Power system sub-super-synchronous oscillation parameter identification method and device, terminal and medium

Abstract

The application discloses a power system sub-super-synchronous oscillation parameter identification method and device, a terminal and a medium. The scheme provided by the application first acquires electrical quantity sampling data of a power system through a wideband synchronous phasor measurement unit, constructs a superposition model containing a fundamental component and multiple oscillation modes, then calculates the residual error of a measured signal and a model prediction value, constructs a nonlinear least square objective function, and generates a parameter update vector in combination with a dynamic damping factor. In the iterative optimization process, the damping factor size is dynamically adjusted by comparing the objective function value change to adapt to the damping factor numerical value requirement of the current iteration stage, and finally, the optimized oscillation parameter is output as the final identification result. Through the dynamic damping factor adjustment mechanism, a larger damping factor is used to suppress the matrix ill-conditioning in the early stage of iteration, and the damping factor is reduced to improve the convergence speed when approaching convergence, so that the stability and efficiency of the optimization process are effectively balanced.

Description

Technical Field

[0001] This application relates to the field of power signal processing technology, and in particular to a method, device, terminal and medium for identifying subsynchronous oscillation parameters in power systems. Background Technology

[0002] With the centralized grid connection of large-scale wind and solar power plants, traditional power systems are exhibiting a "dual-high" characteristic: a high proportion of new energy power generation and high levels of power electronics. The randomness of new energy power generation and the low inertia and low damping characteristics of power electronic equipment have exacerbated the problem of subsynchronous oscillations, which not only lead to increased losses and accelerated aging of grid equipment but also potentially cause protection malfunctions and system-wide resonance, posing a serious threat to the safe and stable operation of the power grid. Therefore, the management of subsynchronous oscillations in the power grid has received considerable attention in recent years.

[0003] Novel subsynchronous / supersynchronous oscillations caused by grid connection of new energy power generation such as wind and solar power are influenced by various factors, including grid parameters, converter control parameters, and even wind speed. These oscillations exhibit complex frequency fluctuations and a wide range of time-varying characteristics. Furthermore, the power electronic converters used in new energy power generation have low overload capacity, easily causing the converter control to enter the limiting range. This leads to oscillations diverging due to negative damping instability, ultimately manifesting as nonlinear oscillations. Traditional damped least squares methods with fixed damping factors are difficult to effectively adapt to the nonlinear characteristics of grid-connected oscillation models of new energy power generation. This results in poor accuracy in identifying subsynchronous / supersynchronous oscillation parameters, failing to meet the demand for rapid and accurate monitoring of wide-frequency oscillations in grids with high penetration of new energy power. Summary of the Invention

[0004] This application provides a method, device, terminal, and medium for identifying subsynchronous / supersynchronous oscillation parameters in power systems, which addresses the technical problem of poor identification accuracy in existing subsynchronous / supersynchronous oscillation parameter identification methods.

[0005] To address the aforementioned technical problems, the first aspect of this application provides a method for identifying subsynchronous oscillation parameters in power systems, comprising:

[0006] Electrical quantity sampling data in the power system is obtained through broadband synchronous phasor acquisition equipment;

[0007] Based on the sub-supersynchronous oscillation signal contained in the electrical quantity sampling data, a sub-supersynchronous oscillation signal data model is constructed, and based on the sub-supersynchronous oscillation signal data model, the predicted values ​​of the oscillation parameters corresponding to each oscillation mode are determined.

[0008] Based on the residual between the measured and predicted values ​​of the oscillation parameters, a nonlinear least squares objective function and a parameter update function based on dynamic damping are constructed to optimize the sub-supersynchronous oscillation signal data model. When the output iteration state of the nonlinear least squares objective function meets the preset iteration termination condition, the sub-supersynchronous oscillation parameter identification result of the power system is obtained based on the optimized sub-supersynchronous oscillation signal data model.

[0009] Preferably, the sub-supersynchronous oscillation signal data model is specifically as follows:

[0010]

[0011] In the formula, The fundamental amplitude, The fundamental frequency, The fundamental phase is N, and the number of oscillation modes is N. Let i be the amplitude of the i-th oscillation signal. Let be the decay time constant of the i-th oscillation signal. Let be the frequency of the i-th oscillation signal. Let be the phase of the i-th oscillation signal.

[0012] Preferably, the nonlinear least squares objective function is as follows:

[0013]

[0014] In the formula, Let be the objective function. Let be the vector of oscillation parameters to be solved. For time points The measured signal value, For the sub-supersynchronous oscillation signal data model at time point The predicted values ​​of the oscillation parameters.

[0015] Preferably, the parameter update function is specifically:

[0016]

[0017]

[0018] In the formula, For Jacobian matrices, For dynamic damping factor, It is the identity matrix. Update the vector for the parameters. For the residual vector, This is the updated oscillation parameter vector.

[0019] Preferably, the dynamic damping factor is updated in the following way:

[0020] Compare the oscillation parameter vectors before and after the update, and combine them with the absolute value of the function output obtained by the nonlinear least squares objective function;

[0021] When the absolute value of the first function output corresponding to the updated oscillation parameter vector is greater than the absolute value of the second function output corresponding to the oscillation parameter vector before the update, the dynamic damping factor is updated based on a preset first multiple. When the absolute value of the first function output is not greater than the absolute value of the second function output, the dynamic damping factor is updated based on a preset second multiple, wherein the first multiple is greater than 1 and the second multiple is less than 1 and greater than 0.

[0022] Preferably, the step of calculating the predicted oscillation parameters corresponding to each oscillation mode based on the sub-supersynchronous oscillation signal data model includes:

[0023] Based on the subsynchronous oscillation signal data model and the electrical quantity sampling data, the local time-frequency characteristics of the oscillation signal are captured by a preset deep learning neural network, and the predicted values ​​of the oscillation parameters corresponding to each oscillation mode are estimated based on the local time-frequency characteristics.

