A method and system for verifying electromagnetic transient modeling of energy storage power stations

Abstract

This application provides a method and system for electromagnetic transient modeling and verification of an energy storage power station, relating to the technical field of model verification. The method includes: during modeling, constructing an electromagnetic transient digital model of the energy storage power station containing a detailed switching model and an average value converter model; calculating the consistency deviation by applying the same disturbance signal to both models and collecting dynamic response data; during simulation, if the consistency deviation is less than a preset threshold, locking the model, performing simulation calculations under specified operating conditions to obtain model simulation data, and acquiring corresponding benchmark data; during verification, performing time-series synchronization processing on the model simulation data and benchmark data, calculating the verification error, and generating a verification pass signal if it is less than a preset error threshold. This application can improve the accuracy or reliability of verification results.

Description

Technical Field

[0001] This application relates to the technical field of model verification, and in particular to a method and system for electromagnetic transient modeling and verification of energy storage power stations. Background Technology

[0002] With the rapid development of new power systems, energy storage power stations, as key facilities for mitigating new energy fluctuations and providing frequency and voltage regulation support, are increasingly being connected to the grid. The core equipment of energy storage power stations includes a large number of power electronic converters, which exhibit complex millisecond-level or even microsecond-level electromagnetic transient characteristics when the power grid experiences faults or disturbances. To accurately assess the impact of energy storage power stations on grid stability, optimize control strategies, and verify the behavior of protection devices, constructing high-precision electromagnetic transient digital models of energy storage power stations and conducting rigorous model verification has become an indispensable foundational task in the field of power system simulation analysis.

[0003] In existing electromagnetic transient modeling and verification practices for energy storage power stations, technicians typically construct either detailed switching models or average-value converter models independently, depending on the simulation objective. Detailed switching models focus on reproducing the physical switching process of power devices and are suitable for analyzing high-frequency harmonics and device stress; average-value models focus on the fundamental dynamic characteristics and are suitable for long-term system-level simulations. When verifying model validity, the conventional procedure is to directly select one model (usually based on empirical judgment of simulation efficiency or accuracy requirements), perform simulation calculations under specific disturbance conditions, obtain simulation waveform data, and then perform time-series alignment and error comparison with test data from a hardware-in-the-loop simulation platform or field-measured data. If the deviation indicators meet relevant standards or preset thresholds, the digital model is considered successfully constructed and verified, and can then be used for subsequent grid stability analysis or control strategy evaluation.

[0004] However, the aforementioned existing technologies lack a pre-verification step to check the inherent consistency of dynamic responses between models of different accuracy levels (such as detailed models and average models) under the same control strategy. They directly skip the self-consistency check between models and rely only on a single comparison between the model and external benchmark data, resulting in one-sidedness and low accuracy of the verification results. Summary of the Invention

[0005] To improve the accuracy or reliability of the verification results, this application provides a method and system for electromagnetic transient modeling and verification of energy storage power stations.

[0006] Firstly, this application provides a method for electromagnetic transient modeling and verification of energy storage power stations, employing the following technical solution: A method for electromagnetic transient modeling and verification of an energy storage power station includes the following steps: Modeling: Construct an electromagnetic transient digital model of the energy storage power station, which includes a detailed switching model and an average value converter model generated based on the same control strategy; Calculate the consistency deviation: Apply the same disturbance signal to the detailed switching model and the average value converter model, collect the dynamic response data of the detailed switching model and the average value converter model during the transient process, and calculate the consistency deviation based on the dynamic response data; Simulation: When the consistency deviation is less than the preset consistency threshold, the electromagnetic transient digital model is locked, and the specified working condition is simulated using the locked electromagnetic transient digital model to obtain model simulation data. The semi-physical simulation test data or measured data corresponding to the specified working condition are obtained and recorded as the reference data. Verification: The model simulation data and the benchmark data are time-synchronized. The verification error between the time-synchronized model simulation data and the time-synchronized benchmark data is calculated. When the verification error is less than a preset error threshold, a verification pass signal is generated.

[0007] This application constructs an electromagnetic transient digital model that includes a detailed switching model and an average value converter model, integrating these two models into a single electromagnetic transient digital model. This approach can take into account both the microscopic details of the energy storage power station and the overall performance.

