillation cancellation, and its corresponding phase trajectory change curve is extracted. This phase inversion trajectory serves as a reference template for the control input of multiple inverters, and is used for subsequent phase conjugate reactive power injection operations.

[0082] Based on parameters such as frequency, phase, and duration of the inversion trajectory, an injection task is constructed for each distributed photovoltaic inverter in a stable operating state with injection capability. This task uses the inverter's current reactive power output capacity as a reference, setting the maximum injection envelope to no more than 20% of its rated capacity, ensuring the injection frequency matches the target frequency, and controlling the phase difference between the injection phase and the inversion trajectory within 180 degrees ± 5 degrees. In actual operation, the waveform of the reactive current injection is gradually phase-shifted by controlling the inverter's internal current phase angle adjustment mechanism, enabling it to construct a stable anti-phase envelope at the target frequency. To ensure the synchronization of signals across multiple inverters, all injection tasks are triggered using a unified time baseline and are completed within 10 milliseconds after the main control command is issued, avoiding phase cancellation failures due to injection time differences. During execution, the system continuously monitors whether the output power, phase angle change rate, and injection frequency of each inverter remain within the set range, triggering a temporary pause and reconfiguration when the set tolerance is exceeded.

[0083] After the injection process of all inverters has been stably executed, the voltage phasors and frequency energy spectra of multiple key coupled nodes in the power grid are continuously observed. By comparing the energy density changes of each node at the target frequency with those before the injection begins, if the energy decrease rate is found to be greater than 30%, and the waveform envelope shape transitions from multi-peak to single-peak or tends to flatten, it is considered that the energy at the current frequency has been partially or completely offset. At the same time, if there are obvious phase lag amplitude contractions and frequency oscillation period extensions on the voltage phase angle change trend curve, it can also be considered that the conjugate energy injection has had a substantial suppressive effect on the original oscillation behavior. The duration of this process is generally set to more than three cycles of the target frequency to ensure that the effect of the control behavior is not only a short-term intervention result, but also a dynamic control based on stable and continuous injection.

[0084] After achieving significant control effectiveness, key adjustment factors in the control process are extracted, generating a write-back response data structure, which is then linked to a unified time baseline to enable subsequent prediction, pre-control, and compensation based on historical factors. The extracted key factors include: the percentage decrease in target frequency response amplitude, inverter injection start time and duration, maximum phase conjugate error, energy decrease slope, minimum voltage phase angle fluctuation bandwidth, and injection effect duration. This data is indexed using time and frequency labels to establish a multi-dimensional relationship, and written into the unified time baseline data structure, forming a response profile that can be directly accessed by subsequent control strategies. This response profile has a clear application scenario, traceability, and parameter replicability, serving as crucial support for achieving closed-loop optimization.

[0085] After the control response data is successfully written, a structural closed-loop regression analysis is performed on the entire frequency coupling control process, combining the original identification data, control input data, and response results. This analysis uses a unified time baseline as the coordinate axis and constructs a complete six-stage chain: "prediction—identification—constraint—compensation—feedback—calibration." The output data of each stage is fed back to the previous stage to form a self-learning path. Specifically: when oscillation characteristics similar to the current frequency appear in the future prediction stage, the previous control path can be recalled in advance, setting an early start time; in the identification stage, the response lag factor is used as an auxiliary variable to improve identification accuracy; in the constraint stage, a more reasonable dynamic boundary is set based on the previous injected power limit; in the compensation stage, the injected phase angle offset speed is adjusted with reference to the conjugate error empirical value; in the feedback stage, the judgment threshold is corrected using known control results; and in the calibration stage, new response data is continuously added to the time baseline structure. Through this process, the power grid frequency response control transforms from a passive response to an active evolution, ultimately forming an intelligent closed-loop control mechanism oriented towards oscillation risks.

[0086] This invention achieves deep alignment and fusion analysis of multi-source heterogeneous data by constructing a unified time baseline and phase reference field, breaking through the time consistency limitations of traditional static evaluation methods. It introduces reversible time-frequency rearrangement and resonance suspect spectrum extraction techniques, enabling the first-ever early identification of latent oscillation signs. Through phase-driven probing and full-network voltage synchronous response analysis, it constructs a dynamic coupling evidence chain, accurately pinpointing key nodes and frequencies that trigger system resonance. Combined with an improved second-order cone programming model, the system can dynamically evaluate the N-1 safety margin and predict the capacity limit under frequency constraints, forming a quantified risk warning window. Furthermore, it introduces virtual admittance shaping, anti-resonance feedforward control, and dynamic energy channel construction techniques to actively reduce and reverse-cancel frequency coupling energy, effectively preventing instability propagation caused by coupling accumulation. Finally, through a time-inversion phase gating mechanism and a conjugate reactive power injection strategy, it achieves real-time and precise suppression of voltage oscillations, dynamically writing the response factor into the unified time baseline, forming a prediction-identification-evaluation-suppression-feedback closed-loop control chain.

