A new energy station voltage and frequency coupling collaborative control performance online evaluation method

Abstract

The application is suitable for the technical field of data processing, and provides a new energy station voltage and frequency coupling collaborative control performance online evaluation method, which comprises the following steps: collecting three-phase voltage signals, frequency signals, unit output power signals and power grid dispatching instruction signals of a new energy station grid-connected point, and calculating a voltage and frequency coupling characteristic matrix; generating a collaborative parameter matrix through collaborative quantitative analysis; performing control performance evaluation on the collaborative parameter matrix based on a collaborative evaluation network structure, and generating collaborative evaluation parameters; constructing a dynamic coupling constraint function, and calculating a collaborative stability boundary representing a station operation state; and performing dynamic correction on the collaborative evaluation parameters, and generating a station control performance optimization sequence. The safety, stability and power quality of the new energy station operation are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a new energy station voltage and frequency coupling collaborative control performance online evaluation method. BACKGROUND

[0002] In the traditional power system, the synchronous generator can effectively maintain the stability of voltage and frequency by virtue of its own inertia and regulation characteristics. However, the characteristics of power electronic devices such as wind turbine generators and photovoltaic inverters in new energy stations are significantly different from those of traditional synchronous generators, and they lack inherent inertia response and frequency regulation and voltage regulation capabilities. When the power grid is disturbed, new energy stations cannot timely and effectively participate in voltage and frequency regulation, which easily leads to excessive voltage fluctuation and frequency deviation, seriously affecting the power quality and operation stability of the power grid.

[0003] At present, the control methods for voltage and frequency of new energy stations are mostly independent, without fully considering the coupling relationship between voltage and frequency. This control method is difficult to achieve precise collaborative control of voltage and frequency under complex working conditions, and cannot meet the growing stability demand of the power grid. For example, when the active power of the new energy station changes rapidly, not only will it cause frequency fluctuation, but also will affect voltage stability through the power flow distribution of the power system; similarly, voltage fluctuations will also affect the output of new energy units, and thus affect the frequency. In addition, the existing control performance evaluation methods are usually offline, which cannot reflect the control performance of new energy stations under different operating conditions in real time, and it is difficult to discover and solve potential problems in a timely manner.

[0004] In view of this, a new energy station voltage and frequency coupling collaborative control performance online evaluation method is proposed. SUMMARY

[0005] The present application provides a new energy station voltage and frequency coupling collaborative control performance online evaluation method, which solves the problem that the control performance of new energy stations under different operating conditions cannot be reflected in real time, and it is difficult to discover and solve potential problems in a timely manner.

[0006] The present application provides a new energy station voltage and frequency coupling collaborative control performance online evaluation method, which comprises:

[0007] Collecting three-phase voltage signals, frequency signals, unit output power signals and grid dispatching instruction signals of the new energy station grid-connected point, and calculating a voltage and frequency coupling characteristic matrix based on the collected signals;

[0008] Performing collaborative quantization analysis on the voltage and frequency coupling characteristic matrix to generate a collaborative parameter matrix;

[0009] The control performance is evaluated based on the collaborative evaluation network structure to generate collaborative evaluation parameters;

[0010] Based on the collaborative evaluation parameters, a dynamic coupling constraint function is constructed, and the collaborative stability boundary characterizing the station's operating state is calculated.

[0011] Based on the mapping relationship between the voltage and frequency coupling feature matrix and the cooperative stability boundary, the cooperative evaluation parameters are dynamically corrected to generate a station control performance optimization sequence.

[0012] Furthermore, the process of collecting three-phase voltage signals, frequency signals, generator output power signals, and grid dispatch command signals from the grid connection point of the new energy power station, and calculating the voltage-frequency coupling feature matrix based on the collected signals, includes:

[0013] Multi-channel synchronous acquisition of three-phase voltage signals, frequency signals, unit output power signals, and grid dispatch command signals at the grid connection point of new energy power plants is performed to generate a coupled observation data matrix.

[0014] The coupled observation data matrix is ​​decomposed in the frequency domain to obtain the fundamental frequency component matrix and the high-frequency disturbance component matrix. The fundamental frequency component matrix is ​​then calculated based on the frequency domain energy integral to generate the initial element values ​​of the coupling degree feature matrix.

[0015] Dynamic phase matching is performed on the initial element values ​​of the coupling degree feature matrix, and the coupling degree feature matrix is ​​corrected by introducing a coupling compensation coefficient to generate a corrected coupling degree feature matrix.

[0016] Based on the multi-source data fusion algorithm, the modified coupling degree feature matrix and the power grid state matrix are correlated and analyzed to generate a coupling response matrix. Then, the coupling response matrix is ​​jointly calculated in the time and frequency domains to generate voltage and frequency coupling feature matrices under different operating modes.

[0017] Furthermore, the step of performing frequency domain decomposition on the coupled observation data matrix to obtain the fundamental frequency component matrix and the high-frequency disturbance component matrix includes:

[0018] The coupled observation data matrix is ​​decomposed into full-band components using discrete Fourier transform to generate initial spectral components;

[0019] Based on the energy percentage threshold, the initial spectral components are filtered for fundamental frequency bands, and the components whose energy percentage in the main frequency band exceeds a preset value are extracted to form a fundamental frequency component matrix.

[0020] The remaining frequency band components are subjected to moving average filtering to generate a high-frequency disturbance component matrix, and the envelope of the high-frequency disturbance components is extracted by Hilbert transform.

[0021] Furthermore, the step of performing cooperative quantization analysis on the voltage and frequency coupling feature matrix to generate a cooperative parameter matrix includes:

[0022] The voltage and frequency coupling feature matrix is ​​dynamically divided into intervals to generate quantization interval parameters. Based on the quantization interval parameters, the coupling features in the voltage and frequency coupling feature matrix are adaptively thresholded to generate a dynamic quantization matrix.

[0023] The coupling features in the voltage and frequency coupling feature matrix are subjected to multimodal classification processing according to the dynamic quantization matrix to generate a classification feature matrix;

[0024] A sliding window statistical analysis is performed on the classification feature matrix, and a dynamic coupling interval is set to generate a statistical feature matrix;

[0025] The statistical feature matrix is ​​smoothed by passing it through a low-pass filter to generate a steady-state feature matrix, and then the steady-state feature matrix is ​​normalized to generate a cooperative parameter matrix.

[0026] Furthermore, based on the collaborative parameter matrix, a collaborative evaluation network structure is constructed, and the voltage and frequency coupling state of the new energy power station is evaluated in real time to generate collaborative evaluation parameters, including:

[0027] The collaborative parameter matrix is ​​divided into a main control feature matrix and an auxiliary feature matrix;

[0028] A master control network is constructed based on the master control feature matrix. The master control network includes an input layer, a hidden layer, and an output layer. The input layer contains voltage, frequency, power, and command signals. The hidden layer uses a dynamic nonlinear activation function. The output layer generates master control evaluation values ​​to obtain the output data of the master control network.