[0008] Subsequently, this application applies the same perturbation signal to both the detailed switching model and the average-value converter model and collects dynamic response data to calculate the consistency deviation. This allows for a quantitative assessment of the consistency between the two models in describing the same electromagnetic transient process. It helps determine whether the average-value converter model excessively sacrifices accuracy during computational simplification and whether it can accurately replace the detailed switching model for analysis and simulation in specific scenarios. A small consistency deviation indicates that the average-value converter model, while maintaining computational efficiency, can better reflect the dynamic characteristics of the system and possesses high reliability.

[0009] This application locks the electromagnetic transient digital model when the consistency deviation is less than a preset consistency threshold, ensuring that the model used in the simulation calculation of the specified working condition is verified and relatively stable, thus minimizing the inaccuracy and unreliability of the results due to the uncertainty of the model during the simulation process, and providing a guarantee for obtaining accurate and reliable model simulation data.

[0010] This application also calculates the verification error after performing time-series synchronization processing on the model simulation data and the benchmark data. The time-series synchronization processing ensures the consistency of the compared data on the time axis, avoids errors caused by time differences, and makes the verification results more accurate and reliable.

[0011] Optionally, calculate the internal consistency deviation and determine the internal logical consistency of the model based on the internal consistency deviation, including: The dynamic response data of the detailed switching model and the average converter model at the same time scale are extracted and denoted as target data. The target data includes four types of time series data: voltage fundamental positive sequence component, current fundamental positive sequence component, active power and reactive power. The root mean square value of the relative error of each type of target data in the transient process is calculated respectively. If the root mean square error of any target data exceeds the preset consistency threshold, the internal logic of the model is determined to be inconsistent; otherwise, the internal logic of the model is determined to be consistent.

[0012] This application extracts four types of time series data—voltage fundamental positive sequence component, current fundamental positive sequence component, active power, and reactive power—as target data, and calculates the root mean square value of the relative error of each target data in the transient process. This quantifies the differences between models. The root mean square value of the relative error comprehensively considers the magnitude and distribution of the error, and can more accurately reflect the consistency of the dynamic response of the model throughout the transient process.

[0013] Optionally, in the process of calculating the consistency deviation based on dynamic response data, the method further includes: The high-frequency residual signal between the dynamic response data of the detailed switching model and the dynamic response data of the average converter model is calculated. The high-frequency residual signal contains the switching frequency and harmonic components at the multiples of the switching frequency. The high-frequency residual signal is filtered and phase-compensated to obtain an additional disturbance source. The additional disturbance source is injected into the average converter model to obtain new dynamic response data output by the average converter model. The consistency deviation is calculated based on the new dynamic response data output from the average value converter model and the dynamic response data from the detailed switching model.

[0014] Detailed switching models can accurately reflect the high-frequency harmonics generated by switching devices during the switching process, while average-value converter models typically simplify the harmonics at the switching frequency and its harmonics. By calculating the high-frequency residual signal, this application can capture the differences in high-frequency characteristics between the two models.

[0015] This application obtains an additional disturbance source by filtering and phase compensation of the high-frequency residual signal, and injects the additional disturbance source into the average value converter model. This can effectively compensate the average value converter model, enabling it to better simulate the high-frequency behavior in the detailed switching model, thereby narrowing the gap between the two models and improving the accuracy of the average value converter model.

[0016] Optionally, during the acquisition of reference data, the method further includes: denoising the collected hardware-in-the-loop simulation test data or measured data; and forcibly synchronizing the start time of the denoised hardware-in-the-loop simulation test data or measured data with the trigger time of the locked electromagnetic transient digital model.

[0017] This application employs denoising processing to effectively remove noise components, resulting in cleaner data that more accurately reflects the actual operation of energy storage power stations. Accurate time synchronization and denoising processing enhance the reliability of the verification results. When both benchmark data and model simulation data are effectively guaranteed in terms of time and quality, the calculated verification error can more realistically reflect the model's performance and accuracy.

[0018] Optionally, in the verification step, the model simulation data and the benchmark data undergo time synchronization processing, including: Identify the characteristic moments when disturbances occur in the model simulation data and the baseline data, mark the characteristic moments as time zero, and extract model simulation data and baseline data of a preset duration based on time zero. Using the relative time zero point as a common time axis reference, it is determined whether the sampling frequency of the model simulation data and the reference data are consistent. If they are inconsistent, the low sampling rate data is readjusted to the resolution of the high sampling rate data using interpolation, generating a dataset to be aligned with the same timestamp density.

[0019] This application identifies the characteristic moments when disturbances occur in model simulation data and benchmark data and marks them uniformly as time zero points, so that subsequent data extraction for a preset duration has a unified time starting point, enabling direct comparison between model simulation data and benchmark data in the time dimension.