[0024] Preferably, after obtaining the sub-supersynchronous oscillation parameter identification results of the power system, the method further includes:

[0025] Based on the subsynchronous oscillation frequency obtained from the subsynchronous oscillation parameter identification results of this iteration, update the data time window parameters for the next iteration.

[0026] Meanwhile, a second aspect of this application provides a device for identifying sub-supersynchronous oscillation parameters in a power system, comprising:

[0027] The data sampling unit is used to acquire electrical quantity sampling data in the power system through broadband synchronous phasor acquisition equipment;

[0028] The oscillation parameter initial value determination unit is used to construct a sub-supersynchronous oscillation signal data model based on the sub-supersynchronous oscillation signal contained in the electrical quantity sampling data, so as to determine the predicted values ​​of oscillation parameters corresponding to each oscillation mode based on the sub-supersynchronous oscillation signal data model.

[0029] The oscillation parameter optimization unit is used to construct a nonlinear least squares objective function and a parameter update function based on the residual between the measured and predicted values ​​of the oscillation parameters. The nonlinear least squares objective function and the parameter update function are used to optimize the sub-supersynchronous oscillation signal data model. When the output iteration state of the nonlinear least squares objective function meets the preset iteration termination condition, the sub-supersynchronous oscillation parameter identification result of the power system is obtained based on the optimized sub-supersynchronous oscillation signal data model.

[0030] A third aspect of this application provides a power system sub-supersynchronous oscillation parameter identification terminal, comprising: a memory and a processor;

[0031] The memory is used to store program code, which corresponds to the power system sub-supersynchronous oscillation parameter identification method provided in the first aspect of this application;

[0032] The processor is used to read and execute the program code to implement the sub-supersynchronous oscillation parameter identification method of the power system.

[0033] The fourth aspect of this application provides a computer-readable storage medium storing program code, which is read and executed by a processor to implement the power system sub-supersynchronous oscillation parameter identification method provided in the first aspect of this application.

[0034] As can be seen from the above technical solutions, this application has the following advantages:

[0035] The proposed solution first collects electrical quantity sampling data of the power system using a broadband synchronous phasor measurement unit, constructs a superposition model including the fundamental component and multiple oscillation modes, then calculates the residual between the measured signal and the model prediction, constructs a nonlinear least squares objective function, and generates a parameter update vector by combining a dynamic damping factor. During iterative optimization, the damping factor is dynamically adjusted by comparing changes in the objective function value to adapt to the damping factor numerical requirements of the current iteration stage. When the output convergence state of the objective function reaches a preset threshold condition, the optimized oscillation parameters are output as the final identification result. This solution, through a dynamic damping factor adjustment mechanism, uses a larger damping factor in the early stages of iteration to suppress the ill-conditioned influence of the matrix, and reduces the damping factor as convergence approaches to improve the convergence speed, effectively balancing the stability and efficiency of the optimization process. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating an embodiment of a method for identifying subsynchronous oscillation parameters in a power system provided in this application.

[0038] Figure 2 The waveform comparison diagram shows the calculation results obtained based on the method of this application combined with the collected 500ms voltage data and the actual values.

[0039] Figure 3 This is a schematic diagram of the architecture of an embodiment of a power system sub-supersynchronous oscillation parameter identification device provided in this application.

[0040] Figure 4 This is a schematic diagram of the architecture of a power system sub-supersynchronous oscillation parameter identification terminal embodiment provided in this application. Detailed Implementation

[0041] In existing technologies, with the increasing proportion of new energy power generation connected to the grid, the power system exhibits characteristics of high proportion of new energy and high power electronics, leading to increasingly prominent subsynchronous oscillation problems. Traditional parameter identification methods rely on fixed damping factors, which are difficult to adapt to the nonlinear oscillation characteristics caused by new energy grid connection, resulting in insufficient parameter identification accuracy and inability to meet the requirements of broadband oscillation monitoring. Specifically, because the subsynchronous oscillation identification model of the power grid often has a large number of parameters to be identified and the model is highly nonlinear, the Jacobian matrix is ​​prone to ill-conditioned behavior, which will produce very large numerical errors during inversion, resulting in inaccurate step size direction. Moreover, because the dynamic process of power grid oscillation is very complex, the coupling relationship between different parameters will change drastically with changes in the operating point or the oscillation mode itself. This will cause the local curvature of the nonlinear least squares objective function to be inconsistent on the iteration path, causing the iteration step size direction to deviate from the optimal solution.

[0042] To address the aforementioned issues, an optimization mechanism capable of dynamically adjusting the parameter update direction is needed. Therefore, this application provides a method, apparatus, terminal, and medium for identifying parameters of subsynchronous / supersynchronous oscillations in power systems. By introducing a dynamic damping factor adjustment strategy, the parameter update direction can be adjusted in real time based on the convergence state of the objective function, alleviating the ill-conditioned problem of the Jacobian matrix. Simultaneously, by combining deep learning neural networks to estimate the initial values ​​of the oscillation signal's time-frequency characteristics, more accurate initial parameters are provided for the optimization algorithm, reducing the risk of getting trapped in local optima during the iteration process. This achieves the goal of solving the technical problem of poor identification accuracy in existing subsynchronous / supersynchronous oscillation parameter identification methods.

[0043] To make the inventive objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0044] First, a detailed description of an embodiment of a power system sub-supersynchronous oscillation parameter identification method provided in this application is as follows:

[0045] Please see Figure 1 This embodiment provides a method for identifying sub-supersynchronous oscillation parameters in a power system, including:

[0046] Step 101: Obtain electrical quantity sampling data from the power system using a broadband synchronous phasor acquisition device;

[0047] Step 102: Based on the sub-supersynchronous oscillation signal contained in the electrical quantity sampling data, construct a sub-supersynchronous oscillation signal data model, and determine the predicted values ​​of the oscillation parameters corresponding to each oscillation mode based on the sub-supersynchronous oscillation signal data model.

