Power fluctuation smoothing control method and system for photovoltaic power station cluster

Abstract

The application discloses a power fluctuation smoothing control method and system for photovoltaic power station cluster, and particularly relates to the technical field of power system control, and is used for solving the power oscillation and instability problem caused by the existing smoothing control method due to neglecting the internal power grid dynamic characteristics of the cluster; by acquiring historical power flow and historical smoothing control instructions, the information transmission entropy distribution between the instructions and the flow under different frequencies is analyzed to identify the dominant oscillation mode frequency, the nonlinear interaction strength of the smoothing control loop of each substation is evaluated, and the time and space distribution inertia center of the node oscillation energy is calculated, the system dynamic instability criterion is constructed based on the nonlinear interaction strength and the inertia center offset trajectory, the dynamic safety constraint set with the target of inhibiting the oscillation mode is constructed when the criterion is established, and the smoothing control instruction optimization model is solved under the constraint set restriction to generate and execute new control instructions to each substation.

Description

Technical Field

[0001] This invention relates to the field of power system control technology, and in particular to a method and system for smoothing power fluctuations in photovoltaic power plant clusters. Background Technology

[0002] Photovoltaic power plant clusters, by coordinating and controlling the output of multiple geographically dispersed power plants, smooth out fluctuations in their total power output, which is an effective way to improve the performance of large-scale photovoltaic grid connection. Existing technologies typically use centralized or distributed control architectures to generate and distribute smoothing control commands to each sub-power plant based on the deviation between the cluster's total power output and the target value. The core lies in smoothing out fluctuations in total power output over time through optimized algorithms or control strategies to meet grid connection requirements. Its design and analysis often focus on the time-series response characteristics of the control strategy itself and its impact on external characteristics at the grid's point of common coupling (PCC).

[0003] However, existing smoothing control methods have limitations when applied to large photovoltaic clusters with tight electrical connections. These methods generally treat the internal power grid of the cluster as an ideal power transmission channel, ignoring its inherent dynamic characteristics as a complex electromagnetic network. When rapidly changing smoothing control commands are applied to each power station, complex circulating currents and power oscillations are induced in the internal power grid of the cluster. This power oscillation mode induced by the control command is coupled with the inherent resonant characteristics of the internal power grid of the cluster, which may lead to the amplification of oscillations or even instability. This not only endangers the safe operation of the internal power grid of the cluster, but may also ultimately lead to the failure of the smoothing control objective. Existing technologies lack effective consideration of the dynamic interaction effect between smoothing control commands and the internal power grid when constructing smoothing control strategies. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a method and system for smoothing power fluctuations in photovoltaic power plant clusters.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] Methods for smoothing power fluctuations in photovoltaic power plant clusters include:

[0007] S1. Obtain the historical power flow and corresponding historical smoothing control commands of the internal power grid of the photovoltaic power station cluster.

[0008] S2. Based on historical power flow and historical smoothing control commands, analyze the information transfer entropy distribution between historical smoothing control commands and historical power flow at different frequencies.

[0009] S3. Based on the information transmission entropy distribution, identify the frequency corresponding to the extreme value of information transmission entropy as the dominant oscillation mode frequency of the internal power grid of the photovoltaic power station cluster.

[0010] S4. Evaluate the nonlinear interaction strength of the smooth control loops of different sub-power stations in the photovoltaic power station cluster at the frequency of the dominant oscillation mode, and calculate the spatiotemporal distribution inertia center of the oscillation energy of each node under the dominant oscillation mode.

[0011] S5. Based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center, construct a dynamic instability criterion for the system;

[0012] S6. When the dynamic instability criterion of the system meets the preset conditions, a dynamic safety constraint set with the goal of suppressing oscillation modes is constructed, and an optimization model of smooth control instructions is solved under the constraints of the dynamic safety constraint set. New smooth control instructions are generated and executed to each sub-power station in the photovoltaic power station cluster.

[0013] Furthermore, the historical power flow and corresponding historical smoothing control commands of the internal power grid of the photovoltaic power plant cluster are obtained, including:

[0014] Historical power flow data are collected from monitoring devices within the power grid of a photovoltaic power plant cluster.

[0015] And extract historical smooth control commands from the control command record;

[0016] Historical power flow and historical smoothing control commands are matched and aligned using timestamps to ensure the time correspondence and integrity of the data.

[0017] Furthermore, based on historical power flow and historical smoothing control commands, the information transfer entropy distribution between historical smoothing control commands and historical power flow at different frequencies is analyzed, including:

[0018] The historical power flow and historical smoothing control commands are transformed in the frequency domain to obtain the power flow spectrum and the control command spectrum, respectively.

[0019] The power flow spectrum and the control command spectrum are divided in the frequency domain according to frequency resolution. For each frequency point, the conditional probability transfer entropy from the historical smooth control command to the historical power flow is calculated. The information transfer entropy value of each frequency point is obtained by comparing the logarithmic expectation values ​​of the conditional probability and the unconditional probability.

[0020] The information transmission entropy values ​​corresponding to all frequencies are combined to form the information transmission entropy distribution.

[0021] Furthermore, based on the information transmission entropy distribution, the frequency corresponding to the extreme value of information transmission entropy is identified as the dominant oscillation mode frequency of the power grid within the photovoltaic power plant cluster, including:

[0022] Locate all local extreme points in the entropy distribution of information transmission;

[0023] The extreme points that satisfy the preset salience condition are selected from the local extreme points as the extreme values ​​of information transmission entropy;

[0024] The frequency corresponding to the extreme value of information transmission entropy on the horizontal axis of the information transmission entropy distribution is determined as the dominant oscillation mode frequency of the power grid within the photovoltaic power plant cluster.

[0025] Furthermore, the nonlinear interaction strength of the smoothed control loops of different sub-power stations in the photovoltaic power plant cluster at the dominant oscillation mode frequency is evaluated, and the spatiotemporal distribution inertia center of the oscillation energy of each node under the dominant oscillation mode is calculated, including:

[0026] Extract the dynamic response signal of the smoothing control loop of each sub-power station near the dominant oscillation mode frequency;

[0027] The strength of the nonlinear interaction of the smooth control loop at the dominant oscillation mode frequency is evaluated by calculating the nonlinear correlation coefficient between the dynamic response signals.

[0028] At the same time, the node oscillation energy is calculated based on the oscillation components of the voltage and current at each node;

[0029] By using the node oscillation energy as the weight, a weighted average of the node position coordinates is performed to obtain the spatiotemporal distribution inertia center of each node's oscillation energy under the dominant oscillation mode.