[0029] An auxiliary network is constructed based on the auxiliary feature matrix. The auxiliary network includes an input layer, a hidden layer, and an output layer. The input layer receives the output data of the main control network, the hidden layer uses a piecewise activation function, and the output layer generates auxiliary evaluation values ​​to obtain the output data of the auxiliary network.

[0030] The main control network and the auxiliary network are jointly trained to generate a collaborative evaluation network structure. Error backpropagation and network sensitivity analysis are then performed on the collaborative evaluation network structure to generate network evaluation results.

[0031] Based on the network evaluation results, the weight coefficients of the master network and auxiliary network in the collaborative evaluation network structure are updated, and the network parameters are adjusted by adaptive learning rate to generate collaborative evaluation parameters.

[0032] Furthermore, the step of performing error backpropagation and network sensitivity analysis on the collaborative evaluation network structure includes: constructing a composite loss function that includes mean squared error and cross-entropy loss, and calculating the error gradient of the network output data.

[0033] Furthermore, the step of constructing a dynamic coupling constraint function based on the collaborative evaluation parameters and calculating the collaborative stability boundary characterizing the station's operating state includes:

[0034] The collaborative evaluation parameters are reconstructed using multi-dimensional tensors, and a coupled state matrix is ​​generated through nonlinear mapping.

[0035] Based on the coupling state matrix, a dynamic energy integral term is designed, multi-interval operations are performed on the coupling variables to generate dynamic constraint terms, and the coupling state matrix and the dynamic constraint terms are combined to construct the station operation constraint function, thereby generating a dynamic coupling constraint function.

[0036] The frequency domain characteristics of the dynamic coupling constraint function are calculated to generate a frequency domain response result. The frequency domain response result is then compared with the amplitude-frequency characteristics of the running state to generate coupling constraint conditions.

[0037] The rated operating range of the new energy power station is dynamically divided according to the coupling constraint conditions, and a collaborative stability index is set to generate a constraint interval. Based on the constraint interval, boundary analysis is performed to generate a collaborative stability boundary for the operating state of the power station.

[0038] Furthermore, the step of dynamically correcting the collaborative evaluation parameters based on the mapping relationship between the voltage and frequency coupling characteristic matrix and the collaborative stability boundary, and generating a station control performance optimization sequence, includes:

[0039] Based on the voltage and frequency coupling feature matrix, an evaluation weight matrix and a compensation correction matrix are constructed, and the evaluation weight matrix and the compensation correction matrix are substituted into the collaborative optimization function to generate a performance optimization objective function;

[0040] Set control response constraints and dynamic adjustment limit conditions for the performance optimization objective function, and convert the cooperative stability boundary into state derivative constraints to generate performance optimization constraints.

[0041] The performance optimization objective function and the performance optimization constraints are input into the online optimization solver to calculate the correction parameters in the continuous time domain and generate the initial sequence of correction parameters.

[0042] The initial sequence of the correction parameters is dynamically adjusted in step size, and the adjustment rate is constrained in combination with the operating conditions of the new energy power station to generate the power station control performance optimization sequence.

[0043] Furthermore, the dynamic step size adjustment of the initial sequence of the correction parameters includes:

[0044] The initial adjustment step size is set based on the convergence rate of historical correction parameters;

[0045] The step size is dynamically reduced or increased based on the distance between the current correction parameter and the cooperative stability boundary, and the golden section method is used to perform nonlinear optimization of the step size adjustment process.

[0046] Furthermore, the method also includes:

[0047] The station control performance optimization sequence is encoded into instructions to generate control instruction signals, and the control instruction signals are then processed for power adaptation to generate execution drive signals.

[0048] Calculate the target voltage and frequency adjustment based on the power grid dispatch instructions, and monitor the operation status of the grid connection point in real time to generate operation monitoring data;

[0049] The difference between the operation monitoring data and the target voltage and frequency adjustment amount is calculated to generate an adjustment deviation value, and the adjustment deviation value is compared with a preset allowable deviation threshold to generate a deviation judgment result;

[0050] Based on the deviation determination result, the control parameters are corrected online to generate updated control parameters. The updated control parameters are then reconstructed into a sequence, and the control parameters are converted into adjustment command values ​​through state space mapping to generate a new control command sequence.

[0051] The new control command sequence is input into the station controller for closed-loop control, and the unit output is dynamically adjusted.

[0052] As can be seen from the above technical solutions, the present invention has the following advantages:

[0053] This invention calculates a voltage-frequency coupling characteristic matrix based on acquired signals, which more accurately reflects the actual operating status of renewable energy power plants compared to traditional methods of monitoring voltage and frequency separately. It performs collaborative quantification analysis on the voltage-frequency coupling characteristic matrix to generate a collaborative parameter matrix. Based on a collaborative evaluation network structure, it evaluates the control performance of the collaborative parameter matrix to generate collaborative evaluation parameters. This allows for a more comprehensive and accurate assessment of the voltage-frequency coupling state of renewable energy power plants, and the generated collaborative evaluation parameters are more valuable. A dynamic coupling constraint function is constructed based on the collaborative evaluation parameters to calculate the collaborative stability boundary characterizing the power plant's operating state. Based on the mapping relationship between the voltage-frequency coupling characteristic matrix and the collaborative stability boundary, the collaborative evaluation parameters are dynamically corrected to generate a power plant control performance optimization sequence. This achieves dynamic monitoring and control of the operating status of renewable energy power plants, enabling timely adjustments to control strategies. This ensures that renewable energy power plants remain within a stable operating range under complex and changing conditions, effectively preventing equipment damage and grid failures caused by excessive voltage or frequency fluctuations, and significantly improving the operational safety and stability of renewable energy power plants. Attached Figure Description

[0054] Figure 1 This is a schematic flowchart of an embodiment of an online evaluation method for voltage and frequency coupling coordinated control performance of a new energy power station according to the present invention;

[0055] Figure 2 This is a schematic flowchart of another embodiment of the online evaluation method for voltage and frequency coupling coordinated control performance of a new energy power station in this invention;

[0056] Figure 3 This is a schematic flowchart of another embodiment of the online evaluation method for voltage and frequency coupling coordinated control performance of a new energy power station in this invention;

[0057] Figure 4 This is a schematic flowchart of another embodiment of the online evaluation method for voltage and frequency coupling coordinated control performance of a new energy power station in this invention;

[0058] Figure 5 This is a schematic flowchart of another embodiment of the online evaluation method for voltage and frequency coupling coordinated control performance of a new energy power station in this invention;

[0059] Figure 6 This is a schematic flowchart of another embodiment of the online evaluation method for voltage and frequency coupling coordinated control performance of a new energy power station in this invention. Detailed Implementation

[0060] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0061] Example 1

[0062] The implementation method in this embodiment can be implemented in a system, on a server, or on a terminal; no specific limitation is made. The following section will introduce the online evaluation method for the voltage and frequency coupling coordinated control performance of new energy power stations from the perspective of system implementation. Please refer to... Figures 1 to 6 The method provided in this application includes the following steps:

[0063] S1. Collect the three-phase voltage signal, frequency signal, unit output power signal and grid dispatch command signal at the grid connection point of the new energy power station, and calculate the voltage and frequency coupling feature matrix based on the collected signals;

[0064] In this embodiment, the voltage-frequency coupling feature matrix is ​​calculated based on the acquired signal, including the following steps:

[0065] S11. Multi-channel synchronous acquisition of three-phase voltage signals, frequency signals, unit output power signals, and grid dispatch command signals at the grid connection point of new energy power plants to generate a coupled observation data matrix;

[0066] In renewable energy power plants, specialized multi-channel synchronous acquisition equipment is used to collect three-phase voltage signals, frequency signals, generator output power signals, and grid dispatch command signals at the grid connection point. The acquisition equipment features high precision and a high sampling frequency, ensuring that the acquired data accurately reflects the true state of the signals. After the signals are acquired, a coupled observation data matrix is ​​generated.