[0020] Subsequently, this application determines whether the sampling frequencies of the model simulation data and the benchmark data are consistent. For cases of inconsistency, interpolation is used to resample the low-sampling-rate data to the resolution of the high-sampling-rate data, generating a dataset to be aligned with the same timestamp density. This solves the problem of data incomparability due to different sampling frequencies. Interpolation can reasonably insert new data points based on the changing trends between low-sampling-rate data points, ensuring that the two sets of data have the same resolution in time. This allows for precise point-by-point comparisons, significantly improving the accuracy of verification. For example, if the sampling frequency of the model simulation data is 10kHz, while the sampling frequency of the benchmark data is 5kHz, interpolation can resample the benchmark data to 10kHz, ensuring a one-to-one correspondence of data points on the time axis.

[0021] Optionally, after generating the dataset to be aligned, the method further includes: Within a pre-defined small neighborhood near the relative time zero point, the cross-correlation coefficient of the dataset to be aligned is calculated using the sliding window method. With the goal of maximizing the cross-correlation coefficient, the optimal time offset is automatically searched, and the simulation data or benchmark data in the dataset to be aligned are translated and corrected.

[0022] This application utilizes a sliding window method to calculate the cross-correlation coefficient within a pre-defined small neighborhood near the relative time zero point, enabling detailed analysis of the similarity between model simulation data and benchmark data within this local range. The cross-correlation coefficient reflects the correlation between two data sequences at different time offsets. By automatically searching for the optimal time offset with the goal of maximizing the cross-correlation coefficient, the time point that maximizes the similarity between the two sets of data can be precisely found, thus achieving high-precision data alignment. Compared to coarse time alignment methods, this application can capture more subtle time differences. By searching within a small neighborhood near the relative time zero point, this application can better handle the rapid changes in the system at the moment of disturbance, improving the accuracy of data alignment and making the aligned data more realistically reflect the actual operation of the system.

[0023] Accurate data alignment is a crucial prerequisite for verifying model accuracy. If there is a temporal discrepancy between the model simulation data and the baseline data, even if the model itself has high accuracy, it will lead to increased verification errors. The translation correction method described above eliminates the errors caused by time offset, making the verification results more accurately reflect the degree of consistency between the model and the actual system.

[0024] Optionally, after generating the verification pass signal, the method further includes: Using the validated electromagnetic transient digital model, the simulation step size is set; Within a preset frequency range, disturbance signals of different frequencies are injected into the verified electromagnetic transient digital model to calculate the positive sequence equivalent impedance amplitude-frequency characteristic curve and phase-frequency characteristic curve of the energy storage power station. The positive-sequence equivalent impedance characteristic curve is superimposed and analyzed with the background impedance characteristic curve of the power grid to obtain the amplitude intersection point. If an amplitude intersection point exists and the phase margin at the amplitude intersection point is less than the preset safety margin, an early warning signal is output.

[0025] This application injects disturbance signals of different frequencies into the model within a preset frequency range and calculates the positive-sequence equivalent impedance amplitude-frequency characteristic curve and phase-frequency characteristic curve of the energy storage power station. This allows for a comprehensive and accurate acquisition of the impedance characteristics of the energy storage power station at different frequencies. The amplitude-frequency characteristic curve reflects the change in impedance amplitude with frequency, while the phase-frequency characteristic curve reflects the change in impedance phase with frequency. The positive-sequence equivalent impedance characteristic curve can reveal the electrical response characteristics of the energy storage power station at different frequencies, helping to identify potential problems in the system.

[0026] This application analyzes the superposition of the positive-sequence equivalent impedance characteristic curve and the grid background impedance characteristic curve to obtain the amplitude intersection point. It then determines whether the phase margin at this intersection point is less than a preset safety margin, effectively assessing the interaction risk between the energy storage power station and the grid. The amplitude intersection point represents the point where the energy storage power station and the grid may resonate or interact strongly at that frequency, while the phase margin reflects the system's stability at that point. If the phase margin is less than the preset safety margin, it indicates a risk of instability in the system at that frequency, potentially leading to problems such as resonant overvoltage and overcurrent, threatening the safe operation of both the energy storage power station and the grid.

[0027] Secondly, this application provides an electromagnetic transient modeling and verification system for energy storage power stations, which adopts the following technical solution: An electromagnetic transient modeling and verification system for an energy storage power station includes: a processor and a memory communicatively connected to the processor; The memory is provided with a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the processor processes a computer program stored on the computer-readable storage medium, it implements the method as described in the first aspect.