[0030] Furthermore, evaluating the nonlinear interaction strength of the smooth control loop at the dominant oscillation mode frequency by calculating the nonlinear correlation coefficient between dynamic response signals includes: reconstructing the phase space of the dynamic response signals of the smooth control loop of each sub-station, calculating the mutual information between each pair of loops based on the reconstructed state vector, obtaining the nonlinear correlation coefficient by normalizing the mutual information, and obtaining the nonlinear interaction strength by combining the nonlinear correlation coefficients between all loops.

[0031] Furthermore, by using the node oscillation energy as a weight to perform a weighted average of the node position coordinates, the spatiotemporal distribution inertia center of each node's oscillation energy under the dominant oscillation mode is obtained by: firstly calculating the instantaneous oscillation energy of each node under the dominant oscillation mode, using the instantaneous oscillation energy as a weighting coefficient, performing a weighted summation of the node's position coordinates in the power grid topology, and then dividing by the total oscillation energy to obtain the coordinate value of the spatiotemporal distribution inertia center.

[0032] Furthermore, based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center, a dynamic instability criterion for the system is constructed, including:

[0033] The nonlinear interaction strength is compared with a preset interaction threshold, and the offset trajectory of the spatiotemporal distribution inertia center is analyzed to see if it exhibits a continuous divergence characteristic.

[0034] When the nonlinear interaction intensity exceeds the preset interaction threshold and the offset trajectory of the spatiotemporal distribution inertia center simultaneously exhibits continuous divergence characteristics, the dynamic instability criterion of the system is determined to be valid.

[0035] Furthermore, when the system's dynamic instability criterion meets preset conditions, a dynamic safety constraint set aimed at suppressing oscillating modes is constructed. Under the constraints of this dynamic safety constraint set, an optimization model for smooth control commands is solved, generating and executing new smooth control commands to each sub-power station in the photovoltaic power station cluster, including:

[0036] When the dynamic instability criterion of the system is met, a dynamic safety constraint set that limits the rate of change of control commands is constructed based on the frequency of the dominant oscillation mode.

[0037] Under the constraints of the dynamic safety constraint set, an optimization model for solving the smooth control command is used to minimize the total power fluctuation of the cluster as the objective function, and new smooth control commands for each sub-power station are obtained.

[0038] The new smooth control command is issued to each sub-power station in the photovoltaic power station cluster for execution.

[0039] On the other hand, the present invention provides a power fluctuation smoothing control system for photovoltaic power plant clusters, comprising:

[0040] The information acquisition module is used to acquire the historical power flow and corresponding historical smoothing control commands of the power grid within the photovoltaic power plant cluster.

[0041] The distribution analysis module is used to analyze the information transfer entropy distribution between historical power flow and historical smoothing control commands at different frequencies, based on historical power flow and historical smoothing control commands.

[0042] The dominant identification module is used to identify the frequency corresponding to the extreme value of information transmission entropy as the dominant oscillation mode frequency of the internal power grid of the photovoltaic power plant cluster based on the information transmission entropy distribution.

[0043] The information analysis module is used to evaluate the nonlinear interaction strength of the smooth control loops of different sub-power stations in the photovoltaic power station cluster at the frequency of the dominant oscillation mode, and to calculate the spatiotemporal distribution of the oscillation energy of each node under the dominant oscillation mode.

[0044] The criterion construction module is used to construct dynamic instability criteria for the system based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center.

[0045] The instruction execution module is used to construct a dynamic safety constraint set with the goal of suppressing oscillation modes when the system dynamic instability criterion meets the preset conditions, and solve the optimization model of the smooth control instruction under the constraints of the dynamic safety constraint set, and generate and execute new smooth control instructions to each sub-power station in the photovoltaic power station cluster.

[0046] The beneficial effects of this invention are:

[0047] 1. By analyzing the dynamic characteristics of the internal power grid of a photovoltaic power plant cluster and the interaction mechanism of control commands, a safer and more effective power fluctuation smoothing control is achieved. By establishing the information transmission entropy distribution between historical power flow and smoothing control commands, the dominant oscillation mode frequency of the internal power grid can be accurately identified, thereby revealing potential oscillation risks that are difficult to detect by traditional control methods. By evaluating the nonlinear interaction strength of the control loops of each sub-power plant at the dominant frequency and combining the spatiotemporal distribution characteristics of node oscillation energy, a dynamic criterion that can provide early warning of system instability is constructed, enabling the system to actively identify and avoid power oscillations induced by control commands. This solves the problem of oscillation instability caused by neglecting the electromagnetic dynamic characteristics of the internal power grid in existing technologies.

[0048] 2. By constructing a dynamic safety constraint set aimed at suppressing oscillating modes, the dynamic safety requirements of the power grid are directly embedded into the optimization process of smooth control commands. This ensures that the generated smooth control commands not only meet the power smoothing requirements but also do not excite system oscillating modes. Thus, circulating currents and power oscillations are avoided at the source of control. Compared with traditional methods that only focus on external characteristic control, this method significantly improves the operational stability and control reliability of photovoltaic power plant clusters during the smoothing of power fluctuations by coordinating the matching relationship between control commands and the dynamic characteristics of the power grid. Attached Figure Description

[0049] Figure 1 This is a flowchart of the power fluctuation smoothing control method for photovoltaic power plant clusters according to the present invention;

[0050] Figure 2 This is a schematic diagram of the power fluctuation smoothing control system for the photovoltaic power plant cluster of the present invention. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] Example 1: Figure 1 The present invention provides a power fluctuation smoothing control method for photovoltaic power plant clusters, comprising:

[0053] S1. Obtain the historical power flow and corresponding historical smoothing control commands of the internal power grid of the photovoltaic power station cluster.

[0054] S2. Based on historical power flow and historical smoothing control commands, analyze the information transfer entropy distribution between historical smoothing control commands and historical power flow at different frequencies.

[0055] S3. Based on the information transmission entropy distribution, identify the frequency corresponding to the extreme value of information transmission entropy as the dominant oscillation mode frequency of the internal power grid of the photovoltaic power station cluster.

[0056] S4. Evaluate the nonlinear interaction strength of the smooth control loops of different sub-power stations in the photovoltaic power station cluster at the frequency of the dominant oscillation mode, and calculate the spatiotemporal distribution inertia center of the oscillation energy of each node under the dominant oscillation mode.

[0057] S5. Based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center, construct a dynamic instability criterion for the system;

[0058] S6. When the dynamic instability criterion of the system meets the preset conditions, a dynamic safety constraint set with the goal of suppressing oscillation modes is constructed, and an optimization model of smooth control instructions is solved under the constraints of the dynamic safety constraint set. New smooth control instructions are generated and executed to each sub-power station in the photovoltaic power station cluster.