[0048] Step 103: Based on the residuals between the measured and predicted values ​​of the oscillation parameters, construct a nonlinear least squares objective function and a parameter update function based on dynamic damping. Optimize the sub-supersynchronous oscillation signal data model using the nonlinear least squares objective function and the parameter update function. When the output iteration state of the nonlinear least squares objective function meets the preset iteration termination condition, obtain the sub-supersynchronous oscillation parameter identification result of the power system based on the optimized sub-supersynchronous oscillation signal data model.

[0049] Among them, wideband synchronous phasor acquisition equipment refers to a device capable of synchronously acquiring voltage or current signals from multiple nodes at a high sampling rate. For example, a PMU device using a GPS synchronous clock to achieve microsecond-level time alignment is used to acquire the original signal containing sub-supersynchronous oscillation components. Sub-supersynchronous oscillation signal data models refer to mathematical models that superimpose the fundamental component with multiple oscillation modes. For example, an exponentially decaying sine function is used to describe each oscillation mode, and its function is to decompose complex oscillation signals into analytical mathematical expressions. Nonlinear least squares objective functions use the sum of squared residuals between the measured signal and the model's predicted values ​​as the optimization objective. For example, the Levenberg-Marquardt algorithm is used to construct the objective function, and its function is to transform parameter identification into a nonlinear optimization problem. Dynamic damping parameter update functions refer to parameter update rules that adaptively adjust the damping factor according to the convergence state during iteration. For example, the damping factor is dynamically scaled by comparing changes in the objective function value, and its function is to balance iteration speed and stability, and suppress error amplification caused by parameter coupling.

[0050] Specifically, firstly, electrical quantity sampling data of the power system, such as voltage or current signals, are acquired using a broadband synchronous phasor measurement unit, and time-series data containing sub-supersynchronous oscillation components are extracted. Then, a superposition model containing the fundamental component and multiple oscillation modes is constructed, with each oscillation mode defined by parameters such as amplitude, frequency, phase, and decay time constant. Based on this model, the time-frequency characteristics of the signal are analyzed, and initial predicted values ​​for each oscillation parameter are output. Next, the residual between the measured signal and the model's predicted values ​​is calculated, a nonlinear least-squares objective function is constructed, and a parameter update vector is generated in conjunction with a dynamic damping factor. During iterative optimization, the damping factor is dynamically adjusted by comparing changes in the objective function value to adapt to the damping factor numerical requirements of the current iteration stage. When the output convergence state of the objective function reaches a preset threshold condition, it is determined to be in a convergence state, and the optimized oscillation parameters are output as the final identification result.

[0051] This scheme employs a dynamic damping factor adjustment mechanism. In the early stages of iteration, a larger damping factor is used to suppress the ill-conditioned effects of the matrix. As convergence approaches, the damping factor is reduced to improve the convergence speed, effectively balancing the stability and efficiency of the optimization process. This technical solution can accurately identify the sub-supersynchronous oscillation parameters induced by new energy grid integration, and is particularly suitable for nonlinear oscillation scenarios when converter control enters a limiting state. By dynamically adjusting the damping factor, the impact of parameter coupling and matrix ill-conditioning on the optimization process is mitigated, improving the identification accuracy of key parameters such as frequency, amplitude, and decay time constant, providing reliable data support for monitoring and suppressing broadband oscillations in the power grid.

[0052] Furthermore, the expression for the sub-supersynchronous oscillation signal data model mentioned in the above basic embodiment is as follows:

[0053]

[0054] In the formula, The fundamental amplitude, The fundamental frequency, The fundamental phase is N, and the number of oscillation modes is N. Let i be the amplitude of the i-th oscillation signal. Let be the decay time constant of the i-th oscillation signal. Let be the frequency of the i-th oscillation signal. Let be the phase of the i-th oscillation signal.

[0055] The oscillation signal can be divided into subsynchronous components. Supersynchronous components And the following relationship usually exists: .

[0056] Specifically, this supersynchronous oscillation signal data model achieves a mathematical representation of complex oscillation signals by decomposing electrical quantity sampling data into a superposition of a fundamental component and multiple sub-supersynchronous oscillation modes. The fundamental component corresponds to the power system's steady-state frequency component, while the superimposed oscillation modes describe the transient oscillation characteristics within the subsynchronous and supersynchronous frequency ranges. Each oscillation mode describes the decay process of the oscillation amplitude through an exponential decay term and expresses the oscillation frequency and phase information through a cosine function. In the parameter identification process, the fundamental component parameters are first extracted based on the electrical quantity sampling data. Then, each sub-supersynchronous oscillation mode is separated using a signal decomposition method. Finally, the amplitude, frequency, phase, and decay time constant corresponding to each oscillation mode are iteratively solved using an optimization algorithm.

[0057] The model in this application, by introducing a decay time constant and superimposing multiple oscillation modes, can more accurately describe the characteristics of multi-frequency, time-varying decaying sub-supersynchronous oscillations in new energy grid-connected scenarios. This provides an accurate initial model foundation for subsequent parameter optimization. The above technical solution effectively solves the parameter identification error problem caused by model simplification in traditional methods. By constructing a multi-mode superimposed model including a decay time constant, the dynamic decay process of sub-supersynchronous oscillations in actual power systems can be more realistically reflected, thereby improving the identification accuracy of key parameters such as frequency, amplitude, and decay time constant, and providing a reliable basis for the design of subsequent damping control strategies.

[0058] Furthermore, the expression for the nonlinear least squares objective function mentioned in the above basic embodiment is as follows:

[0059]

[0060] In the formula, Let be the objective function. The oscillation parameter vector to be solved includes , , N , , and , For time points The measured signal value, For the sub-supersynchronous oscillation signal data model at time point The predicted values ​​of the oscillation parameters.