[0028] In summary, this application includes at least one of the following beneficial technical effects: 1. This application applies the same perturbation signal to both the detailed switching model and the average-value converter model and collects dynamic response data, calculating the consistency deviation. This allows for a quantitative assessment of the consistency between the two models in describing the same electromagnetic transient process. It helps determine whether the average-value converter model excessively sacrifices accuracy during computational simplification and whether it can accurately replace the detailed switching model for analysis and simulation in specific scenarios. A small consistency deviation indicates that the average-value converter model, while maintaining computational efficiency, can better reflect the dynamic characteristics of the system and possesses high reliability.

[0029] 2. This application locks the electromagnetic transient digital model when the consistency deviation is less than a preset consistency threshold, so that the model used in the simulation calculation of the specified working condition is verified and relatively stable, and avoids inaccurate and unreliable results due to the uncertainty of the model during the simulation process as much as possible, thus providing a guarantee for obtaining accurate and reliable model simulation data.

[0030] 3. This application also calculates the verification error after performing time synchronization processing on the model simulation data and the benchmark data. Time synchronization processing ensures the consistency of the compared data on the time axis, avoids errors caused by time differences, and makes the verification results more accurate and reliable. Attached Figure Description

[0031] Figure 1This is a flowchart of the method in Embodiment 1 of this application. Detailed Implementation

[0032] The following combination Figure 1 This application will be described in further detail.

[0033] Example 1: This example discloses a method for electromagnetic transient modeling and verification of an energy storage power station, referring to... Figure 1 The method includes: S1 modeling is used to construct an electromagnetic transient digital model of the energy storage power station. The electromagnetic transient digital model includes a detailed switching model and an average value converter model generated based on the same control strategy.

[0034] The detailed switching model retains the actual switching timing and turn-on / turn-off characteristics of the power semiconductor devices (such as IGBTs and MOSFETs) inside the converter, and can truly reflect the fine electrical characteristics of the converter at the switching frequency and its multiples, such as high-frequency harmonics, switching transient spikes, and loss distribution.

[0035] The main circuit of the detailed switching model adopts the actual three-phase two-level converter circuit structure. In the power device modeling of the detailed switching model, each bridge arm includes a specific IGBT / MOSFET and its anti-parallel diode. The device model in the detailed switching model includes nonlinear on-resistance, junction capacitance, reverse recovery characteristics, and dead-time logic. The detailed switching model includes a real LCL or LC filter, with inductors considering saturation characteristics and capacitors considering equivalent series resistance (ESR). The DC side of the detailed switching model includes the Thevenin equivalent circuit of the battery pack (open-circuit voltage + internal resistance + RC parallel network), which can accurately simulate the DC bus voltage ripple.

[0036] The detailed switching model's driving and switching logic is as follows: it receives discrete PWM pulse signals (0 or 1) from the control layer, and within each simulation step (usually at the microsecond level, such as 1μs-5μs), it strictly determines the on and off states of the device based on the pulse state, generating a stepped voltage / current waveform.

[0037] The detailed switching model can output rich harmonic components including the switching frequency (e.g., 2kHz-10kHz) and its harmonics.

[0038] The average value converter model ignores the high-frequency switching process, uses the average value equivalent circuit to represent the external characteristics of the converter, and replaces the actual switching topology with a controlled voltage source / current source.

[0039] In the average value converter model, the AC side no longer uses switching devices, but instead uses three controlled voltage sources to replace the bridge arms. The amplitude of the controlled voltage sources is determined by the modulation wave, as expressed in the following formula:

[0040] in, It is the equivalent instantaneous voltage of the AC side A-phase output terminal of the converter relative to the DC side neutral point (or virtual neutral point); This refers to the total voltage of the DC bus of the converter in the energy storage power station. The modulated wave signal of phase A, when its value is 1, indicates that the upper bridge arm is fully conducting, and the output is... When its value is -1, it indicates that the lower bridge arm is fully conductive, and the output is... When its value is 0, the output is 0.

[0041] In the average value converter model, the DC side is replaced by a controlled current source at the DC inlet of the converter. The current injected into the DC bus is calculated based on the power conservation principle or current mapping relationship, as shown in the following formula:

[0042] in, This is the DC side current; , , These represent the instantaneous output voltages of phases A, B, and C on the AC side of the converter, respectively. , , These represent the instantaneous currents flowing through phases A, B, and C of the converter, respectively. This represents the sum of the instantaneous power of the three phases; This is the total voltage of the DC bus of the converter in the energy storage power station.