[0067] S12. Perform frequency domain decomposition on the coupled observation data matrix to obtain the fundamental frequency component matrix and the high-frequency disturbance component matrix, and calculate the initial element values ​​of the coupling degree characteristic matrix based on the frequency domain energy integral.

[0068] The frequency domain decomposition of the coupled observation data matrix includes the following steps:

[0069] 1. The coupled observation data matrix is ​​decomposed into full-band components using Discrete Fourier Transform to generate initial spectral components;

[0070] 2. Based on the energy proportion threshold, the initial spectral components are screened for the fundamental frequency band, and the components whose energy proportion in the main frequency band exceeds the preset value are extracted to form the fundamental frequency component matrix;

[0071] 3. Perform moving average filtering on the remaining frequency band components to generate a high-frequency disturbance component matrix, and extract the envelope of the high-frequency disturbance components through Hilbert transform.

[0072] Specifically, the coupled observation data matrix is ​​decomposed across the entire frequency band using the Discrete Fourier Transform technique to obtain initial spectral components. By setting an energy percentage threshold (which can be determined based on actual energy conditions), components with a main frequency band energy percentage exceeding a preset value are selected; these components constitute the fundamental frequency component matrix. The remaining frequency band components are then subjected to moving average filtering to obtain the high-frequency disturbance component matrix, and the envelope of the high-frequency disturbance components is extracted using a Hilbert transform. The fundamental frequency component matrix is ​​calculated based on the frequency domain energy integral to obtain the initial element values ​​of the coupling degree characteristic matrix.

[0073] S13. Perform dynamic phase matching on the initial element values ​​of the coupling degree feature matrix, and correct it by introducing a coupling compensation coefficient to generate the corrected coupling degree feature matrix;

[0074] S14. Based on the multi-source data fusion algorithm, the modified coupling degree feature matrix and the power grid state matrix are correlated and analyzed to generate the coupling response matrix. The coupling response matrix is ​​then jointly calculated in the time and frequency domains to generate the voltage and frequency coupling feature matrices under different operating modes.

[0075] Specifically, coupling compensation coefficient The following steps were used to calculate the result:

[0076] First, the coupling degree feature matrix before correction is... Coupled with a baseline matrix constructed from historical stable operating data Perform element-wise complex correlation operations to extract the instantaneous phase deviation matrix. :

[0077]

[0078] in, and For matrix element indexing, To obtain the complex phase angle, It is the conjugate of the corresponding elements of the base matrix.

[0079] Secondly, a phase tracker based on Kalman filtering is used to... Smoothing and prediction are performed to generate a dynamic phase compensation matrix. Coupling compensation coefficient matrix It is generated by the following formula:

[0080]

[0081] in: The imaginary unit is used to perform phase rotation correction on the initial element values ​​in the complex field to compensate for phase lag or lead caused by coefficient perturbation.

[0082] Corrected Coupling Feature Matrix It can be obtained through the following formula:

[0083]

[0084] in: This represents the Hadamard product of a matrix (element-by-element multiplication).

[0085] In actual operation, the signal may have phase deviation. It is necessary to perform dynamic phase matching on the initial element values ​​of the coupling degree feature matrix. After introducing the coupling compensation coefficient for correction, the multi-source data fusion algorithm is used to correlate and analyze it with the power grid state matrix to obtain the coupling response matrix. The voltage and frequency coupling feature matrices under different operating modes are generated by joint calculation in the time and frequency domains.

[0086] Large wind farms are connected to the grid connection point via cables. The multi-channel synchronous acquisition equipment installed here can accurately capture real-time changes in three-phase voltage, frequency, unit output power, and grid dispatch command signals, and integrate them to generate a coupled observation data matrix.

[0087] Frequency domain decomposition of the coupled observation data matrix: Initial spectral components are obtained by using discrete Fourier transform to decompose the entire frequency band; an energy percentage threshold (e.g., 70%) is set, and the components with energy percentages exceeding the threshold in the main frequency band are extracted to form the fundamental frequency component matrix; the remaining frequency band components are filtered by moving average to remove noise and abnormal fluctuations, generating a high-frequency disturbance component matrix, and then the envelope is extracted by Hilbert transform.

[0088] The fundamental frequency component matrix is ​​calculated based on the frequency domain energy integral, yielding the initial element values ​​of the coupling degree characteristic matrix. Since interference in the power system may cause signal phase deviation, a coupling compensation coefficient derived from historical wind farm data and theoretical analysis is introduced to correct the initial element values, resulting in the corrected coupling degree characteristic matrix.

[0089] The modified coupling feature matrix is ​​correlated with the power grid state matrix stored in the monitoring center and analyzed using a multi-source data fusion algorithm to obtain the coupling response matrix. The voltage-frequency coupling feature matrix under different operating modes is generated through joint time-frequency domain calculation, providing key data support for subsequent analysis and evaluation of the voltage and frequency coupling coordinated control performance of wind farms.

[0090] S2. Perform co-quantization analysis on the voltage and frequency coupling characteristic matrix to generate a co-parameter matrix;

[0091] In this embodiment, generating the cooperative parameter matrix includes the following steps:

[0092] S21. Dynamically divide the voltage and frequency coupling feature matrix into intervals, generate quantization interval parameters, and adaptively set thresholds for the coupling features in the voltage and frequency coupling feature matrix based on the quantization interval parameters to generate a dynamic quantization matrix.

[0093] Determining the quantization interval parameters is an adaptive process based on data-driven density estimation:

[0094] For each coupling feature sequence in the voltage and frequency coupling feature matrix The probability density function of its eigenvalues ​​is calculated using kernel density estimation. The kernel function used is a Gaussian kernel with a bandwidth of [missing information]. Adaptive selection based on the standard deviation of historical data; for the estimated probability density function Peak and valley detection is performed to identify local maxima of the density function. Identify the central pattern of the data distribution and identify local minima. As candidate boundaries for interval division; based on the detected boundary points, combined with the preset number of intervals. (The number of intervals is set to 3-5 depending on the number of operating modes.) Interval merging and optimization are performed. The final quantization interval parameters are defined as a set of suitable boundary values. ,in and The minimum and maximum values ​​of the data are such that each interval contains a significant data pattern (peak).