[0059] S1. Obtain the historical power flow and corresponding historical smoothing control commands of the internal power grid of the photovoltaic power plant cluster. The specific implementation is as follows:

[0060] In implementing the power fluctuation smoothing control method for photovoltaic power plant clusters, the first step is data acquisition. This step specifically involves collecting historical power flow data from monitoring devices within the power grid of the photovoltaic power plant cluster. These monitoring devices are deployed at key nodes of the power grid, such as substations and combiner stations. These devices integrate sensors and data acquisition units, enabling real-time monitoring of electrical parameters. The data acquisition unit reads data from the sensors at fixed time intervals, which can be set according to application requirements, such as 1 second or 1 minute, to ensure the time resolution of the data meets the requirements of subsequent analysis. The collected data includes active power, reactive power, voltage, and current values. These parameters are stored in a database in time-series format, with each data point accompanied by a timestamp accurate to the millisecond level to record the precise moment of data acquisition. During data acquisition, the monitoring devices connect to the upper-level system via communication protocols, such as Modbus or IEC 61850, to achieve automatic data upload and storage. The collected historical power flow data needs to undergo preliminary verification, such as checking whether the data range is reasonable and removing obvious outliers, such as power values ​​of zero or exceeding rated limits, to ensure the reliability and availability of the data. The data range is set based on power grid operation specifications. For example, the active power value should be between zero and the maximum allowable power, the reactive power value should be set with positive and negative limits according to the power grid demand, the voltage value should be within the allowable deviation range of the rated voltage, and the current value should not exceed the current carrying capacity of the equipment. These limits are determined through power grid design parameters and operating experience to ensure that the data truly reflects the power grid status.

[0061] Next, historical smoothing control commands are extracted from the control command records. These records originate from the control system database of the photovoltaic power plant cluster, which records all smoothing control commands issued to each sub-power station. The extraction process involves querying the database, filtering command data corresponding to the historical power flow acquisition period based on the time range. Command data includes command type, command parameters, and timestamps. The command type involves power setpoint adjustments, and command parameters are represented numerically, such as power change rate limits or target power values. The timestamps are formatted identically to those in the historical power flow, accurate to the millisecond level, to ensure the feasibility of subsequent matching. During extraction, the completeness of the command records must be verified, such as checking for missing time periods. If missing commands are found, they are supplemented through log entry or interpolation methods to ensure that the command data covers the entire analysis period. Linear interpolation can be used, calculating missing values ​​based on command values ​​at adjacent time points to ensure the continuity of the command sequence. The generation of control commands is based on the smooth control strategy of the photovoltaic power plant cluster. For example, the power allocation of each sub-power plant is calculated through optimization algorithms. The command parameters are dynamically adjusted according to the grid operating conditions and fluctuation characteristics. For example, the power change rate limit is set based on the response capability of the sub-power plant and the grid stability requirements, and the target power value is determined based on the predicted photovoltaic output.

[0062] Finally, historical power flow and historical smoothing control commands are matched and aligned using timestamps. The matching process employs a time alignment algorithm. This algorithm first sorts the historical power flow data and historical smoothing control command data in ascending order of timestamps. Then, for each power flow data point, it finds the control command data point with the smallest time difference for pairing. A tolerance value, such as 100 milliseconds, is allowed for the time difference. The tolerance value is set based on the clock synchronization accuracy and communication latency of the data acquisition system. For example, it ensures time synchronization of each device through a network time protocol. The tolerance value is determined according to the actual error range, and data points within the tolerance range are considered matchable. During the alignment process, if timestamps are not completely consistent, an interpolation method is used to calculate the approximate value of the command value at the corresponding time point. For example, linear interpolation calculates the intermediate value based on the timestamps and values ​​of adjacent command data points to ensure that each power flow data point has a corresponding control command value. After matching is completed, a data integrity check is performed, such as statistically analyzing the proportion of successfully matched data. If the matching rate is lower than a preset matching rate threshold, such as 95%, the time tolerance or data acquisition parameters are readjusted. The preset matching rate threshold is set based on data analysis needs, such as ensuring sufficient data volume for subsequent entropy calculations. The preset matching rate threshold is determined through historical data verification. It is calculated by comparing the ratio of successfully matched data points to the total number of data points with the preset matching rate threshold. If the ratio is lower than the threshold, a readjustment process is triggered. Finally, the matched and aligned dataset is stored in a structured format for subsequent analysis steps, ensuring the temporal correspondence and integrity of the data. The data storage format includes timestamps, power flow values, and corresponding control command values, forming a complete time-series dataset.

[0063] S2. Based on historical power flow and historical smoothing control commands, analyze the information transfer entropy distribution between historical smoothing control commands and historical power flow at different frequencies. The specific implementation is as follows:

[0064] In the implementation of the power fluctuation smoothing control method for photovoltaic power plant clusters, an information transfer entropy distribution analysis step is performed. This step analyzes the information transfer entropy distribution between historical power flow and historical smoothing control commands at different frequencies, based on historical power flow and historical smoothing control commands. First, frequency domain transformations are performed on the historical power flow and historical smoothing control commands to obtain the power flow spectrum and control command spectrum, respectively. The frequency domain transformation uses the Fast Fourier Transform algorithm, which converts time-series data into a frequency domain representation. The input data are the time series of historical power flow and historical smoothing control commands, which are derived from the output of the data acquisition step, ensuring data continuity and time alignment.

[0065] The sampling frequency of the Fast Fourier Transform (FFT) is set based on the data acquisition time interval. For example, if the data acquisition time interval is 1 second, the sampling frequency is set to 1 Hz. The number of transformation points is determined according to the data length and frequency resolution requirements. For example, 1024 points are chosen to balance frequency resolution and computational efficiency. The number of points can be adjusted from 256 to 4096 points, depending on the total data length and the required frequency accuracy. For example, a longer data sequence can use a larger number of points to improve resolution. Before the transformation, a window function, such as the Hanning window, is applied to reduce spectral leakage. The parameters of the window function are adjusted according to the signal characteristics; for example, the window length is consistent with the data length. The output of the power flow spectrum includes amplitude and phase spectra. The amplitude spectrum represents the power magnitude of each frequency component, and the unit can be per unit or a practical unit such as kilowatt. The phase spectrum represents phase information, and the unit is radians. The control command spectrum also outputs amplitude and phase spectra. The amplitude unit is set according to the command type; for example, the power command value is kilowatts per second. Frequency resolution is calculated from the sampling frequency and the number of transformation points. For example, with a sampling frequency of 1 Hz and 1024 transformation points, the frequency resolution is approximately 0.001 Hz. The spectrum data is stored in complex form, including real and imaginary parts, for subsequent calculations.