[0061] The nonlinear least squares objective function optimizes the oscillation parameters by minimizing the sum of squared residuals. This can be solved using gradient descent or the Gauss-Newton method, and its mathematical expression is the sum of the squared differences between the measured signal values ​​and the model predictions. The oscillation parameter vector to be solved includes the fundamental amplitude, fundamental frequency, fundamental phase, and amplitude, decay time constant, frequency, and phase parameters of each oscillation mode. It can be mathematically modeled in vector form to uniformly describe all the oscillation characteristics to be identified. The measured signal value refers to the actual measurement result of the power system voltage or current signal at a specific time point obtained through a broadband synchronous phasor acquisition device. This can be stored and processed in discrete time series form. The predicted oscillation parameter value refers to the theoretical signal output at the corresponding time point calculated based on the sub-supersynchronous oscillation signal data model. This can be numerically calculated by substituting the parameters into the model formula.

[0062] Specifically, the nonlinear least squares objective function constructs the mathematical foundation for parameter optimization by summing the squared differences between the measured signal values ​​and the model predictions. In each iteration, the vector of oscillation parameters to be solved is dynamically adjusted to minimize the objective function value. The measured signal values ​​are compared point-by-point with the model predictions as input data, and the residual vector is used to calculate the Jacobian matrix and generate the parameter update direction. The predicted oscillation parameters are obtained by substituting the current parameter estimates into the sub-supersynchronous oscillation signal data model; the calculation process involves the superposition of the fundamental component and multiple oscillation modes. The construction of this objective function transforms the parameter optimization process into a nonlinear least squares problem, gradually approximating the true parameters through numerical optimization algorithms.

[0063] This scheme directly quantifies model prediction errors through a nonlinear least squares objective function. Combined with a dynamic damping factor adjustment mechanism, it adaptively adjusts the parameter update step size according to the iteration stage, effectively mitigating the ill-conditioned problem of the Jacobian matrix. Compared to the fixed damping method, this objective function more accurately reflects the changes in parameter coupling relationships, avoiding error accumulation caused by linear assumptions. Through this technical solution, the nonlinear characteristics of sub-supersynchronous oscillation signals can be described more accurately, improving parameter identification accuracy. The objective function, through the direct minimization of the residual sum of squares, ensures that the parameter optimization process always revolves around the measured data, avoiding divergence problems caused by model bias. The overall optimization method of the parameter vector effectively handles the mutual influence between multiple oscillation modes, providing a reliable mathematical basis for subsequent dynamic damping adjustment.

[0064] Furthermore, the expression for the parameter update function mentioned in the above basic embodiment is as follows:

[0065]

[0066]

[0067] In the formula, For Jacobian matrices, For dynamic damping factor, It is the identity matrix. Update the vector for the parameters. For the residual vector, This is the updated oscillation parameter vector.

[0068] The Jacobian matrix is ​​a matrix composed of the partial derivatives of the objective function with respect to each parameter. It can be calculated using the finite difference method or automatic differentiation techniques and is used to characterize the sensitivity of the residuals to changes in parameters. More specifically, to avoid complex analytical differentiation, the Jacobian matrix can be calculated using the numerical difference method, and can be expressed as:

[0069]

[0070] in, For the difference step size, It is a unit vector.

[0071] Parameter update vector This refers to the parameter adjustment amount during each iteration, specifically obtained by solving a system of linear equations, used to drive the parameters towards the optimal solution. The residual vector is the sequence of differences between the measured signal and the model's predicted values, specifically calculated point-by-point, used to quantify the model fitting error. The parameter update equations are obtained by solving the damped least squares method. , obtain the parameter update vector Then update the equation through parameters. This yields the updated oscillation parameter vector.

[0072] Correspondingly, the objective function can then be computed using the updated parameter vector. and residual vector The iteration process checks if the termination condition is met. If it is, the iteration stops; otherwise, the damping factor is dynamically adjusted to continue iteration. The termination condition mainly includes: the objective function converges, i.e. The residuals converge, that is... Or it could also include: exceeding the limit on the number of iterations, i.e. ,in, To determine the threshold for convergence, a value close to 0 can be used.

[0073] Specifically, in the iterative optimization process, the Jacobian matrix and residual vector under the current parameters are first calculated, and then a regularization matrix is ​​constructed based on the dynamic damping factor. The parameter update vector is obtained by solving the linear equations and is then superimposed on the current parameter vector to form the updated parameter estimates. The dynamic damping factor is set to a relatively large value in the early stages of iteration to suppress oscillations caused by the ill-conditioned matrix, and is gradually decreased as the iteration converges to improve the sensitivity of parameter adjustment. This process, through adaptive adjustment of the damping factor, effectively alleviates the numerical instability problem caused by the deterioration of the Jacobian matrix condition number, while avoiding the imbalance of the fixed damping factor at different stages of iteration.

[0074] This scheme introduces a dynamic damping mechanism, enabling the parameter update process to autonomously adjust the regularization intensity based on the current iteration state, thereby improving convergence efficiency while ensuring numerical stability. This technical solution effectively overcomes the iterative instability problem inherent in the traditional damped least squares method for identifying parameters of strongly nonlinear oscillations, improving the identification accuracy of key parameters such as sub-supersynchronous oscillation frequency and amplitude, and providing reliable technical support for broadband oscillation monitoring in new energy grid-connected systems.