[0095] After obtaining the voltage-frequency coupling feature matrix, it is dynamically divided into intervals. Based on the data distribution in the matrix and practical application requirements, quantization interval parameters are determined. Using these parameters, adaptive threshold settings are applied to the coupling features in the voltage-frequency coupling feature matrix to generate a dynamic quantization matrix.

[0096] Obtaining the quantization interval parameters Then, the adaptive threshold is set according to the following rules:

[0097] For each quantization interval Set a corresponding dynamic threshold This threshold is not a fixed value, but rather a function of the statistical characteristics of the data within the interval; for real-time data streams, within the sliding time window, the threshold... The calculation rules are as follows: ,in: Within the current window, the range falls within... The mean of all data. The standard deviation of the corresponding data. This is a sensitivity coefficient, preset according to the stability and response speed requirements of the power grid; for example, when high stability is required... A smaller value is chosen to tighten the threshold. Finally, a dynamic quantization matrix is ​​generated. Its elements Indicates the first The feature in the first Interval index of the time step And the dynamic threshold of this interval This will be used for subsequent judgments.

[0098] S22. Perform multimodal classification processing on the coupling features in the voltage and frequency coupling feature matrix according to the dynamic quantization matrix to generate a classification feature matrix;

[0099] Multimodal classification processing based on dynamic quantization matrix and the original coupling feature matrix The following criteria shall be adopted:

[0100] Based on the dynamic quantization matrix Each feature sampling point is initially divided into its corresponding quantization interval mode. Building upon the initial classification, further analysis is conducted on the trajectory of each feature parameter within a short time window. Morphological features of the trajectory are extracted, such as slope sign (positive / negative / zero), curvature (convex / concave), and whether it crosses adjacent intervals. This leads to the definition of a set of trajectory patterns. The final classification result is a composite modality. It is determined by both the primary interval pattern and the secondary trajectory pattern: For example, one feature point might be classified as a (interval 2, rapidly increasing) mode, while another might be classified as a (interval 2, stationary) mode. Finally, a classification feature matrix is ​​generated. Each element is no longer the original scalar feature value, but represents the composite mode to which the sampling point belongs. The encoding.

[0101] S23. Perform sliding window statistics on the classification feature matrix and set a dynamic coupling interval to generate a statistical feature matrix;

[0102] Based on the dynamic quantization matrix, multimodal classification is performed on the coupling features in the voltage-frequency coupling feature matrix. The coupling features are divided into different categories according to different feature attributes and threshold ranges, generating a classification feature matrix. The classification feature matrix is ​​then processed using a sliding window statistical method. During the sliding window process, a dynamic coupling interval is set, and the data features within the window are statistically analyzed to generate a statistical feature matrix.

[0103] S24. The statistical feature matrix is ​​smoothed through a low-pass filter to generate a steady-state feature matrix, and the steady-state feature matrix is ​​normalized to generate a cooperative parameter matrix.

[0104] The statistical feature matrix is ​​smoothed using a low-pass filter to remove high-frequency noise and fluctuations from the signal, yielding the steady-state feature matrix. To facilitate subsequent analysis and calculation, the steady-state feature matrix is ​​normalized, mapping the data to a specific interval and generating a co-parameter matrix.

[0105] S3. Based on the collaborative evaluation network structure, the control performance of the collaborative parameter matrix is ​​evaluated, and collaborative evaluation parameters are generated;

[0106] In this embodiment, generating collaborative evaluation parameters includes the following steps:

[0107] S31. Divide the collaborative parameter matrix into a main control feature matrix and an auxiliary feature matrix;

[0108] S32. Construct a master control network based on the master control feature matrix. The master control network includes an input layer, a hidden layer and an output layer. The input layer contains voltage, frequency, power and command signals. The hidden layer uses a dynamic nonlinear activation function. The output layer generates master control evaluation values ​​and obtains the output data of the master control network.

[0109] S33. Construct an auxiliary network based on the auxiliary feature matrix. The auxiliary network includes an input layer, a hidden layer, and an output layer. The input layer receives the output data of the main control network, the hidden layer uses a piecewise activation function, and the output layer generates auxiliary evaluation values ​​to obtain the output data of the auxiliary network.

[0110] S34. Jointly train the master network and the auxiliary network to generate a collaborative evaluation network structure, and perform error backpropagation and network sensitivity analysis on the collaborative evaluation network structure to generate network evaluation results; wherein, performing error backpropagation and network sensitivity analysis on the collaborative evaluation network structure includes: constructing a composite loss function that includes mean square error and cross-entropy loss, and calculating the error gradient on the network output data;

[0111] S35. Based on the network evaluation results, update the weight coefficients of the master network and auxiliary network in the collaborative evaluation network structure, adjust the network parameters through adaptive learning rate, and generate collaborative evaluation parameters.

[0112] A collaborative evaluation network is constructed based on the collaborative parameter matrix, which is first divided into a master control feature matrix and an auxiliary feature matrix. The master control network (containing an input layer, hidden layer, and output layer) is constructed based on the master control feature matrix: the input layer has four neurons corresponding to voltage, frequency, power, and command signals; the hidden layer uses a dynamic nonlinear activation function, which can adaptively adjust the activation method according to the characteristics of the input data, enhancing nonlinear processing capabilities; the output layer generates the master control evaluation value, obtaining the output data of the master control network. An auxiliary network (also containing an input layer, hidden layer, and output layer) is constructed based on the auxiliary feature matrix: the input layer receives the output data of the master control network; the hidden layer uses a piecewise activation function (activated according to the characteristics of different data intervals); the output layer generates auxiliary evaluation values, obtaining the output data of the auxiliary network.

[0113] Two networks are jointly trained to achieve collaborative optimization: During training, parameters are adjusted through error backpropagation and network sensitivity analysis. A composite loss function containing mean squared error and cross-entropy loss is constructed. The error gradient is calculated to adjust the weight coefficients. Combined with adaptive learning rate, parameters are optimized, and finally, collaborative evaluation parameters are generated. Taking a 50 MW large-scale photovoltaic power plant (connected to the grid via transmission lines) as an example, the relevant processing flow is as follows:

[0114] Starting with the voltage and frequency coupling feature matrix, a dynamic interval division is first performed (by combining historical operating data and the parameter change patterns under different conditions to determine the quantization interval parameters, and then setting an adaptive threshold for the coupling features to generate a dynamic quantization matrix); then, the coupling features are classified into multiple modes according to the dynamic quantization matrix to generate a classification feature matrix; the classification feature matrix is ​​processed using a sliding window statistical method (by setting the window size and time interval, statistically analyzing the voltage, frequency, and other data features in a specific interval within the window to generate a statistical feature matrix); the statistical feature matrix is ​​then subjected to spectral smoothing using a low-pass filter to obtain a steady-state feature matrix, which is then normalized (mapped to the [0,1] interval) to generate a coordinating parameter matrix.