[0066] Next, the power flow spectrum and control command spectrum are divided in the frequency domain according to frequency resolution. For each frequency point, the conditional probability transfer entropy from historical smoothed control commands to historical power flow is calculated. The information transfer entropy value of each frequency point is obtained by comparing the logarithmic expectation values ​​of the conditional probability and the unconditional probability. The frequency division is based on the frequency axis of the spectrum, starting from zero frequency and ending at the Nyquist frequency, dividing each frequency point at frequency resolution intervals. For example, the frequency point sequence is 0 Hz, 0.001 Hz, 0.002 Hz, etc., up to the Nyquist frequency of 0.5 Hz. For each frequency point, the calculation of the conditional probability transfer entropy involves probability estimation. The conditional probability is defined as the probability of the power flow spectrum value given a control command spectrum value, and the unconditional probability is defined as the marginal probability of the power flow spectrum value.

[0067] The probability estimation employs a histogram method, dividing the control command spectrum and power flow spectrum into multiple intervals. The number of intervals is determined based on the data volume and distribution characteristics; for example, the Sturgess rule is used to determine the number of intervals. The frequency of spectrum values ​​falling into each interval in historical data is statistically analyzed to estimate the probability distribution. The expected logarithmic value is obtained by calculating a weighted average of the probability logarithms, with the weights corresponding to the probability values. Specifically, the natural logarithm is used in the calculation. The conditional probability transfer entropy value represents the mutual information between the conditional and unconditional probability distributions, achieved by multiplying the logarithm of the sum of the ratios of conditional and unconditional probabilities by the conditional probability. During the calculation, the input parameters are the amplitude values ​​of the power flow spectrum and control command spectrum, derived from the frequency domain transformation output. Constraints include that the spectrum values ​​must be real and non-negative, and the output information transfer entropy value is a scalar, expressed in nanots, used to quantify the information transfer intensity. Data preprocessing includes spectral value normalization, such as scaling amplitude values ​​to between 0 and 1 to avoid numerical overflow. Post-processing includes entropy smoothing, such as using moving average filtering to reduce fluctuations. The filter window size is set according to the frequency resolution, for example, a window size of 5 frequency points. Boundary condition handling includes avoiding zero-probability cases, such as adding a small offset, such as 0.001, to the probability estimation to prevent invalid logarithmic calculations.

[0068] Finally, the information transfer entropy values ​​corresponding to all frequencies are combined to form an information transfer entropy distribution. This combination process is achieved by creating a mapping structure between frequency and entropy value. For example, a two-dimensional array can be constructed, where one dimension stores the frequency values ​​and the other dimension stores the information transfer entropy value at the corresponding frequency point. The frequency values ​​are arranged in ascending order, from zero frequency to the Nyquist frequency, increasing with frequency resolution. The information transfer entropy values ​​are filled into the array in the order of calculation, forming a continuous distribution. The distribution data can be stored in list or matrix format for analysis in subsequent steps. The horizontal axis of the information transfer entropy distribution represents frequency, and the vertical axis represents the information transfer entropy value. The distribution characteristics reflect the changing patterns of information transfer from historical smoothing control commands to historical power flow at different frequencies, providing a basis for identifying the dominant oscillation mode frequency.

[0069] S3. Based on the information transmission entropy distribution, identify the frequency corresponding to the extreme value of information transmission entropy as the dominant oscillation mode frequency of the power grid within the photovoltaic power station cluster. The specific implementation is as follows:

[0070] In the implementation of the power fluctuation smoothing control method for photovoltaic power plant clusters, a dominant oscillation mode frequency identification step is performed. This step identifies the frequency corresponding to the extreme value of information transfer entropy as the dominant oscillation mode frequency of the internal power grid of the photovoltaic power plant cluster based on the information transfer entropy distribution. First, all local extreme points are located in the information transfer entropy distribution. The information transfer entropy distribution is derived from the output of the information transfer entropy distribution analysis step, and this distribution is a sequence of correspondences between frequency values ​​and information transfer entropy values. The local extreme points are located using a sliding window comparison method. The window size is set based on the frequency resolution of the information transfer entropy distribution, which is derived from the calculation results of the frequency domain transformation step. For example, if the frequency resolution is 0.001 Hz, the window size is set to 5 frequency points. The window size can be adjusted from 3 to 11 points, depending on the frequency resolution and the smoothness of the distribution curve. For example, a higher frequency resolution corresponds to a smaller window to capture details, while a lower frequency resolution corresponds to a larger window to smooth noise.

[0071] The sliding window comparison method involves traversing each frequency point in the information transfer entropy distribution. For each frequency point, its information transfer entropy value is compared with the information transfer entropy values ​​of its neighboring frequency points within the window. The number of neighboring frequency points is determined by the window size; for example, when the window size is 5, the information transfer entropy values ​​of the current point are compared with the two points before and after it. If the information transfer entropy value of the current point is greater than the information transfer entropy values ​​of all neighboring points, the point is identified as a local maximum. If the information transfer entropy value of the current point is less than the information transfer entropy values ​​of all neighboring points, the point is identified as a local minimum. Boundary frequency points are handled using a one-sided comparison method; for example, the first frequency point is compared only with its subsequent neighboring points, and the last frequency point is compared only with its preceding neighboring points. All located local extrema include both frequency value and information transfer entropy value. The frequency value is measured in Hertz, and the information transfer entropy value is measured in Nats. The local extrema are stored in a list format for subsequent filtering steps.

[0072] Next, extreme points that meet the preset salience criteria are selected from the local extreme points as information transmission entropy extremes. The preset salience criteria consist of a relative salience threshold and an absolute entropy threshold. The relative salience threshold is used to evaluate the relative significance of the extreme points, while the absolute entropy threshold ensures the absolute strength of the extreme points. The relative salience threshold is set based on the statistical characteristics of local extreme points in the historical information transmission entropy distribution. For example, the average difference in information transmission entropy values ​​of all local extreme points in historical data is calculated, and 1.5 times this average is used as the relative salience threshold. The threshold can be adjusted from 1.2 times to 2.0 times, depending on the characteristics of the distribution and the required recognition sensitivity. For example, a lower threshold is used when the distribution is relatively smooth to retain more extreme points. The absolute entropy threshold is set based on the global distribution of information transmission entropy values. For example, the percentile of all information transmission entropy values ​​is calculated, and the 75th percentile is used as the absolute entropy threshold. The threshold can be adjusted from the 60th to the 90th percentile, depending on the required significance level. For example, a higher percentile is used when higher significance is required.