[0043] The filtering stage of the average value converter model retains the same LCL or LC filter parameters as the detailed switching model.

[0044] The average value converter model receives a continuous modulated wave signal (sine reference wave) or a duty cycle signal.

[0045] The average value converter model outputs a smooth sinusoidal fundamental component, free of switching frequency harmonics.

[0046] The two models share the same set of control algorithm code, control parameters, and logic modules, differing only in the mathematical expression of the main circuit topology: Outer loop: The control logic of the power loop / voltage loop / current loop is consistent; Inner loop: Modulation wave generation, phase-locked loop, and limiting logic are consistent; Protection logic: The locking, unlocking, and fault traversal logics are consistent; This ensures that the two models have a consistent foundation in terms of fundamental component, dynamic response trend, and control command output.

[0047] S2 calculates the consistency deviation, including S21 data acquisition, S22 high-frequency residual compensation, and S23 consistency judgment.

[0048] S21 data acquisition applies identical external stimuli and operating conditions to both the detailed switching model and the average value converter model, including: grid voltage step disturbance, system frequency fluctuation, three-phase / single-phase short-circuit fault, and converter unlocking, blocking, and switching processes.

[0049] The electrical quantities of the two models during the transient process are collected simultaneously, including but not limited to: grid connection point voltage and current, active power and reactive power, DC side voltage and current, control command output and modulation wave.

[0050] S22 high-frequency residual compensation calculates the high-frequency residual signal between the dynamic response data of the detailed switching model and the average value model. The calculation model for the high-frequency residual signal is as follows:

[0051] in, The high-frequency residual signal of the i-th type of dynamic response data; Let be the value of the i-th dynamic response data output by the detailed switching model at time t; The value of the i-th dynamic response data output by the converter model at time t is the average value.

[0052] Theoretically, if the fundamental frequency control logic of the two models is completely identical, then after subtraction, the fundamental frequency component and the low-frequency dynamic component will cancel each other out. It can express the ripple caused by switching action, the distortion caused by dead zone effect, and the small phase error caused by discretization delay.

[0053] A Fast Fourier Transform (FFT) is performed on the high-frequency residual signal to obtain the amplitude spectrum. From the amplitude spectrum, the switching frequency and its harmonics are extracted. Specifically, the frequency corresponding to the maximum peak value in the amplitude spectrum is identified as the switching frequency. Search for 2 in sequence 3 ... n The local maximum value in the vicinity is determined to be an overtone component.

[0054] A Chebyshev IIR filter is used to filter the high-frequency residual signal, removing the 50Hz and low-frequency dynamic components to obtain the clean switching frequency and harmonic components. During the filtering of the high-frequency residual signal, an all-pass filter is connected in series with the Chebyshev IIR filter, and the phase compensation parameters of the all-pass filter are set so that the phase lead of the phase compensation parameter at each of the aforementioned characteristic frequency points is completely equivalent to the phase lag of the Chebyshev IIR filter, achieving complete cancellation of phase errors and obtaining the processed high-frequency residual signal. .

[0055] An additional disturbance source is constructed based on the processed high-frequency residual signal. The process involves: fine-tuning the amplitude of the processed high-frequency residual signal, setting an amplitude correction coefficient, the value of which is controlled between 0.95 and 1.05, and finally generating the additional disturbance source. The calculation model of the additional disturbance source is shown below:

[0056] in, For additional disturbance sources; This is the amplitude correction factor; This is the processed high-frequency residual signal.

[0057] The additional disturbance source is added to the dynamic response parameters output by the average value converter model to obtain new dynamic response data output by the average value converter model. The calculation model is as follows:

[0058] in, Let be the value of the new i-th type of dynamic response data output by the converter model at time t; The average value of the i-th dynamic response data output by the converter model at time t; This is an additional source of disturbance.

[0059] S23 Consistency Judgment: Dynamic response data from the detailed switching model and new dynamic response data from the average converter model at the same time scale are extracted and denoted as target data. Target data includes four time series data: voltage fundamental positive sequence component, current fundamental positive sequence component, active power, and reactive power. The root mean square error (RMSRE) of the relative error of each target data in the transient process is calculated, i.e., the consistency deviation. The calculation model is as follows:

[0060] in, The root mean square error of the i-th type of target data; Let i be the number of the i-th type of target data; This refers to the j-th data point in the i-th type of target data output by the detailed switching model; This refers to the j-th data point in the i-th type of target data output by the average converter model.