[0075] Furthermore, this embodiment also proposes a method for updating the dynamic damping factor, specifically including: comparing the oscillation parameter vectors before and after the update, and combining the absolute value of the function output obtained by the nonlinear least squares objective function; when the absolute value of the first function output corresponding to the updated oscillation parameter vector is greater than the absolute value of the second function output corresponding to the oscillation parameter vector before the update, updating the dynamic damping factor based on a preset first multiple; when the absolute value of the first function output is not greater than the absolute value of the second function output, updating the dynamic damping factor based on a preset second multiple, wherein the first multiple is greater than 1, and the second multiple is less than 1 but greater than 0.

[0076] The dynamic damping factor refers to the adaptive coefficient used to adjust the parameter update step size. Specifically, it can be implemented using a preset multiple adjustment strategy, such as the first multiple. It can be 10, a multiple of 10. It could be 0.1, or a multiple of 1. It is 2, the second multiple. The damping factor is set to 0.5. By comparing the changes in the objective function output, the damping factor is dynamically adjusted to balance iterative stability and convergence speed. The parameter update vector can be further refined as: the parameter adjustment direction calculated from the Jacobian matrix and the residual vector. This can be achieved through matrix operations, such as multiplying the transpose of the Jacobian matrix with the residual vector, and then adjusting the parameter update magnitude in conjunction with the dynamic damping factor. The absolute value of the function output refers to the calculation result of the nonlinear least squares objective function under different parameter vectors. It can be calculated using the sum of squared residuals and is used to measure the model fit before and after the parameter update.

[0077] Specifically, during parameter iteration, the absolute value change of the objective function output needs to be calculated after each update. If the objective function value increases after the update, it indicates that the current iteration step size is too large or the direction is off. In this case, the damping factor is increased to reduce the step size and avoid divergence. If the objective function value decreases, the damping factor is decreased to accelerate the convergence speed. For example, when the absolute value of the first function output is greater than the absolute value of the second function output, the damping factor is multiplied by a first factor. (e.g., multiply by 10) to reduce the step size of subsequent iterations; conversely, multiply the damping factor by a second factor. (e.g., 0.1 times), gradually relax the step size limit. By dynamically adjusting the damping factor, oscillations caused by ill-conditioned Jacobian matrix can be suppressed in the early stage of iteration, and the efficiency of parameter optimization can be improved when approaching convergence.

[0078] More specifically, the damping factor adjustment method can be expressed as:

[0079]

[0080] in, and They refer to the first multiple and the second multiple, respectively.

[0081] This scheme introduces a dynamic damping adjustment mechanism based on the change of the objective function. In the early stage of iteration, a larger damping factor is used to suppress the influence of the ill-conditioned matrix. In the near convergence stage, the damping factor is gradually reduced to accelerate the approximation of the optimal solution. This overcomes the defect that a fixed damping factor cannot adapt to the dynamic characteristics of nonlinear systems. Through the above technical solution, this application can adaptively adjust the parameter update step size according to the iteration stage, effectively alleviate the error accumulation problem caused by the ill-conditioned Jacobian matrix and parameter coupling. While ensuring the stability of the algorithm, it improves the convergence speed and accuracy of sub-supersynchronous oscillation parameter identification, and is especially suitable for time-varying, nonlinear broadband oscillation scenarios in new energy grid-connected systems.

[0082] Furthermore, regarding the initial value calculation method for the oscillation parameter prediction mentioned in step 102 of the basic embodiment, this embodiment specifically proposes the following example: based on the sub-supersynchronous oscillation signal data model and electrical quantity sampling data, a method is used to capture the local time-frequency features of the oscillation signal through a preset deep learning neural network, and to estimate the oscillation parameter prediction values ​​corresponding to each oscillation mode based on the local time-frequency features. The deep learning neural network uses multi-layer convolution and pooling operations to adaptively capture the local time-frequency features of the oscillation signal. The order of each layer is: input layer, convolutional layer 1, pooling layer 1, convolutional layer 2, pooling layer 2, convolutional layer 3, global average pooling layer, and fully connected layer.

[0083] Deep learning neural networks refer to machine learning models containing multiple hidden layers, specifically implemented using convolutional neural network structures. They automatically learn the time-frequency distribution patterns of oscillating signals by extracting signal features layer by layer and abstracting them progressively. Local time-frequency features refer to the dynamic changes of oscillating signals in the time and frequency dimensions. This can be achieved by calculating feature responses through sliding convolutional kernels in the time-frequency domain, used to characterize the transient behavior of different oscillation modes. Multi-layer convolution and pooling operations refer to alternating stacks of convolutional and pooling layers, implemented using convolutional kernels of different sizes and max pooling. Convolutional layers extract local time-frequency features, while pooling layers reduce feature dimensionality and enhance translation invariance. Global average pooling layers perform spatial dimensional averaging on feature maps, specifically by calculating the global mean for each feature channel, used to compress feature dimensionality while retaining key information. Fully connected layers map the globally average pooled feature vectors to predicted oscillation parameters, implemented through a combination of linear transformations and nonlinear activation functions.

[0084] Specifically, after the electrical quantity sampling data is input into the deep learning neural network, it first undergoes data standardization processing through the input layer. Convolutional layer 1 uses a narrow-band convolutional kernel that slides along the time axis to capture the short-term fluctuation characteristics of the oscillating signal; pooling layer 1 reduces the time resolution through downsampling while retaining significant features. Convolutional layer 2 uses a wideband convolutional kernel that expands in the frequency dimension to identify oscillation components in different frequency bands; pooling layer 2 further compresses the frequency dimension to suppress noise interference. Convolutional layer 3 fuses time-frequency features through multi-scale convolutional kernels to enhance the representation ability of nonlinear oscillation modes. The global average pooling layer compresses the three-dimensional feature map into a one-dimensional feature vector, and the fully connected layer maps this vector to predicted values ​​of oscillation parameters, including frequency, amplitude, and decay time constant, through a weight matrix.