[0073] The screening process involves calculating the relative salience of each local extremum. Relative salience is defined as the ratio of the minimum difference between the information transmission entropy value of the extremum and the information transmission entropy values ​​of its two adjacent extremums. The selection of adjacent extremums is based on frequency distance; for example, selecting the two extremums closest to the current extremum. Then, the relative salience is compared to a relative salience threshold, and simultaneously, the information transmission entropy value is checked to ensure it exceeds an absolute entropy threshold. Only local extremums that simultaneously meet both conditions are selected as information transmission entropy extrema. If no local extremum meets the preset salience criteria, an alternative strategy is employed, such as selecting the local extremum with the largest information transmission entropy value as the information transmission entropy extremum, ensuring that at least one extremum is selected. The screening results are stored as a list of information transmission entropy extrema, including frequency values ​​and information transmission entropy values.

[0074] Finally, the frequencies corresponding to the extreme values ​​of information transfer entropy on the horizontal axis of the information transfer entropy distribution are determined as the dominant oscillation mode frequencies of the internal power grid of the photovoltaic power plant cluster. The horizontal axis of the information transfer entropy distribution represents the frequency axis, with frequency values ​​being a continuous sequence from zero frequency to the Nyquist frequency. For each extreme value of information transfer entropy, its corresponding frequency value is extracted, and this frequency value is directly used as the dominant oscillation mode frequency. The unit of the dominant oscillation mode frequency is Hertz, representing the characteristic frequency of a significant oscillation mode existing in the internal power grid of the photovoltaic power plant cluster. During the determination process, if there are multiple extreme values ​​of information transfer entropy, there are multiple corresponding dominant oscillation mode frequencies, which are arranged in ascending order of frequency value to form a frequency list. The accuracy of the dominant oscillation mode frequency is determined by the frequency resolution of the information transfer entropy distribution; for example, if the frequency resolution is 0.001 Hertz, the accuracy of the dominant oscillation mode frequency is 0.001 Hertz. All determined dominant oscillation mode frequencies are stored in sequence format, and the corresponding information transfer entropy values ​​are recorded for subsequent analysis and control steps, such as evaluating the nonlinear interaction strength of smoothing control loops. If the list of extreme values ​​for information transfer entropy is empty, the frequency of the dominant oscillation mode is set to empty, and a process for re-analyzing the information transfer entropy distribution is triggered to ensure data availability.

[0075] S4. Evaluate the nonlinear interaction strength of the smooth control loops of different sub-power stations in the photovoltaic power station cluster at the dominant oscillation mode frequency, and calculate the spatiotemporal distribution inertia center of oscillation energy of each node under the dominant oscillation mode. Specifically, this is implemented as follows:

[0076] In the implementation of the power fluctuation smoothing control method for photovoltaic power plant clusters, the steps of nonlinear interaction intensity assessment and spatiotemporal distribution inertia center calculation are performed. First, the dynamic response signal of the smoothing control loop of each sub-power plant near the dominant oscillation mode frequency is extracted. The dominant oscillation mode frequency is derived from the output of the dominant oscillation mode frequency identification step, and the dynamic response signal comes from the real-time monitoring data of the smoothing control loop of each sub-power plant, including power output response and control command response. The extraction process uses a digital bandpass filtering method. The filter is designed as a Butterworth filter with an order of 4. The filter center frequency is set to the dominant oscillation mode frequency; for example, if the dominant oscillation mode frequency is 1 Hz, the center frequency is set to 1 Hz. The bandwidth is set to 10% of the dominant oscillation mode frequency, for example, 0.1 Hz. The bandwidth adjustment range can be from 5% to 20%, based on signal processing theory and practical application requirements. For example, a narrower bandwidth is used for accurate frequency analysis, while a wider bandwidth is used to cover frequency fluctuations. The bandwidth setting is determined by comparing the deviation of the actual signal frequency components with the center frequency; signal components with deviations within the bandwidth range are retained. The filtered dynamic response signal retains the oscillation components near the dominant oscillation mode frequency. The signal length is determined according to the analysis requirements, for example, taking data points corresponding to 10 oscillation cycles. The signal sampling frequency is consistent with the data acquisition steps, for example, 1 kHz. The extracted dynamic response signal is stored in time series format for subsequent nonlinear interaction strength assessment.

[0077] Next, the nonlinear interaction strength of the smoothing control loop at the dominant oscillation mode frequency is evaluated by calculating the nonlinear correlation coefficient between the dynamic response signals. Phase space reconstruction is performed on the dynamic response signals of the smoothing control loops of each sub-station. The parameters for phase space reconstruction include the embedding dimension and time delay. The embedding dimension is determined using the spurious nearest neighbor method based on signal characteristics. For example, the proportion of spurious nearest neighbors under different embedding dimensions is calculated, and the embedding dimension is determined when the proportion is lower than a preset threshold. The preset threshold is set to 5%, and its setting is based on standard practices for phase space reconstruction, such as selecting the minimum proportion that can stably reconstruct the phase space through historical data testing. The preset threshold can be adjusted from 3% to 10%, depending on the signal noise level and reconstruction accuracy requirements. The comparison method is to calculate the proportion of spurious nearest neighbors under each embedding dimension; if the proportion is lower than the preset threshold, the embedding dimension is stopped from increasing. The time delay is determined using the autocorrelation function method, for example, by calculating the time delay when the autocorrelation function drops to 1 / e of its initial value. The time delay ranges from 1 to 10 sampling points. The mutual information between each pair of loops is calculated based on the reconstructed state vectors. The mutual information calculation uses a histogram estimation method, dividing the state vectors into multiple intervals. The number of intervals is determined by the Sturgess rule based on the data volume; for example, 15 intervals are used when there are 1000 data points. The nonlinear correlation coefficient is obtained by normalizing the mutual information by dividing it by the geometric mean of the entropy values ​​of the two signals. The nonlinear correlation coefficient ranges from zero to one. The nonlinear interaction strength is obtained by combining the nonlinear correlation coefficients of all loops. The combination method is to calculate the weighted average of the nonlinear correlation coefficients of all loops. The weights of the weighted average are set according to the rated power capacity of each loop; for example, a loop with a rated power capacity of 1 MW has a weight of 1, and a loop with a rated power capacity of 2 MW has a weight of 2. The weights are set based on the relative importance of each loop in the cluster, and the weights are proportional to the rated power capacity to ensure that larger capacity loops contribute more to the nonlinear interaction strength. The nonlinear interaction strength is a dimensionless value used to characterize the coupling degree between smoothing control loops.