[0061] If the RMSRE of any target data exceeds the preset consistency threshold (e.g., 5%), the internal logic of the model is determined to be inconsistent; otherwise, the internal logic of the model is determined to be consistent, and S3 deviation analysis is performed.

[0062] In this embodiment, the total simulation duration is set to 2 seconds. At t=0.5s, a phase A ground fault is simultaneously applied to both the detailed switch model and the average value converter model, and the fault lasts for 0.2 seconds. During the simulation, the grid-connected three-phase current, DC bus voltage, and d-axis current reference command of both models are recorded synchronously at a sampling rate of 50kHz.

[0063] The high-frequency residual signal is obtained by subtracting the current data from the average value model (AVM) from the raw current data of the detailed switching model (DSM).

[0064] Fast Fourier Transform (FFT) spectrum analysis is performed on the high-frequency residual signal to identify the switching frequency and its harmonic components.

[0065] A bandpass filter is designed to extract the aforementioned switching frequency components, and phase compensation is performed on the residual signal based on the system delay characteristics to generate an additional disturbance source. This additional disturbance source is then superimposed on the AC output of the average-value converter model as a current injection term (or as a correction term for the controlled source), causing the average-value model to output new dynamic response data.

[0066] Dynamic response data from the detailed switching model and the new dynamic response data from the compensated average model at the same time scale were extracted separately. Four key target data time series were selected: voltage fundamental positive sequence component, current fundamental positive sequence component, active power, and reactive power. For each target data, the root mean square value of its relative error in the transient process was calculated.

[0067] If the root mean square value of the relative error of any target data is greater than the preset error threshold, the internal logic of the model is determined to be inconsistent. Even if the high-frequency residual is compensated, there is still a significant difference in the fundamental dynamic characteristics of the two. This may be due to errors in the implementation of the control strategy, parameter mismatch, or defects in the topological equivalence principle of the average value model, and an alarm signal will be issued.

[0068] If the root mean square error of the relative errors of all four target data is not greater than the preset error threshold, the internal logic of the model is considered consistent. This indicates that the average value model, after high-frequency residual compensation, can accurately reproduce the dynamic and static characteristics of the detailed switching model.

[0069] In S3 simulation, when the consistency deviation of various target data is less than the preset consistency threshold, the electromagnetic transient digital model is locked (i.e., the model structure, control logic, main circuit parameters, simulation step size, and solver configuration are fixed).

[0070] The locked model is used to perform simulation calculations for specified operating conditions, resulting in model simulation data, which includes time series data such as transient voltage, current, power, and DC voltage. The specified operating conditions include: symmetrical / asymmetrical grid faults, large-capacity load switching, voltage dips, voltage surges, weak grids, and high-impedance grid connection scenarios.

[0071] Obtain the hardware-in-the-loop test data or measured data corresponding to the specified working conditions mentioned above, and record them as the reference data. Use wavelet transform to denoise the collected reference data.

[0072] The initial synchronization of the start time of the denoised reference data with the trigger time of the locked electromagnetic transient digital model is forcibly achieved. This includes: identifying the simulation trigger times of fault input, load switching, and voltage disturbances in the simulation data and recording them as the start times; and identifying the starting points of electrical quantity abrupt changes (such as voltage sag, current abrupt change, and power jump) for the corresponding operating conditions in the reference data and recording them as the trigger times. The time difference between the original trigger time of the reference data and the zero point of the simulation reference is calculated, which is the timing offset. Based on this time offset, the entire reference data is linearly shifted in time, forcibly aligning the disturbance trigger times of the reference data with the simulation trigger times of the locked model. This unifies the start times of the two sets of data, resulting in the model simulation data and the reference data after the start time synchronization.

[0073] S4 verification identifies the characteristic moments (such as the voltage drop start point and the fault occurrence time) of disturbances in the model simulation data and the reference data after the start time synchronization in the S3 simulation. The characteristic moments are uniformly marked as time zero points. Based on time zero points, model simulation data and reference data of preset durations are extracted respectively.

[0074] Using zero point as a common time axis reference, it is determined whether the sampling frequencies of the model simulation data and the reference data within a preset time period are consistent. If they are inconsistent, an interpolation method (such as cubic spline interpolation) is used to readjust the low sampling rate data to the resolution of the high sampling rate data, generating a dataset to be aligned with the same timestamp density. The dataset to be aligned includes two sets of data: model simulation data and reference data within the preset time period.