[0085] This scheme automatically learns the local time-frequency characteristics of oscillating signals through a deep learning network. The convolution kernel parameters are adaptively adjusted during training, effectively capturing frequency fluctuations and amplitude abrupt changes. Existing initial value estimation methods based on fixed models are susceptible to parameter coupling, while this scheme abstracts features step-by-step through multi-layer convolution and pooling operations, separating the coupling components of different oscillation modes and providing more accurate initial values ​​for subsequent parameter optimization. This approach solves the problem of insufficient accuracy of traditional initial value estimation methods in nonlinear time-varying scenarios. The deep learning network automatically extracts time-frequency features strongly correlated with oscillation parameters through end-to-end training, avoiding subjective biases from manually designed features. The multi-layer convolution structure effectively handles high-frequency noise and mode aliasing in the signal, while the global average pooling layer suppresses the risk of overfitting. This method provides high-precision initial parameter values ​​for subsequent nonlinear least squares optimization, significantly improving the convergence speed and accuracy of sub-hypersynchronous oscillation parameter identification.

[0086] In addition, regarding the initial value of the predicted oscillation parameter, besides the method based on deep learning networks mentioned in the above embodiments, basic frequency domain or time domain algorithms, such as FFT or Prony algorithm, can also be used for estimation.

[0087] Furthermore, after obtaining the sub-supersynchronous oscillation parameter identification result of the power system in step 103 mentioned in this embodiment, it may further include:

[0088] Based on the subsynchronous oscillation frequency obtained from the subsynchronous oscillation parameter identification results of this iteration, update the data time window parameters for the next iteration.

[0089] It should be noted that after each iteration, the calculation result of this iteration is output. The data time window for the next iteration can be dynamically adjusted based on the frequency of the subsynchronous oscillation component in the subsynchronous component of the calculation result. For example, if the frequency of the lowest frequency subsynchronous component in the current calculation result is... Then the data window length for the next iteration can be adjusted to .

[0090] This scheme dynamically adjusts the time window length based on the lowest subsynchronous frequency, ensuring that the time window length matches the signal period and provides sufficient "effective information" for parameter estimation of each frequency component. This fundamentally improves the identification accuracy. By dynamically adjusting the time window, the identification method can adapt to different frequency oscillation modes, avoiding the problems of "high-frequency information loss and low-frequency accuracy deficiency" caused by a fixed time window. This enhances the robustness of the method in complex scenarios and avoids unnecessary resource consumption, achieving a dynamic balance between "accuracy and efficiency".

[0091] To further verify the effectiveness of the technical solution of this application, this application also provides verification embodiments including specific parameters, as follows:

[0092] Suppose that the electrical quantity data collected by the broadband synchronous phasor acquisition device contains one subsynchronous component, one supersynchronous component, and grid voltage sampling data with noise approximately 0.1% of the fundamental voltage amplitude, and the sampling period is 100μs. The specific expression is as follows:

[0093]

[0094] Based on this sampling data, the data model established according to the steps of this embodiment is as follows:

[0095]

[0096] The model contains a maximum of two subsynchronous components and two supersynchronous components.

[0097] In this embodiment, the Prony algorithm with a 40ms data window is used to obtain the initial value vector as follows:

[0098] x0 = [71,49,4.1,1.2,41,0.4,3.2,1.5,59,-0.8,0.01,0.01,1401,3.7,0.01,0.01,581,1.3]

[0099] Specifically, the initial value calculation in this embodiment shows that there are two main oscillation modes. Therefore, in order to reduce the computational scale, the oscillation modes in the data model are adjusted to two.

[0100] In this embodiment, the nonlinear least squares objective function is constructed as follows:

[0101]

[0102] Specifically, the data window is 40ms, and the number of data points is 400, which is 2 power frequency cycles.

[0103] In this embodiment, the damped least squares parameter update equation is established as follows:

[0104] Specifically, the parameter update equation is as follows:

[0105]

[0106] Where, is the damping factor. .

[0107] Next, the damped least squares parameter update equation is solved, and the parameter vector is updated using the following formula:

[0108]

[0109] After 6 iterations, the residual convergence condition is met, and the calculation results are as follows:

[0110] x=[81.775, 50.01, 4.964, -1.937, 44.871, 0.503, 4.029, 2.139, 55.381,-0.494]

[0111] The model calculation results and the actual values ​​are shown in Table 1 below:

[0112]

[0113] In this embodiment, apart from a certain calculation error in the decay time constant, the calculation errors of other parameters are very small. In this embodiment, the invented algorithm was used to calculate voltage data for up to 500ms, and the calculation results were compared with the actual values. Figure 2 As shown.

[0114] The above is a detailed description of an embodiment of a power system sub-supersynchronous oscillation parameter identification method provided in this application. The following is a detailed description of an embodiment of a power system sub-supersynchronous oscillation parameter identification device provided in this application.

[0115] Please see Figure 3 This embodiment provides a power system sub-supersynchronous oscillation parameter identification device, comprising:

[0116] The data sampling unit 201 is used to acquire electrical quantity sampling data in the power system through a broadband synchronous phasor acquisition device;

[0117] The oscillation parameter initial value determination unit 202 is used to construct a sub-supersynchronous oscillation signal data model based on the sub-supersynchronous oscillation signal contained in the electrical quantity sampling data, so as to determine the predicted values ​​of the oscillation parameters corresponding to each oscillation mode based on the sub-supersynchronous oscillation signal data model.

[0118] The oscillation parameter optimization unit 203 is used to construct a nonlinear least squares objective function and a parameter update function based on the residual between the measured and predicted values ​​of the oscillation parameters. The nonlinear least squares objective function and the parameter update function are used to optimize the sub-supersynchronous oscillation signal data model. When the output state of the nonlinear least squares objective function meets the preset iteration termination condition, the sub-supersynchronous oscillation parameter identification result of the power system is obtained based on the optimized sub-supersynchronous oscillation signal data model.