[0115] The aforementioned collaborative parameter matrix is ​​used to construct a collaborative evaluation network. The resulting collaborative evaluation parameters can reflect the voltage and frequency coupling control performance of the photovoltaic power station, providing a basis for decision-making and adjustments.

[0116] S4. Construct a dynamic coupling constraint function based on the collaborative evaluation parameters, and calculate the collaborative stability boundary characterizing the station's operating state;

[0117] In this embodiment, obtaining the cooperative stability boundary characterizing the station's operating state includes the following steps:

[0118] S41. Reconstruct the collaborative evaluation parameters using multi-dimensional tensors, and generate a coupled state matrix through nonlinear mapping;

[0119] Specifically, collaborative evaluation parameters It is a vector that changes over time. ,in Representing the Several collaborative evaluation sub-parameters, such as voltage stability score and frequency response agility score; time series Reconstructed into a three-dimensional tensor This tensor characterizes the cooperative state of the system from the upper dimension, where dimension one ( - Operation mode dimension), which divides historical operation data into clusters using a clustering algorithm. Typical operating modes, such as high power output and low voltage ride-through, each mode corresponds to a slice of the tensor; dimension two ( -Time series dimension) The collaborative evaluation parameters of each mode in a fixed-length time window The sequence within; dimension three ( - Feature parameter dimension), i.e. Its own Sub-parameters. For tensors Perform modulo-1 matrix transformation, expanding it into a two-dimensional matrix. This is the coupling state matrix. The rows of this matrix are indexed by a combination of (operation mode, time point), and the columns correspond to feature parameters. Subsequently, singular value decomposition is performed on this matrix: Take the left singular matrix The former The three principal column vectors (corresponding to the largest singular values) constitute the dimension-reduced coupling state matrix that can characterize the core coupling dynamics of the system. ,in It includes a joint feature vector of operating mode and time series. It is a diagonal matrix, and the singular values ​​characterize the energy intensity of each coupling mode. It contains coupling structure information in the dimension of feature parameters.

[0120] S42. Based on the coupled state matrix, design a dynamic energy integral term, perform multi-interval operations on the coupled variables to generate dynamic constraint terms, and combine the coupled state matrix and dynamic constraint terms to construct the station operation constraint function, thereby generating a dynamic coupled constraint function.

[0121] The dynamic energy integral term is used to quantify the energy of a system deviating from its equilibrium point under specific coupled states. It is defined as a quadratic integral of the state variables:

[0122]

[0123] in: For the current moment, The dynamic integral window is adaptively adjusted according to the period of the main oscillation mode of the system, for example, taking 2-3 times the dominant oscillation period; , for The deviation of the voltage and frequency at any given time from their rated values. This is the coupling state matrix obtained in the previous step, which encodes the coupling relationship between voltage and frequency; It is a positive definite diagonal weight matrix, whose elements are derived from the coupling state matrix. singular values The result after normalization is used to measure the relative importance of different coupling modes. Here... It includes not only the amplitude of voltage and frequency deviations, but more importantly, through... The dynamic dissipation relationship between the two was captured; the larger the value of t, the higher the unstable energy accumulated by the system under this coupling mode.

[0124] Station operation constraint functions Defined as dynamic energy integral term It must be less than a dynamic threshold : Among them, dynamic threshold It is not a fixed value, but the current running point. The function (including output, load, etc.) is obtained by querying a pre-trained deep neural network model, which operates at a certain point. The input is the maximum coupling energy that can be tolerated at that point, and the output is the maximum coupling energy that can be tolerated at that point.

[0125] S43. Calculate the frequency domain characteristics of the dynamic coupling constraint function, generate the frequency domain response result, and compare the frequency domain response result with the amplitude-frequency characteristics of the running state to generate coupling constraint conditions;

[0126] For dynamic coupling constraint functions Linearization is performed near the rated operating point to obtain its frequency domain transfer function. Its frequency domain response It can be obtained through frequency sweeping or direct calculation; here we are focusing on its amplitude-frequency characteristics. Simultaneously, through the processing in steps S1 and S2, the amplitude-frequency characteristics of the actual operating state of the new energy power station can be extracted. (For example, derived from the voltage-frequency coupling characteristic matrix). Coupling constraints are determined by comparison. and In key frequency bands The relationship within (covering the main oscillation modes) is used to determine this. The specific rules are as follows: ,in The stability margin coefficient is a safety factor used to ensure sufficient stability margin. Let be a very small positive number, which serves as the decision threshold. This condition requires that, at all critical frequencies, the amplitude of the constraint function must be higher than the amplitude of the controlled object after scaling with a certain margin. The integral operation ensures that the requirement is met throughout the entire frequency band, thereby generating the coupled constraint conditions for stable system operation.

[0127] S44. Dynamically divide the rated operating range of the new energy power station according to the coupling constraint conditions, set the collaborative stability index, generate the constraint interval, and perform boundary analysis based on the constraint interval to generate the collaborative stability boundary of the power station's operating state.

[0128] The collaborative evaluation parameters are reconstructed using multi-dimensional tensors, and the coupled state matrix is ​​obtained through mathematical transformation. Based on this, a dynamic energy integral term is designed, and dynamic constraint terms are generated by operating the coupled variables in multiple intervals. The two are combined to construct the dynamic coupling constraint function for the operation of the power station. The frequency domain characteristics of the dynamic coupling constraint function are calculated, and the results are compared with the amplitude-frequency characteristics of the new energy power station's operating state to determine the coupling constraint conditions. Based on this, the rated operating range is dynamically divided, and a collaborative stability index is set to determine the constraint interval. The collaborative stability boundary of the power station's operating state is obtained through boundary analysis.

[0129] Taking a 30 MW small hydropower station as an example, the specific processing procedure is as follows:

[0130] 1. Obtaining Collaborative Evaluation Parameters: Assuming a parameter vector is obtained. ;

[0131] 2. Tensor Reconstruction and Coupled State Matrix: Identifying Operating modes: full-load stable mode, isolated network frequency regulation mode, and low-output mode during dry season; constructing tensors (3 modes, 100 time point windows, 3 parameters); through SVD decomposition, the first 2 principal components are taken to obtain the coupling state matrix. This study revealed that in this hydropower station, the coupling between voltage and frequency is mainly characterized by a frequency-dominated inertial coupling mode.

[0132] 3. Dynamic Energy Integral and Constraint Function: Dynamic Energy Integral Term The weight matrix in according to The singular value setting highlights the frequency-dominant mode; dynamic threshold The calculation is performed by a deep learning model that has been trained with historical data. This model can predict the stable energy limit under the current operating conditions based on the current head height, guide vane opening, and grid load.