[0078] Simultaneously, node oscillation energy is calculated based on the oscillation components of voltage and current at each node. The oscillation components of node voltage and current are extracted using bandpass filtering, with filtering parameters consistent with those used in the dynamic response signal extraction. The center frequency is set to the dominant oscillation mode frequency, and the bandwidth is set to 10% of the dominant oscillation mode frequency. The node oscillation energy is calculated using the instantaneous energy method, defined as the integral of the product of the voltage oscillation component and the current oscillation component. The integration interval is a complete oscillation cycle; for example, when the dominant oscillation mode frequency is 1 Hz, the integration interval is 1 second. The unit of node oscillation energy is joules, representing the degree of energy participation of the node in the dominant oscillation mode. During energy calculation, the phase relationship between voltage and current must be considered. The instantaneous phase is calculated using Hilbert transform to ensure accuracy. When the phase difference between voltage and current exceeds 90 degrees, the energy value is treated as an absolute value. The 90-degree setting is based on the power theory that a phase difference exceeding 90 degrees indicates a reversal of energy flow direction, determined by comparing the instantaneous phase difference with 90 degrees.

[0079] Finally, a weighted average of the node position coordinates is performed using the node oscillation energy as the weight to obtain the spatiotemporal distribution inertia center of each node's oscillation energy under the dominant oscillation mode. First, the instantaneous oscillation energy of each node under the dominant oscillation mode is calculated, and this instantaneous oscillation energy is used as the weighting coefficient. The node position coordinates in the power grid topology are obtained from the power grid geographic information system or topology database, using a Cartesian coordinate system with meters as the unit. The weighted summation process involves multiplying the position coordinates of each node by its node oscillation energy, summing the results, and then dividing by the total oscillation energy to obtain the coordinate value of the spatiotemporal distribution inertia center. The total oscillation energy is the sum of the oscillation energies of all nodes. The coordinate value of the spatiotemporal distribution inertia center represents the distribution center of the oscillation energy in the power grid space and is used to analyze the spatial propagation characteristics of the oscillation. During the calculation, it is necessary to ensure that the node oscillation energy is non-negative; if a negative value occurs, its absolute value is taken. The result of the spatiotemporal distribution inertia center is stored in coordinate pairs for subsequent construction of system dynamic instability criteria.

[0080] S5. Based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center, a dynamic instability criterion for the system is constructed, specifically as follows:

[0081] In the implementation of the power fluctuation smoothing control method for photovoltaic power plant clusters, a step of constructing a system dynamic instability criterion is performed. This step constructs a system dynamic instability criterion based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center. First, the nonlinear interaction strength is compared with a preset interaction threshold. The nonlinear interaction strength comes from the output of the nonlinear interaction strength evaluation step and is a dimensionless value characterizing the degree of coupling between smoothing control loops. The preset interaction threshold is set based on the statistical characteristics of nonlinear interaction intensity under historical stable operating conditions. Specifically, it is achieved by collecting time series of nonlinear interaction intensity from historical data, calculating the mean and standard deviation of the time series, and then adding twice the standard deviation to the mean as the preset interaction threshold. For example, if the historical data mean is 0.5 and the standard deviation is 0.1, the preset interaction threshold is 0.7. The threshold can be adjusted from 1.5 times the standard deviation to 3 times the standard deviation, depending on the system's stability requirements and risk tolerance. For example, a higher multiple, such as 3 times the standard deviation, is used in scenarios requiring high stability to reduce the risk of misjudgment, while a lower multiple, such as 1.5 times the standard deviation, is used in scenarios that allow for some fluctuation to improve detection sensitivity. The comparison method is to directly compare the current nonlinear interaction intensity value with the preset interaction threshold. If the current value is greater than the preset interaction threshold, the first criterion condition is met.

[0082] Simultaneously, the analysis examines whether the offset trajectory of the spatiotemporal distributed inertia center exhibits a continuous divergence characteristic. The spatiotemporal distributed inertia center is derived from the output of the calculation step and is a coordinate sequence arranged in chronological order. Offset trajectory analysis is achieved by calculating the movement distance and directional change characteristics of the spatiotemporal distributed inertia center within a continuous time window. The movement distance is calculated using the Euclidean distance formula, calculating the straight-line distance between coordinate values ​​of adjacent time points, with the distance unit consistent with the coordinate unit. The directional change characteristic is evaluated by calculating the angle change of continuous movement vectors, for example, calculating the angle between two vectors formed by three adjacent time points. The threshold for the angle change rate is set to 30 degrees per time step, and the threshold can be adjusted from 20 degrees to 45 degrees, based on the system oscillation characteristics and trajectory stability requirements. For example, a higher threshold is used in systems with frequent oscillations to reduce the impact of noise, while a lower threshold is used in stable systems to capture subtle changes. The comparison method is to calculate the directional change rate at each time step and compare it with the directional change rate threshold. If the directional change rate exceeds the threshold, the directional change is considered significant. The determination of a continuous divergence feature requires the simultaneous fulfillment of two conditions: a continuous increase in the moving distance and a change rate in direction exceeding a set threshold. The determination of a continuous increase in the moving distance is achieved by comparing the difference in moving distance over consecutive time periods. For example, if the difference in moving distance over three consecutive time periods is greater than zero and the magnitude of the difference exceeds the distance change threshold, the threshold is set to 10% of the initial moving distance. The threshold can be adjusted from 5% to 20%, depending on the system's sensitivity to distance changes and the accuracy requirements of trajectory analysis. For example, a lower threshold is used in scenarios requiring rapid response to detect divergence as early as possible, while a higher threshold is used in scenarios requiring accuracy to avoid false alarms. The comparison method is to calculate the difference in moving distance over each time period and compare it with the distance change threshold. If the difference exceeds the threshold and the trend continues, it is considered that the moving distance is continuously increasing.

[0083] The system dynamic instability criterion is established when the nonlinear interaction strength exceeds a preset interaction threshold and the offset trajectory of the spatiotemporal inertia center simultaneously exhibits a continuous divergence characteristic. The determination process employs a logical AND operation, meaning both conditions must be met simultaneously. The establishment of the system dynamic instability criterion indicates that the photovoltaic power station cluster is in a dynamically unstable state, requiring the initiation of subsequent control measures. A data verification step is included in the criterion determination process, such as checking the completeness and temporal continuity of the input data. If the data is incomplete, the determination is delayed until the data meets the requirements. All determination results are recorded, including the determination timestamp, nonlinear interaction strength value, offset trajectory parameters, and the final criterion state, for subsequent control decisions and historical data analysis.