[0119] Specifically, the data sampling unit acquires grid bus voltage or grid-connected current signals through a high-precision ADC converter, converts the analog signals into digital signals, and transmits them to the processing unit. The oscillation parameter initial value determination unit performs time-frequency analysis on the sampled data, establishes a signal model including fundamental amplitude, frequency, phase, and parameters of multiple oscillation modes, and preliminarily estimates the parameters of each oscillation component through time-domain fitting. The oscillation parameter optimization unit calculates the residual between the measured data and the model prediction values, constructs a nonlinear least squares optimization problem, and iteratively updates the model parameters by combining a dynamic damping factor adjustment mechanism. When the change in the objective function value is lower than a set threshold, convergence is determined, and the optimized oscillation frequency, amplitude, and decay time constant, among other key parameters, are output.

[0120] This scheme introduces a dynamic damping adjustment mechanism during parameter updates by setting up an independent optimization unit. The damping factor can be adaptively adjusted according to the iteration stage. Larger damping is used in the initial stage to ensure algorithm stability, while damping is reduced near convergence to accelerate the optimization process, effectively overcoming the ill-conditioned problem of the Jacobian matrix. This technical solution can accurately identify broadband oscillation parameters induced by power electronic equipment. Particularly in the scenario of grid connection of new energy power generation, it can effectively handle the nonlinear oscillation characteristics generated when the converter control enters the amplitude-limiting state, providing a reliable parameter identification basis for the safe and stable control of the power grid.

[0121] Furthermore, this application also provides detailed descriptions of power system subsynchronous oscillation parameter identification terminal embodiments and computer-readable storage medium embodiments associated with the aforementioned method embodiments.

[0122] like Figure 4 As shown, the subsynchronous oscillation parameter identification terminal for power systems provided in this embodiment includes: a memory 33 and a processor 31, which can be connected via a communication bus 34;

[0123] Memory 33 is used to store program code, which corresponds to the power system sub-supersynchronous oscillation parameter identification method provided in the above embodiments;

[0124] The processor 31 is used to read and execute program code to implement a method for identifying subsynchronous oscillation parameters in power systems.

[0125] In this context, memory refers to the hardware storage unit used to store program code, which can be implemented using ROM, flash memory, or solid-state drives. Its function is to provide the processor with an executable instruction set, ensuring the complete loading and execution of the parameter identification algorithm. The processor is the computational control unit that executes the program code, which can be implemented using a multi-core CPU or GPU. Its function is to accelerate the iterative optimization process of the nonlinear least squares objective function and parameter update function through parallel computing, solving the slow convergence problem caused by the ill-conditioned Jacobian matrix in traditional methods.

[0126] Specifically, after acquiring voltage or current sampling data through a broadband synchronous phasor acquisition device, the terminal's processor calls program code from memory to construct a sub-supersynchronous oscillation signal data model containing the fundamental component and multi-mode oscillation components. During the model parameter optimization phase, the processor iteratively solves the nonlinear least-squares objective function based on a dynamic damping factor adjustment strategy and the predicted oscillation parameters output by a deep learning neural network. When the convergence of the residual vector reaches a preset threshold, the processor outputs optimized key parameters such as oscillation frequency, amplitude, and decay time constant, completing the sub-supersynchronous oscillation identification.

[0127] In some specific implementations, the memory can be configured as a distributed storage architecture, for example, dividing the program code into a data preprocessing module, a neural network initial value estimation module, and a parameter optimization module, which are stored in different storage areas respectively; the processor can adopt a hybrid computing architecture, for example, the CPU performs data sampling and model building tasks, while the GPU undertakes neural network inference and matrix inverse operation tasks.

[0128] This embodiment provides a computer-readable storage medium containing program code, which is read and executed by a processor to implement the power system sub-supersynchronous oscillation parameter identification method provided in the above embodiment.

[0129] In this context, computer-readable storage media refers to a physical carrier capable of persistently storing program code, such as solid-state drives, USB flash drives, or optical discs. Its function is to provide a portable and reusable code storage foundation for the execution of the parameter identification method. Program code refers to a set of computer languages ​​containing executable instructions, such as Python, C++, or Java. Its function is to translate the steps of the parameter identification method into a processor-executable logical flow. The processor is a hardware unit with computing capabilities, such as a multi-core CPU or GPU. Its function is to read and execute the program code in the storage medium to complete the mathematical operations and logical judgments for oscillation parameter identification.

[0130] Specifically, the computer-readable storage medium stores program code, enabling the processor to sequentially complete the following steps when executing the code: First, it acquires electrical quantity sampling data from the power system, such as voltage or current signals; then, it constructs a sub-supersynchronous oscillation signal data model and estimates the predicted values ​​of the oscillation parameters using a deep learning neural network; further, it constructs a nonlinear least squares objective function based on the residuals between the measured and predicted values, updates the parameters using a dynamic damping factor, until the objective function converges; finally, it outputs the optimized oscillation parameter identification result. During this process, the program code controls the processor's computation direction through a preset algorithm flow, such as dynamically adjusting the damping factor to optimize the iteration step size, thereby ensuring the stability and accuracy of parameter identification. Those skilled in the art will understand that, for convenience and brevity, the specific working processes of the terminals, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0131] In the several embodiments provided in this application, it should be understood that the disclosed terminals, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units 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 an indirect coupling or communication connection between devices or units through some interfaces, and may be electrical, mechanical, or other forms.

[0132] 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 particular order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a 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.

[0133] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0134] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0135] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0136] If the integrated unit is implemented as a software functional unit 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 the present invention, in essence, or the part that contributes to the prior art, or all or part 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 described in the various embodiments of the present invention. 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.