[0133] 4. Frequency Domain Comparison and Boundary Generation: Calculating Constraint Functions Amplitude-frequency characteristics It was found that it has a high gain in the 0.1Hz to 1Hz frequency band (corresponding to the turbine inertial regulation frequency band); the amplitude-frequency characteristics of the power station in actual operation were obtained. The coupling constraint is set as follows: within the frequency band of 0.1Hz to 1Hz, Must always be greater than (Right now This ensures a 50% stability margin to suppress power oscillations caused by water hammer; ultimately, this condition is used to optimize the hydropower station's... The (power-frequency) operating range is dynamically divided to generate clear cooperative stability boundaries to guide operation. For example, in islanded mode, this boundary strictly limits the rate of power fluctuation.

[0134] S5. Based on the mapping relationship between the voltage and frequency coupling characteristic matrix and the cooperative stability boundary, the cooperative evaluation parameters are dynamically corrected to generate the station control performance optimization sequence.

[0135] In this embodiment, generating the station control performance optimization sequence includes the following steps:

[0136] S51. Construct an evaluation weight matrix and a compensation correction matrix based on the voltage and frequency coupling characteristic matrix, and substitute the evaluation weight matrix and the compensation correction matrix into the collaborative optimization function to generate the performance optimization objective function;

[0137] S52. Set control response constraints and dynamic adjustment limit conditions for the performance optimization objective function, and convert the cooperative stability boundary into state derivative constraints to generate performance optimization constraints;

[0138] S53. Input the performance optimization objective function and performance optimization constraints into the online optimization solver, calculate the correction parameters in the continuous time domain, and generate the initial sequence of correction parameters;

[0139] S54. The initial sequence of correction parameters is dynamically adjusted in step size, and the regulation rate is constrained in combination with the operating conditions of the new energy power station to generate the power station control performance optimization sequence.

[0140] Evaluation weight matrices and compensation correction matrices are constructed based on the voltage-frequency coupling characteristic matrix. These matrices are built upon the analysis of the importance of different elements in the coupling characteristic matrix and practical control requirements. Substituting the evaluation weight matrices and compensation correction matrices into the collaborative optimization function yields the performance optimization objective function.

[0141] An evaluation weight matrix and a compensation correction matrix are constructed based on the voltage and frequency coupling feature matrix. These matrices are then substituted into a collaborative optimization function to generate a performance optimization objective function. The specific implementation method is as follows:

[0142] First, based on the influence of each characteristic parameter in the voltage-frequency coupling characteristic matrix on the voltage-frequency coupling coordinated control performance of new energy power plants, the weight coefficients of each characteristic parameter are determined, and an evaluation weight matrix is ​​constructed. The evaluation weight matrix for A dimensional matrix, where The number of feature parameters, The number of sampling points, matrix elements Indicates the first The characteristic parameter at the th ... The weight coefficients of each sampling point, and satisfy the following conditions: .

[0143] Secondly, the deviations in voltage and frequency coupling characteristics of new energy power stations under different operating conditions are analyzed, and a compensation correction matrix is ​​constructed. The compensation correction matrix Also 3D matrix, matrix elements Indicates the first The characteristic parameter at the th ... The compensation correction coefficient for each sampling point is used to correct deviations in characteristic parameters caused by equipment characteristics, environmental factors, etc.

[0144] Then, construct the collaborative optimization function. The collaborative optimization function This is a matrix operation function, and its specific expression is:

[0145]

[0146] in, The first element in the voltage-frequency coupling characteristic matrix The characteristic parameter at the th ... The actual value of each sampling point For the first The characteristic parameter at the th ... The target value for each sampling point.

[0147] Evaluation weight matrix and compensation correction matrix Substituting the above collaborative optimization function In this process, the performance optimization objective function is calculated through matrix multiplication and summation. This function represents the weighted sum of squares of the deviations between the actual and target values ​​of the voltage and frequency coupling characteristics of the new energy power station, and is used to measure the target of control performance optimization.

[0148] To ensure the optimization is reasonable and feasible, control response constraints and dynamic adjustment limits need to be applied to the performance optimization objective function. Simultaneously, the cooperative stability boundary is transformed into a state derivative constraint, which together constitutes the performance optimization constraints. The objective function and constraints are input into an online optimization solver. The solver uses an efficient algorithm (such as a genetic algorithm) to calculate the correction parameters in the continuous time domain, obtaining the initial sequence. Considering actual dynamic changes, the initial sequence is dynamically adjusted in step size: the initial step size is set based on the convergence rate of historical correction parameters, and the step size is dynamically scaled according to the distance between the current parameters and the cooperative stability boundary. The step size is nonlinearly optimized using the golden section method, and the adjustment rate constraint is applied in conjunction with the operating conditions of the new energy power station to generate the control performance optimization sequence.

[0149] Taking a 100 MW centralized photovoltaic power plant (composed of multiple photovoltaic arrays, connected to the high-voltage grid via a step-up substation) as an example, its optimization process is as follows:

[0150] The monitoring center technicians construct an evaluation weight matrix and a compensation correction matrix based on the voltage and frequency coupling characteristic matrix. In the evaluation weight matrix, the proportions of voltage and frequency are adjusted according to illumination and load (e.g., voltage has a higher proportion under strong light and heavy load); the compensation correction matrix refers to equipment characteristics and experience (e.g., the inverter's deviation in a specific temperature range and power range). Substituting the two matrices into the co-optimization function yields the performance optimization objective function. Then, control response constraints, dynamic adjustment limits, and state derivative constraints (from the co-stability boundary) are applied as constraints, and input into the online optimization solver to obtain the initial sequence of correction parameters. During dynamic step size adjustment, the initial step size is set according to the historical convergence rate, and the step size is scaled according to the distance from the stability boundary. After optimization using the golden section method, the adjustment rate constraint is combined with real-time illumination, temperature, load, and other operating conditions to finally generate a control performance optimization sequence adapted to the current operating conditions, thereby improving the power plant's operational stability and power quality.

[0151] In addition to the steps described above, this embodiment also includes the following steps:

[0152] 1. Encode the station control performance optimization sequence into instructions to generate control command signals, and perform power adaptation processing on the control command signals to generate execution drive signals;

[0153] 2. Calculate the target voltage and frequency adjustment based on the power grid dispatch instructions, and monitor the operation status of the grid connection point in real time to generate operation monitoring data;

[0154] 3. Calculate the difference between the operation monitoring data and the target voltage and frequency adjustment amount to generate the adjustment deviation value, and compare the adjustment deviation value with the preset allowable deviation threshold to generate the deviation judgment result;

[0155] 4. Based on the deviation determination results, the control parameters are corrected online to generate updated control parameters. The updated control parameters are then reconstructed into a sequence. The control parameters are converted into adjustment command values ​​through state space mapping to generate a new control command sequence.

[0156] 5. Input the new control command sequence into the station controller for closed-loop control and dynamically adjust the unit output.

[0157] Specifically, after obtaining the optimized control performance sequence for the facility, it is encoded into commands. The optimized sequence is then converted into control command signals, enabling them to be recognized and executed by the facility equipment. Power adaptation processing is performed on the control command signals to ensure that the power of the command signals matches the power requirements of the facility equipment, generating execution drive signals.