[0084] S6. When the system's dynamic instability criterion meets the preset conditions, a dynamic safety constraint set is constructed with the goal of suppressing oscillation modes. Under the constraints of the dynamic safety constraint set, an optimization model for smooth control commands is solved, and new smooth control commands are generated and executed to each sub-power station in the photovoltaic power station cluster. The specific implementation is as follows:

[0085] In the implementation of the power fluctuation smoothing control method for photovoltaic power plant clusters, when the system dynamic instability criterion meets the preset conditions, the steps of constructing a dynamic safety constraint set and optimizing smoothing control commands are executed. First, when the system dynamic instability criterion is established, a dynamic safety constraint set limiting the rate of change of control commands is constructed based on the dominant oscillation mode frequency. The system dynamic instability criterion comes from the output of the system dynamic instability criterion construction step, and the dominant oscillation mode frequency comes from the output of the dominant oscillation mode frequency identification step. The construction of the dynamic safety constraint set is achieved by setting upper and lower limits for the rate of change of control commands. The calculation of the rate of change constraint value is based on the reciprocal relationship of the dominant oscillation mode frequency. For example, when the dominant oscillation mode frequency is 1 Hz, the rate of change constraint value is set to 0.5 times the rated power per second. The adjustment range of the rate of change constraint value can be from 0.3 times to 0.8 times the rated power per second, depending on the system stability and response speed requirements. For example, in scenarios with high stability requirements, a stricter constraint value, such as 0.3 times the rated power per second, is used to better suppress oscillations; in scenarios with high response speed requirements, a looser constraint value, such as 0.8 times the rated power per second, is used to accelerate the control response. The constraint set is specifically in the form that the absolute value of the rate of change of the control command is less than or equal to the rate of change constraint value. This constraint condition applies to the control commands of all sub-power stations to ensure that the control command changes smoothly and does not trigger oscillation modes.

[0086] Next, under the constraints of the dynamic security constraint set, an optimization model for smooth control commands is solved with the objective function of minimizing the total power fluctuation of the cluster, resulting in new smooth control commands for each sub-station. The objective function is defined as the sum of squares of the deviations between the total power output of the cluster and the target power, with the target power determined based on grid dispatch commands or predicted load. The decision variables of the optimization model are the smooth control command values ​​of each sub-station. The constraints include the rate of change constraint defined by the dynamic security constraint set and the power output capacity constraint of each sub-station, which is determined based on the real-time available power and equipment limits of the sub-station. The optimization solution employs a quadratic programming algorithm, which iteratively calculates to find the optimal solution that satisfies the constraints and minimizes the objective function. The iteration termination condition is set to the relative change of the objective function value between two adjacent iterations being less than 0.001 percentage points. During the solution process, the weight coefficient of each sub-power station is set according to its response characteristics. For example, sub-power stations with faster response speeds are given higher weights to better suppress fluctuations. The weight coefficient adjustment range is from 0.5 to 2.0, based on the dynamic response capability of each sub-power station and its importance in the cluster. The weight coefficient is specifically determined through expert evaluation. For example, system operators score the response capability of each sub-power station based on historical operating experience, and the weight coefficient is obtained after normalizing the scores.

[0087] Finally, the new smooth control command is issued to each sub-station in the photovoltaic power station cluster for execution. Command issuance is achieved through power system communication protocols, such as the IEC 60870-5-104 protocol for data transmission. The execution process includes command verification and confirmation. After receiving the command, each sub-station verifies the command format and numerical range. If correct, it performs power adjustment. Execution results are fed back in real time through the monitoring system. If an abnormal command execution is detected, a retransmission mechanism is initiated, limited to three retransmissions. Exceeding this limit switches to manual intervention mode. The time delay of the entire command generation and execution process is controlled within one second to ensure real-time control and effectiveness. Before command issuance, data preprocessing is performed to check if the command value is within the sub-station's allowable operating range. If it exceeds the range, a limiting effect is applied, with the limiting value determined based on the sub-station's equipment parameters.

[0088] Example 2: Figure 2 A schematic diagram of the power fluctuation smoothing control system for a photovoltaic power plant cluster according to the present invention is provided. The power fluctuation smoothing control system for a photovoltaic power plant cluster includes:

[0089] The information acquisition module is used to acquire the historical power flow and corresponding historical smoothing control commands of the power grid within the photovoltaic power plant cluster.

[0090] The distribution analysis module is used to analyze the information transfer entropy distribution between historical power flow and historical smoothing control commands at different frequencies, based on historical power flow and historical smoothing control commands.

[0091] The dominant identification module is used to identify the frequency corresponding to the extreme value of information transmission entropy as the dominant oscillation mode frequency of the internal power grid of the photovoltaic power plant cluster based on the information transmission entropy distribution.

[0092] The information analysis module is used to evaluate the nonlinear interaction strength of the smooth control loops of different sub-power stations in the photovoltaic power station cluster at the frequency of the dominant oscillation mode, and to calculate the spatiotemporal distribution of the oscillation energy of each node under the dominant oscillation mode.

[0093] The criterion construction module is used to construct dynamic instability criteria for the system based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center.

[0094] The instruction execution module is used to construct a dynamic safety constraint set with the goal of suppressing oscillation modes when the system dynamic instability criterion meets the preset conditions, and solve the optimization model of the smooth control instruction under the constraints of the dynamic safety constraint set, and generate and execute new smooth control instructions to each sub-power station in the photovoltaic power station cluster.

[0095] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0096] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0097] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0098] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0099] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0100] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0101] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0102] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0103] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0104] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for smoothing power fluctuations in a photovoltaic power plant cluster, characterized in that, include: S1. Obtain the historical power flow and corresponding historical smoothing control commands of the internal power grid of the photovoltaic power station cluster. S2. Based on historical power flow and historical smoothing control commands, analyze the information transfer entropy distribution between historical smoothing control commands and historical power flow at different frequencies, including: The historical power flow and historical smoothing control commands are transformed in the frequency domain to obtain the power flow spectrum and the control command spectrum, respectively. The power flow spectrum and the control command spectrum are divided in the frequency domain according to frequency resolution. For each frequency point, the conditional probability transfer entropy from the historical smooth control command to the historical power flow is calculated. The information transfer entropy value of each frequency point is obtained by comparing the logarithmic expectation values ​​of the conditional probability and the unconditional probability. The information transmission entropy values ​​corresponding to all frequencies are combined to form an information transmission entropy distribution. S3. Based on the information transmission entropy distribution, identify the frequency corresponding to the extreme value of information transmission entropy as the dominant oscillation mode frequency of the internal power grid of the photovoltaic power station cluster. S4. Evaluate the nonlinear interaction strength of the smooth control loops of different sub-power stations in the photovoltaic power station cluster at the frequency of the dominant oscillation mode, and calculate the spatiotemporal distribution inertia center of the oscillation energy of each node under the dominant oscillation mode. S5. Based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center, construct a dynamic instability criterion for the system; S6. When the dynamic instability criterion of the system meets the preset conditions, a dynamic safety constraint set with the goal of suppressing oscillation modes is constructed, and an optimization model of smooth control instructions is solved under the constraints of the dynamic safety constraint set. New smooth control instructions are generated and executed to each sub-power station in the photovoltaic power station cluster.