[0137] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for identifying sub-supersynchronous oscillation parameters in a power system, characterized in that, include: Electrical quantity sampling data in the power system is obtained through broadband synchronous phasor acquisition equipment; Based on the sub-supersynchronous oscillation signal contained in the electrical quantity sampling data, a sub-supersynchronous oscillation signal data model is constructed, and based on the sub-supersynchronous oscillation signal data model, the predicted values ​​of the oscillation parameters corresponding to each oscillation mode are determined. Based on the residual between the measured and predicted values ​​of the oscillation parameters, a nonlinear least squares objective function and a parameter update function based on dynamic damping are constructed to optimize the sub-supersynchronous oscillation signal data model. When the output iteration state of the nonlinear least squares objective function meets the preset iteration termination condition, the sub-supersynchronous oscillation parameter identification result of the power system is obtained based on the optimized sub-supersynchronous oscillation signal data model. The specific data model for the sub-supersynchronous oscillation signal is as follows: ； In the formula, The fundamental amplitude, The fundamental frequency, The fundamental phase is N, and the number of oscillation modes is N. Let i be the amplitude of the i-th oscillation signal. Let be the decay time constant of the i-th oscillation signal. Let be the frequency of the i-th oscillation signal. Let be the phase of the i-th oscillation signal; The parameter update function is specifically as follows: ； ； In the formula, For Jacobian matrices, For dynamic damping factor, It is the identity matrix. Update the vector for the parameters. For the residual vector, This is the updated oscillation parameter vector; The specific method for updating the dynamic damping factor is as follows: Compare the oscillation parameter vectors before and after the update, and combine them with the absolute value of the function output obtained by the nonlinear least squares objective function; When the absolute value of the first function output corresponding to the updated oscillation parameter vector is greater than the absolute value of the second function output corresponding to the oscillation parameter vector before the update, the dynamic damping factor is updated based on a preset first multiple. When the absolute value of the first function output is not greater than the absolute value of the second function output, the dynamic damping factor is updated based on a preset second multiple, wherein the first multiple is greater than 1 and the second multiple is less than 1 and greater than 0.

2. The method for identifying sub-supersynchronous oscillation parameters in a power system according to claim 1, characterized in that, The nonlinear least squares objective function is specifically as follows: ； In the formula, Let be the objective function. Let be the vector of oscillation parameters to be solved. For time points The measured signal value, For the sub-supersynchronous oscillation signal data model at time point The predicted values ​​of the oscillation parameters.

3. The method for identifying sub-supersynchronous oscillation parameters in a power system according to claim 1, characterized in that, The calculation of the predicted oscillation parameters for each oscillation mode based on the sub-supersynchronous oscillation signal data model includes: Based on the subsynchronous oscillation signal data model and the electrical quantity sampling data, the local time-frequency characteristics of the oscillation signal are captured by a preset deep learning neural network, and the predicted values ​​of the oscillation parameters corresponding to each oscillation mode are estimated based on the local time-frequency characteristics.

4. The method for identifying sub-supersynchronous oscillation parameters in a power system according to claim 1, characterized in that, After obtaining the sub-supersynchronous oscillation parameter identification results of the power system, the process also includes: Based on the subsynchronous oscillation frequency obtained from the subsynchronous oscillation parameter identification results of this iteration, update the data time window parameters for the next iteration.

5. A device for identifying sub-supersynchronous oscillation parameters in a power system, characterized in that, include: The data sampling unit is used to acquire electrical quantity sampling data in the power system through broadband synchronous phasor acquisition equipment; The oscillation parameter initial value determination unit is used to construct a sub-supersynchronous oscillation signal data model based on the sub-supersynchronous oscillation signal contained in the electrical quantity sampling data, so as to determine the predicted values ​​of oscillation parameters corresponding to each oscillation mode based on the sub-supersynchronous oscillation signal data model. The oscillation parameter optimization unit is used to construct a nonlinear least squares objective function and a parameter update function based on the residual between the measured value and the predicted value of the oscillation parameter. The nonlinear least squares objective function and the parameter update function are used to optimize the sub-supersynchronous oscillation signal data model. When the output iteration state of the nonlinear least squares objective function meets the preset iteration termination condition, the sub-supersynchronous oscillation parameter identification result of the power system is obtained based on the sub-supersynchronous oscillation signal data model after oscillation parameter optimization. The specific data model for the sub-supersynchronous oscillation signal is as follows: ； In the formula, The fundamental amplitude, The fundamental frequency, The fundamental phase is N, and the number of oscillation modes is N. Let i be the amplitude of the i-th oscillation signal. Let be the decay time constant of the i-th oscillation signal. Let be the frequency of the i-th oscillation signal. Let be the phase of the i-th oscillation signal; The parameter update function is specifically as follows: ； ； In the formula, For Jacobian matrices, For dynamic damping factor, It is the identity matrix. Update the vector for the parameters. For the residual vector, This is the updated oscillation parameter vector; The specific method for updating the dynamic damping factor is as follows: Compare the oscillation parameter vectors before and after the update, and combine them with the absolute value of the function output obtained by the nonlinear least squares objective function; When the absolute value of the first function output corresponding to the updated oscillation parameter vector is greater than the absolute value of the second function output corresponding to the oscillation parameter vector before the update, the dynamic damping factor is updated based on a preset first multiple. When the absolute value of the first function output is not greater than the absolute value of the second function output, the dynamic damping factor is updated based on a preset second multiple, wherein the first multiple is greater than 1 and the second multiple is less than 1 and greater than 0.

6. A terminal for identifying sub-supersynchronous oscillation parameters in a power system, characterized in that, include: Memory and processor; The memory is used to store program code, which corresponds to the power system sub-supersynchronous oscillation parameter identification method as described in any one of claims 1 to 4; The processor is used to read and execute the program code to implement the sub-supersynchronous oscillation parameter identification method of the power system.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that is read and executed by a processor to implement the subsynchronous oscillation parameter identification method for power systems as described in any one of claims 1 to 4.

Images