[0087] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A distributed photovoltaic grid-connected adaptability assessment system, characterized in that, It includes a benchmark construction module, a time-frequency analysis module, a coupling identification module, a capacity assessment module, a risk mitigation module, and a closed-loop control module: The benchmark construction module constructs a unified time baseline and phase reference field based on the time synchronization device, integrates meteorological data, user power generation data and power grid operation data, extracts the regional natural frequency distribution, and determines the zero phase anchor point. The time-frequency analysis module, under the constraint of zero-phase anchor point, collects the reactive power trajectory and voltage phasor timing of the inverter, performs reversible time-frequency rearrangement processing, extracts micro-amplitude oscillation characteristics, and constructs the resonance suspect spectrum; The coupling identification module injects a micro-amplitude phase traction signal into the inverter based on the resonance suspicion spectrum, monitors the voltage response under a unified time baseline, and identifies the key coupling nodes and their corresponding natural frequencies. The capacity assessment module, in conjunction with the identification results, embeds coupling constraints into an improved second-order cone programming model to calculate the N-1 safety margin of distribution transformers and lines, assess the upper limit of photovoltaic grid connection capacity, and generate high-risk time windows. The risk suppression module performs virtual admittance shaping within high-risk time windows, constructs energy channel diversion paths, and generates anti-resonance feedforward commands in conjunction with energy storage control to establish dynamic power transmission paths. The closed-loop control module, based on the stability of the reverse compensation path, performs time-inverse phase gating according to the inherent frequency, synchronously injects the phase conjugate reactive power envelope signal into multiple inverters, forms a reverse channel with the energy storage energy, suppresses voltage oscillation in real time, and writes the regulation calibration factor into the unified time baseline.

2. The distributed photovoltaic grid-connected adaptability assessment system according to claim 1, characterized in that, The process for determining the zero-phase anchor point is as follows: Deploy a time synchronization device to unify the clocks of multiple data acquisition channels at the meteorological acquisition end, photovoltaic power generation end, and power grid operation measurement end, and achieve time synchronization through a time synchronization module and time synchronization protocol; Based on the time synchronization results, spatial location matching and time point alignment are performed on all time series data, a two-dimensional matrix is ​​constructed, and standardization and filtering are performed to complete the construction of a unified time baseline. The local intrinsic frequency distribution of each spatial region is extracted based on the time baseline, and a short-time Fourier transform is performed using a sliding window. The dominant frequency set is confirmed based on the consistency of frequency peaks. The voltage channel with the strongest steady-state performance was selected, and phase extraction and cross-validation were performed. The phase inflection point was calibrated as the zero-phase anchor point, which served as a unified reference frame for subsequent analysis.

3. The distributed photovoltaic grid-connected adaptability assessment system according to claim 2, characterized in that, The steps for generating the suspected resonance spectrum are as follows: Under the zero-phase anchor point constraint, the reactive power regulation trajectory and three-phase voltage phasor timing of the inverter are extracted, and the uniform interpolation and resampling are performed to a fixed sampling period and aligned with a unified time baseline. A sliding window Fourier transform is performed based on the resampled data. The Blackman-Harris window function is used to suppress frequency leakage. The frequency domain energy is focused in the time domain through a reversible rearrangement method, and the waveform is reconstructed by performing an inverse Fourier transform. The instantaneous envelope and oscillation segments are extracted from the reconstructed waveform to identify high-confidence time periods with oscillation amplitude characteristics and synchronous response characteristics. The frequency components corresponding to the high-confidence time period are quantified and statistically analyzed, and a two-dimensional resonance suspected spectrum is plotted by combining energy density and reactive power trajectory to construct a linked expression of frequency and response characteristics.