2. The power fluctuation smoothing control method for photovoltaic power plant clusters according to claim 1, characterized in that, Obtain historical power flow data and corresponding historical smoothing control commands from the internal power grid of the photovoltaic power plant cluster, including: Historical power flow data are collected from monitoring devices within the power grid of a photovoltaic power plant cluster. And extract historical smooth control commands from the control command record; Historical power flow and historical smoothing control commands are matched and aligned using timestamps to ensure the time correspondence and integrity of the data.

3. The power fluctuation smoothing control method for photovoltaic power plant clusters according to claim 1, characterized in that, Based on the information transmission entropy distribution, the frequency corresponding to the extreme value of information transmission entropy is identified as the dominant oscillation mode frequency of the power grid within the photovoltaic power plant cluster, including: Locate all local extreme points in the entropy distribution of information transmission; The extreme points that satisfy the preset salience condition are selected from the local extreme points as the extreme values ​​of information transmission entropy; The frequency corresponding to the extreme value of information transmission entropy on the horizontal axis of the information transmission entropy distribution is determined as the dominant oscillation mode frequency of the power grid within the photovoltaic power plant cluster.

4. The power fluctuation smoothing control method for photovoltaic power plant clusters according to claim 1, characterized in that, The nonlinear interaction strength of smoothed control loops of different sub-power stations in a photovoltaic power plant cluster is evaluated at the dominant oscillation mode frequency, and the spatiotemporal distribution inertia center of oscillation energy of each node under the dominant oscillation mode is calculated, including: Extract the dynamic response signal of the smoothing control loop of each sub-power station near the dominant oscillation mode frequency; The strength of the nonlinear interaction of the smooth control loop at the dominant oscillation mode frequency is evaluated by calculating the nonlinear correlation coefficient between the dynamic response signals. At the same time, the node oscillation energy is calculated based on the oscillation components of the voltage and current at each node; By using the node oscillation energy as the weight, a weighted average of the node position coordinates is performed to obtain the spatiotemporal distribution inertia center of each node's oscillation energy under the dominant oscillation mode.

5. The power fluctuation smoothing control method for photovoltaic power plant clusters according to claim 4, characterized in that, The nonlinear interaction strength of the smooth control loop at the dominant oscillation mode frequency is evaluated by calculating the nonlinear correlation coefficient between dynamic response signals. This includes: reconstructing the phase space of the dynamic response signals of the smooth control loop of each sub-station; calculating the mutual information between each pair of loops based on the reconstructed state vectors; obtaining the nonlinear correlation coefficient by normalizing the mutual information; and obtaining the nonlinear interaction strength by combining the nonlinear correlation coefficients between all loops.

6. The power fluctuation smoothing control method for photovoltaic power plant clusters according to claim 4, characterized in that, The spatiotemporal distribution inertia center of each node's oscillation energy under the dominant oscillation mode is obtained by weighting the node's position coordinates with the node's oscillation energy as the weight. This includes: first, calculating the instantaneous oscillation energy of each node under the dominant oscillation mode; using the instantaneous oscillation energy as a weighting coefficient; then, weighting and summing the node's position coordinates in the power grid topology; and finally, dividing by the total oscillation energy to obtain the coordinate value of the spatiotemporal distribution inertia center.

7. The power fluctuation smoothing control method for photovoltaic power plant clusters according to claim 1, characterized in that, Based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center, a dynamic instability criterion for the system is constructed, including: The nonlinear interaction strength is compared with a preset interaction threshold, and the offset trajectory of the spatiotemporal distribution inertia center is analyzed to see if it exhibits a continuous divergence characteristic. When the nonlinear interaction intensity exceeds the preset interaction threshold and the offset trajectory of the spatiotemporal distribution inertia center simultaneously exhibits continuous divergence characteristics, the dynamic instability criterion of the system is determined to be valid.

8. The power fluctuation smoothing control method for photovoltaic power plant clusters according to claim 1, characterized in that, When the system's dynamic instability criterion meets preset conditions, a dynamic safety constraint set aimed at suppressing oscillating modes is constructed. Under the constraints of this dynamic safety constraint set, an optimization model for smooth control commands is solved, and new smooth control commands are generated and executed to each sub-power station in the photovoltaic power station cluster, including: When the dynamic instability criterion of the system is met, a dynamic safety constraint set that limits the rate of change of control commands is constructed based on the frequency of the dominant oscillation mode. Under the constraints of the dynamic safety constraint set, an optimization model for solving the smooth control command is used to minimize the total power fluctuation of the cluster as the objective function, and new smooth control commands for each sub-power station are obtained. The new smooth control command is issued to each sub-power station in the photovoltaic power station cluster for execution.

9. A power fluctuation smoothing control system for a photovoltaic power plant cluster, used to implement the power fluctuation smoothing control method for a photovoltaic power plant cluster as described in any one of claims 1-8, characterized in that, include: The information acquisition module is used to acquire the historical power flow and corresponding historical smoothing control commands of the power grid within the photovoltaic power plant cluster. The distribution analysis module is used to analyze the information transfer entropy distribution between historical power flow and historical smoothing control commands at different frequencies, based on historical power flow and historical smoothing control commands. The dominant identification module is used to identify the frequency corresponding to the extreme value of information transmission entropy as the dominant oscillation mode frequency of the internal power grid of the photovoltaic power plant cluster based on the information transmission entropy distribution. The information analysis module is used to evaluate the nonlinear interaction strength of the smooth control loops of different sub-power stations in the photovoltaic power station cluster at the frequency of the dominant oscillation mode, and to calculate the spatiotemporal distribution of the oscillation energy of each node under the dominant oscillation mode. The criterion construction module is used to construct dynamic instability criteria for the system based on the nonlinear interaction strength and the offset trajectory of the spatiotemporal distribution inertia center. The instruction execution module is used to construct a dynamic safety constraint set with the goal of suppressing oscillation modes when the system dynamic instability criterion meets the preset conditions, and solve the optimization model of the smooth control instruction under the constraints of the dynamic safety constraint set, and generate and execute new smooth control instructions to each sub-power station in the photovoltaic power station cluster.

Images