[0075] Within a predefined small neighborhood near time zero, the cross-correlation coefficient of the dataset to be aligned is calculated using the sliding window method. The calculation model is as follows:

[0076] in, Time offset The corresponding cross-correlation coefficients; This represents the time offset value; The sampled values ​​of the model simulation data at time t; This represents the average value of the model simulation data; For benchmark data in The sampled value at time; This is the average value of the baseline data.

[0077] Calculate the optimal time offset with the goal of maximizing the cross-correlation coefficient. The value of the time offset is denoted as the optimal time offset. Based on the optimal time offset, the simulation data or reference data is linearly shifted and corrected to achieve time alignment of the data at each sampling point, thus obtaining the model simulation data and reference data after time synchronization processing.

[0078] The verification error (which can be root mean square error or maximum absolute error) between the model simulation data after time-series synchronization processing and the benchmark data is calculated. In this embodiment, the root mean square error is selected as the verification error. The calculation model of the verification error is the same as that of the consistency deviation, and will not be repeated here. If the verification error is less than a preset error threshold (e.g., 3%), a verification pass signal is generated, indicating that the model is reliable. Otherwise, an alarm signal is issued to the operation and maintenance personnel.

[0079] S5 power grid adaptability analysis utilizes a verified electromagnetic transient digital model, sets a simulation step size ranging from 1 to 5 microseconds, and injects disturbance signals of different frequencies into the verified electromagnetic transient digital model within a preset frequency range (e.g., 10Hz-2000Hz). In this embodiment, the disturbance signal refers to a small-amplitude positive-sequence voltage disturbance signal with an amplitude of 3%-5% of the rated voltage.

[0080] After each frequency point disturbance injection, wait 10 signal cycles, and then synchronously collect the three-phase voltage and three-phase current time-series data of the grid-connected point. Calculate the positive-sequence equivalent impedance amplitude and phase for each frequency point. Specifically, decompose the collected three-phase voltage and current data using the symmetrical component method, extract the fundamental positive-sequence voltage and positive-sequence current time-series components for each swept frequency point, perform a fast Fourier transform on the positive-sequence voltage and positive-sequence current, and extract the voltage amplitude corresponding to the current swept frequency point. Voltage phase Current amplitude I, current phase The magnitude of the positive-sequence equivalent impedance is equal to the ratio of the voltage magnitude to the current magnitude, and the phase of the positive-sequence equivalent impedance is equal to the difference between the voltage phase and the current phase.

[0081] Based on the calculated positive-sequence equivalent impedance amplitude and phase, the positive-sequence equivalent impedance amplitude-frequency characteristic curve and phase-frequency characteristic curve of the energy storage power station are plotted. The positive-sequence equivalent impedance amplitude-frequency characteristic curve is plotted with frequency on the horizontal axis and impedance amplitude on the vertical axis, representing the variation of the impedance amplitude of the energy storage power station with frequency. The phase-frequency characteristic curve is plotted with frequency on the horizontal axis and impedance phase value on the vertical axis, representing the variation of the impedance phase of the energy storage power station with frequency.

[0082] The positive-sequence equivalent impedance characteristic curve of the energy storage power station and the background impedance characteristic curve of the power grid are superimposed and analyzed in the same coordinate system to find the amplitude intersection point of the two curves. The process is as follows: traverse the amplitude-frequency characteristic curves of the entire frequency band, and find the intersection point of the energy storage impedance amplitude curve and the background impedance amplitude curve of the power grid through numerical interpolation. This point is the amplitude intersection point, and the corresponding frequency is recorded as the resonance risk frequency.

[0083] If an amplitude intersection point exists, the phase margin at the amplitude intersection point is calculated, and the calculation model is as follows:

[0084]

[0085] in, The positive sequence impedance phase of the energy storage power station Phase with the positive sequence impedance of the power grid background The total phase difference between them; This represents the phase margin.

[0086] If the phase margin is less than the preset safety margin (e.g., 30 degrees), the system is determined to have an oscillation risk and an early warning signal is output; if the phase margin is not less than the preset safety margin (e.g., 30 degrees), the system is determined not to have an oscillation risk.

[0087] Example 2: This example discloses an electromagnetic transient modeling and verification system for an energy storage power station. The system includes a processor and a memory communicatively connected to the processor. The memory is provided with a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the processor processes the computer program stored on the computer-readable storage medium, it implements the method described in Embodiment 1.