[0158] The target voltage and frequency regulation amounts are calculated based on the power grid dispatch instructions. Operational monitoring data is obtained by real-time monitoring of the grid connection points. The difference between the operational monitoring data and the target voltage and frequency regulation amounts is calculated to obtain the regulation deviation value. The regulation deviation value is compared with a preset allowable deviation threshold, and a deviation judgment result is generated based on the comparison result.

[0159] If the deviation determination result indicates that the adjustment deviation exceeds the allowable range, the control parameters are corrected online based on this. The control parameters are adjusted according to the deviation to obtain updated control parameters. The updated control parameters are then reconstructed into a sequence, and through state-space mapping, the control parameters are converted into adjustment command values, generating a new control command sequence. This new control command sequence is input into the power station controller for closed-loop control. The power station controller dynamically adjusts the unit output according to the commands, achieving online optimization of voltage and frequency coupled coordinated control performance, ensuring stable and efficient operation of the new energy power station under various operating conditions.

[0160] Taking a large wind farm located on the edge of a desert as an example, the wind farm has an installed capacity of 80 megawatts, consisting of 100 wind turbine generators, each with a capacity of 800 kilowatts, and is connected to the regional power grid via transmission lines. In actual operation, this wind farm uses the method described in this embodiment to achieve online optimization of voltage and frequency coupling and coordinated control performance.

[0161] After obtaining the optimized control performance sequence of the wind farm, the wind farm operation monitoring system encodes the commands. Assume the optimized sequence includes a set of parameters such as turbine pitch angle adjustment and generator excitation current adjustment. The monitoring system converts these adjustments into control command signals according to specific encoding rules, such as binary encoding, converting the value of each adjustment into a corresponding binary code, so that the command signals can be recognized and executed by various equipment in the wind farm.

[0162] The control command signals are then processed for power adaptation. Each wind turbine in a wind farm has its specific power output range and characteristic curve. Taking a certain wind turbine as an example, its rated power is 800 kilowatts, and its power output varies at different wind speeds. According to the power characteristic formula of the wind turbine... (in, This indicates the output power of the fan. It is air density. It's wind speed. It is the area swept by the wind turbine. It is the wind energy utilization coefficient, which is the tip speed ratio. and propeller pitch angle The function combines real-time wind speed, air density, and other data at the current location of the wind turbine with the adjustment amount required in the control command to calculate a suitable execution drive signal. This ensures that the power of the command signal matches the actual power demand of the wind turbine, enabling the wind turbine to operate stably and output the required electrical energy.

[0163] The system calculates target voltage and frequency adjustments based on grid dispatch instructions. For example, a grid dispatch instruction might require the wind farm to increase the grid connection voltage by 5% and stabilize the frequency within the range of 50Hz ± 0.1Hz over the next 15 minutes. The wind farm monitoring system, based on this instruction and the actual voltage and frequency data at the current grid connection point, calculates the specific target voltage and frequency adjustments. Simultaneously, various monitoring devices installed at the grid connection point monitor its operational status in real time, acquiring operational monitoring data such as voltage, frequency, and current.

[0164] The difference between the operational monitoring data and the target voltage and frequency adjustment values ​​is calculated to obtain the adjustment deviation value. Assuming the current actual voltage at the grid connection point is 380V, and the target voltage is 380V × (1 + 5%) = 399V, then the voltage adjustment deviation value is 399V - 380V = 19V. If the current actual frequency is 49.9Hz, and the target frequency range is [49.9Hz, 50.1Hz], then the frequency adjustment deviation value is 0Hz (within the allowable range). The adjustment deviation value is compared with the preset allowable deviation thresholds. Assuming the voltage allowable deviation threshold is ±10V and the frequency allowable deviation threshold is ±0.2Hz, since the voltage adjustment deviation value of 19V is greater than the allowable deviation threshold of 10V, and the frequency adjustment deviation value of 0Hz is within the allowable range, the generated deviation judgment result is that the voltage adjustment is not up to standard, and the frequency adjustment is up to standard.

[0165] The control parameters are corrected online based on the deviation determination results. Because the voltage regulation is not up to standard, the monitoring system adjusts the control parameters related to voltage regulation according to preset correction rules, such as increasing the generator excitation current setpoint, to obtain updated control parameters. The updated control parameters are then reconstructed into a sequence, rearranged and combined, and converted into regulation command values ​​through state-space mapping, generating a new control command sequence.

[0166] Finally, the new control command sequence is input into the wind farm controller for closed-loop control. Upon receiving the commands, the wind farm controller dynamically adjusts the turbine pitch angle and generator excitation current. For example, it increases the pitch angle to increase the wind energy captured by the turbine, while simultaneously adjusting the generator excitation current to control the output voltage and frequency. Through continuous monitoring, calculation, correction, and adjustment, the online optimization of voltage and frequency coupled control performance is achieved, ensuring stable and efficient operation of the wind farm under various operating conditions and meeting the power quality requirements of the power grid.

[0167] It is understood that those skilled in the art can combine various implementation methods in the above embodiments under the guidance of the above examples to obtain technical solutions with multiple implementation methods.

[0168] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements 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 online evaluation of the voltage and frequency coupling coordinated control performance of a new energy power station, characterized in that, include: The system collects three-phase voltage signals, frequency signals, generator output power signals, and grid dispatch command signals at the grid connection point of new energy power plants, and calculates the voltage and frequency coupling feature matrix based on the collected signals. The voltage and frequency coupling characteristic matrix is ​​subjected to co-quantization analysis to generate a co-parameter matrix, including: The voltage and frequency coupling feature matrix is ​​dynamically divided into intervals to generate quantization interval parameters. Based on the quantization interval parameters, the coupling features in the voltage and frequency coupling feature matrix are adaptively thresholded to generate a dynamic quantization matrix. The coupling features in the voltage and frequency coupling feature matrix are subjected to multimodal classification processing according to the dynamic quantization matrix to generate a classification feature matrix; A sliding window statistical analysis is performed on the classification feature matrix, and a dynamic coupling interval is set to generate a statistical feature matrix; The statistical feature matrix is ​​smoothed by passing a low-pass filter to generate a steady-state feature matrix, and the steady-state feature matrix is ​​normalized to generate a cooperative parameter matrix. Based on the aforementioned collaborative parameter matrix, a collaborative evaluation network structure is constructed, and the voltage and frequency coupling state of the new energy power station is evaluated in real time to generate collaborative evaluation parameters. Based on the collaborative evaluation parameters, a dynamic coupling constraint function is constructed, and the collaborative stability boundary characterizing the station's operating state is calculated. Based on the mapping relationship between the voltage and frequency coupling feature matrix and the cooperative stability boundary, the cooperative evaluation parameters are dynamically corrected to generate a station control performance optimization sequence.