4. The distributed photovoltaic grid-connected adaptability assessment system according to claim 3, characterized in that, The process of identifying key coupling nodes and their corresponding natural frequencies is as follows: The target frequency is selected based on the frequency energy peak in the resonance suspect spectrum, and a micro-amplitude phase traction test signal with the same frequency as the target frequency and an amplitude less than one percent of the reactive power output capacity is injected into a single photovoltaic inverter. Under a unified time baseline, voltage response data of all nodes are collected, and a three-dimensional response matrix is ​​constructed across nodes and phases. The energy response of the target frequency band is analyzed by short-time Fourier transform to identify the strong response node with the earliest, highest intensity and longest duration of energy change. Frequency response feature comparison and path tracking analysis are performed on adjacent nodes in the topology path around the strong response node to screen out coupling nodes with frequency consistency higher than a preset threshold, and finally the key coupling trigger point and corresponding resonance frequency are locked.

5. The distributed photovoltaic grid-connected adaptability assessment system according to claim 4, characterized in that, In the process of identifying strong response nodes, phase change curves within the same frequency range as the injected test signal are used for matching analysis. The matching index is the phase correlation coefficient, and the matching degree exceeding the set threshold is used as the criterion for determining the causal response relationship.

6. The distributed photovoltaic grid-connected adaptability assessment system according to claim 4, characterized in that, The steps for generating a high-risk time window are as follows: The inherent frequencies corresponding to key nodes are used as frequency coupling sensitive parameters and embedded into the dynamic operating parameters of lines and distribution transformers on the electrical topology path. They are then aligned with a unified time baseline to construct a complete time-space electrical parameter data body. Based on this data volume, a second-order cone optimization model with frequency constraints is constructed. Current, voltage and frequency-related influencing factors are set as capacity constraints. The maximum connectable capacity is dynamically calculated and a safety margin of N minus one is retained. The obtained capacity limit value is extended over time. Combined with the statistical trends of load rate, voltage level and reactive power regulation capability, the time period with significant capacity reduction is identified, marked as a high-risk time window, and a pre-control strategy template containing reactive power regulation trajectory is output.

7. The distributed photovoltaic grid-connected adaptability assessment system according to claim 6, characterized in that, The steps for constructing a dynamic power transmission path are as follows: Before the high-risk time window, a pre-adjustment period is set. Based on the frequency coupling path and key response nodes, the electrical parameters of the backup power channel are obtained. The reactive current phase angle of the inverter is adjusted to achieve admittance structure shaping and break the original frequency conduction channel. After admittance shaping is completed, a path with energy storage and load regulation capabilities is selected as the shunt path based on electrical characteristics, and the main power is naturally transferred through phase shift and current guiding mechanisms. The reserved capacity of the energy storage device is used to generate a feedforward control command with a reverse phase, which limits the frequency response range and output rate, forming a reverse energy wave at the target frequency to weaken frequency coupling. Based on the energy storage output, the main inverter constructs an anti-phase compensation channel, precisely adjusts the phase angle of the reactive power output trajectory, forms a conjugate energy cancellation structure, monitors the electrical response of key nodes in real time and dynamically switches the compensation state, and completes the frequency stability control closed loop.

8. The distributed photovoltaic grid-connected adaptability assessment system according to claim 7, characterized in that, Based on the stability of the anti-phase compensation path, time-inversion phase gating is performed according to the natural frequency, and the phase conjugate reactive power envelope signal is injected synchronously to construct an anti-channel with the stored energy, suppressing voltage oscillations. The calibration factor is then written into the unified time baseline. The steps are as follows: Retrieve the list of identified natural frequencies, select the frequency with the highest coupling strength and the longest historical oscillation duration, extract the corresponding phasor trajectory and perform time inversion, and generate the reverse control trajectory as the conjugate injection reference. Based on the inversion trajectory, the injection tasks of multiple inverters are constructed, the injection frequency and phase parameters are set, and a unified time baseline is used as the start-up reference so that each inverter can synchronously complete the phase conjugate reactive power injection. Monitor the frequency energy spectrum and phase angle change trends of key coupling nodes, compare the energy density decrease and waveform changes before and after injection, and judge the oscillation suppression effect of injection behavior; Key response factors in the control process are extracted and written into the response data structure in a unified time baseline to form a frequency response profile for subsequent calls. By combining data from six stages—prediction, identification, constraint, compensation, feedback, and calibration—regression analysis is performed on the entire frequency coupling process to construct a closed-loop control path with self-learning capabilities.

Images