[0088] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for electromagnetic transient modeling and verification of an energy storage power station, characterized in that, include: Modeling: Construct an electromagnetic transient digital model of the energy storage power station, which includes a detailed switching model and an average value converter model generated based on the same control strategy; Calculate the consistency deviation: Apply the same disturbance signal to the detailed switching model and the average value converter model, collect the dynamic response data of the detailed switching model and the average value converter model during the transient process, and calculate the consistency deviation based on the dynamic response data; Simulation: When the consistency deviation is less than the preset consistency threshold, the electromagnetic transient digital model is locked, and the specified working condition is simulated using the locked electromagnetic transient digital model to obtain model simulation data. The semi-physical simulation test data or measured data corresponding to the specified working condition are obtained and recorded as the reference data. Verification: The model simulation data and the benchmark data are time-synchronized. The verification error between the time-synchronized model simulation data and the time-synchronized benchmark data is calculated. When the verification error is less than a preset error threshold, a verification pass signal is generated.

2. The electromagnetic transient modeling and verification method for energy storage power stations according to claim 1, characterized in that, Calculate the internal consistency deviation and determine the internal logical consistency of the model based on the internal consistency deviation, including: The dynamic response data of the detailed switching model and the average converter model at the same time scale are extracted and denoted as target data. The target data includes four types of time series data: voltage fundamental positive sequence component, current fundamental positive sequence component, active power and reactive power. The root mean square value of the relative error of each type of target data in the transient process is calculated respectively. If the root mean square error of any target data exceeds the preset consistency threshold, the internal logic of the model is determined to be inconsistent; otherwise, the internal logic of the model is determined to be consistent.

3. The electromagnetic transient modeling and verification method for energy storage power stations according to claim 1 or 2, characterized in that, In the process of calculating consistency deviation based on dynamic response data, the method further includes: The high-frequency residual signal between the dynamic response data of the detailed switching model and the dynamic response data of the average converter model is calculated. The high-frequency residual signal contains the switching frequency and harmonic components at the multiples of the switching frequency. The high-frequency residual signal is filtered and phase-compensated to obtain an additional disturbance source. The additional disturbance source is injected into the average converter model to obtain new dynamic response data output by the average converter model. The consistency deviation is calculated based on the new dynamic response data output from the average value converter model and the dynamic response data from the detailed switching model.

4. The electromagnetic transient modeling and verification method for energy storage power stations according to claim 1 or 2, characterized in that, In the process of acquiring benchmark data, the method further includes: denoising the collected hardware-in-the-loop simulation test data or measured data; and forcibly synchronizing the start time of the denoised hardware-in-the-loop simulation test data or measured data with the trigger time of the locked electromagnetic transient digital model.

5. The electromagnetic transient modeling and verification method for energy storage power stations according to claim 1 or 2, characterized in that, In the verification process, the model simulation data and the baseline data undergo time synchronization processing, including: Identify the characteristic moments when disturbances occur in the model simulation data and the baseline data, mark the characteristic moments as time zero, and extract model simulation data and baseline data of a preset duration based on time zero. Using the relative time zero point as a common time axis reference, it is determined whether the sampling frequency of the model simulation data and the reference data are consistent. If they are inconsistent, the low sampling rate data is readjusted to the resolution of the high sampling rate data using interpolation, generating a dataset to be aligned with the same timestamp density.

6. The electromagnetic transient modeling and verification method for energy storage power stations according to claim 5, characterized in that, After generating the dataset to be aligned, the process also includes: Within a pre-defined small neighborhood near the relative time zero point, the cross-correlation coefficient of the dataset to be aligned is calculated using the sliding window method. With the goal of maximizing the cross-correlation coefficient, the optimal time offset is automatically searched, and the simulation data or benchmark data in the dataset to be aligned are translated and corrected.

7. The electromagnetic transient modeling and verification method for energy storage power stations according to claim 1 or 2, characterized in that, After generating the verification pass signal, the process also includes: Using the validated electromagnetic transient digital model, the simulation step size is set; Within a preset frequency range, disturbance signals of different frequencies are injected into the verified electromagnetic transient digital model to calculate the positive sequence equivalent impedance amplitude-frequency characteristic curve and phase-frequency characteristic curve of the energy storage power station. The positive-sequence equivalent impedance characteristic curve is superimposed and analyzed with the background impedance characteristic curve of the power grid to obtain the amplitude intersection point. If an amplitude intersection point exists and the phase margin at the amplitude intersection point is less than the preset safety margin, an early warning signal is output.

8. An electromagnetic transient modeling and verification system for an energy storage power station, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory is provided with a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the processor processes a computer program stored on the computer-readable storage medium, it implements the method as described in any one of claims 1-7.

Images