2. The online evaluation method for voltage and frequency coupling coordinated control performance of new energy power stations according to claim 1, characterized in that, The process involves collecting three-phase voltage signals, frequency signals, generator output power signals, and grid dispatch command signals from the grid connection point of the new energy power station, and calculating the voltage-frequency coupling characteristic matrix based on the collected signals, including: Multi-channel synchronous acquisition of three-phase voltage signals, frequency signals, unit output power signals, and grid dispatch command signals at the grid connection point of new energy power plants is performed to generate a coupled observation data matrix. The coupled observation data matrix is ​​decomposed in the frequency domain to obtain the fundamental frequency component matrix and the high-frequency disturbance component matrix. The fundamental frequency component matrix is ​​then calculated based on the frequency domain energy integral to generate the initial element values ​​of the coupling degree feature matrix. Dynamic phase matching is performed on the initial element values ​​of the coupling degree feature matrix, and the coupling degree feature matrix is ​​corrected by introducing a coupling compensation coefficient to generate a corrected coupling degree feature matrix. Based on the multi-source data fusion algorithm, the modified coupling degree feature matrix and the power grid state matrix are correlated and analyzed to generate a coupling response matrix. Then, the coupling response matrix is ​​jointly calculated in the time and frequency domains to generate voltage and frequency coupling feature matrices under different operating modes.

3. The online evaluation method for voltage and frequency coupling coordinated control performance of new energy power stations according to claim 2, characterized in that, The frequency domain decomposition of the coupled observation data matrix to obtain the fundamental frequency component matrix and the high-frequency disturbance component matrix includes: The coupled observation data matrix is ​​decomposed into full-band components using discrete Fourier transform to generate initial spectral components; Based on the energy percentage threshold, the initial spectral components are filtered for fundamental frequency bands, and the components whose energy percentage in the main frequency band exceeds a preset value are extracted to form a fundamental frequency component matrix. The remaining frequency band components are subjected to moving average filtering to generate a high-frequency disturbance component matrix, and the envelope of the high-frequency disturbance components is extracted by Hilbert transform.

4. The online evaluation method for voltage and frequency coupling coordinated control performance of new energy power stations according to claim 1, characterized in that, Based on the collaborative parameter matrix, a collaborative evaluation network structure is constructed, and the voltage and frequency coupling state of the new energy power station is evaluated in real time to generate collaborative evaluation parameters, including: The collaborative parameter matrix is ​​divided into a main control feature matrix and an auxiliary feature matrix; A master control network is constructed based on the master control feature matrix. The master control network includes an input layer, a hidden layer, and an output layer. The input layer contains voltage, frequency, power, and command signals. The hidden layer uses a dynamic nonlinear activation function. The output layer generates master control evaluation values ​​to obtain the output data of the master control network. An auxiliary network is constructed based on the auxiliary feature matrix. The auxiliary network includes an input layer, a hidden layer, and an output layer. The input layer receives the output data of the main control network, the hidden layer uses a piecewise activation function, and the output layer generates auxiliary evaluation values ​​to obtain the output data of the auxiliary network. The main control network and the auxiliary network are jointly trained to generate a collaborative evaluation network structure. Error backpropagation and network sensitivity analysis are then performed on the collaborative evaluation network structure to generate network evaluation results. Based on the network evaluation results, the weight coefficients of the master network and auxiliary network in the collaborative evaluation network structure are updated, and the network parameters are adjusted by adaptive learning rate to generate collaborative evaluation parameters.

5. The online evaluation method for voltage and frequency coupling coordinated control performance of new energy power stations according to claim 4, characterized in that, The process of backpropagating errors and performing network sensitivity analysis on the collaborative evaluation network structure includes: constructing a composite loss function that includes mean squared error and cross-entropy loss, and calculating the error gradient of the network output data.

6. The online evaluation method for voltage and frequency coupling coordinated control performance of new energy power stations according to claim 1, characterized in that, The step of constructing a dynamic coupling constraint function based on the collaborative evaluation parameters and calculating the collaborative stability boundary characterizing the station's operating state includes: The collaborative evaluation parameters are reconstructed using multi-dimensional tensors, and a coupled state matrix is ​​generated through nonlinear mapping. Based on the coupling state matrix, a dynamic energy integral term is designed, multi-interval operations are performed on the coupling variables to generate dynamic constraint terms, and the coupling state matrix and the dynamic constraint terms are combined to construct the station operation constraint function, thereby generating a dynamic coupling constraint function. The frequency domain characteristics of the dynamic coupling constraint function are calculated to generate a frequency domain response result. The frequency domain response result is then compared with the amplitude-frequency characteristics of the running state to generate coupling constraint conditions. The rated operating range of the new energy power station is dynamically divided according to the coupling constraint conditions, and a collaborative stability index is set to generate a constraint interval. Based on the constraint interval, boundary analysis is performed to generate a collaborative stability boundary for the operating state of the power station.

7. The online evaluation method for voltage and frequency coupling coordinated control performance of new energy power stations according to claim 1, characterized in that, The process of dynamically correcting the collaborative evaluation parameters based on the mapping relationship between the voltage and frequency coupling characteristic matrix and the collaborative stability boundary, and generating a station control performance optimization sequence, includes: Based on the voltage and frequency coupling feature matrix, an evaluation weight matrix and a compensation correction matrix are constructed, and the evaluation weight matrix and the compensation correction matrix are substituted into the collaborative optimization function to generate a performance optimization objective function; Set control response constraints and dynamic adjustment limit conditions for the performance optimization objective function, and convert the cooperative stability boundary into state derivative constraints to generate performance optimization constraints. The performance optimization objective function and the performance optimization constraints are input into the online optimization solver to calculate the correction parameters in the continuous time domain and generate the initial sequence of correction parameters. The initial sequence of the correction parameters is dynamically adjusted in step size, and the adjustment rate is constrained in combination with the operating conditions of the new energy power station to generate the power station control performance optimization sequence.

8. The online evaluation method for voltage and frequency coupling coordinated control performance of new energy power stations according to claim 7, characterized in that, The dynamic step size adjustment of the initial sequence of the correction parameters includes: The initial adjustment step size is set based on the convergence rate of historical correction parameters; The step size is dynamically reduced or increased based on the distance between the current correction parameter and the cooperative stability boundary, and the golden section method is used to perform nonlinear optimization of the step size adjustment process.

9. The online evaluation method for voltage and frequency coupling coordinated control performance of new energy power stations according to claim 1, characterized in that, The method further includes: The station control performance optimization sequence is encoded into instructions to generate control instruction signals, and the control instruction signals are then processed for power adaptation to generate execution drive signals. Calculate the target voltage and frequency adjustment based on the power grid dispatch instructions, and monitor the operation status of the grid connection point in real time to generate operation monitoring data; The difference between the operation monitoring data and the target voltage and frequency adjustment amount is calculated to generate an adjustment deviation value, and the adjustment deviation value is compared with a preset allowable deviation threshold to generate a deviation judgment result; Based on the deviation determination result, the control parameters are corrected online to generate updated control parameters. The updated control parameters are then reconstructed into a sequence, and the control parameters are converted into adjustment command values ​​through state space mapping to generate a new control command sequence. The new control command sequence is input into the station controller for closed-loop control, and the unit output is dynamically adjusted.

Images