Method and apparatus for identifying parameters of power system node frequency response model

Abstract

This invention provides a method and apparatus for identifying parameters of a power system node frequency response model, relating to the field of data processing technology. The method includes: acquiring operational data of multiple nodes in a power system under stable operating conditions; constructing an independent disturbance estimation sequence for each of the multiple nodes based on the operational data; wherein the independent disturbance estimation sequence is obtained by minimizing the residual correlation between different nodes; constructing an equivalent input correlation function and an output correlation function for identifying the node frequency response model based on the independent disturbance estimation sequence; establishing an optimization objective function for identifying the parameters of the node frequency response model based on the equivalent input correlation function and the output correlation function; and solving for the parameters of the node frequency response model based on the optimization objective function. The method provided by this invention achieves accurate modeling of the dynamic frequency characteristics of the system without relying on external excitation.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for identifying parameters of a power system node frequency response model. Background Technology

[0002] In the field of power system dynamic modeling and parameter identification technology, the node frequency response model is an important equivalent model describing the dynamic relationship between active power disturbances and node frequencies. The accuracy of its parameter identification directly affects the effectiveness of applications such as system inertia assessment and frequency regulation capability analysis. Traditional node frequency response (NFR) model parameter identification methods usually rely on artificially applying step or pulse disturbances and estimating parameters by analyzing the system response. However, it is difficult to apply such external disturbances in actual power grid operation, and the system operating conditions are constantly changing. Therefore, such intervention methods are greatly limited in engineering applications.

[0003] In recent years, some studies have attempted to use random fluctuation data under stable power system operation conditions for non-interventional identification. However, due to the superposition and propagation of disturbances from multiple nodes in the power grid, strong correlations exist between power disturbances at different nodes, making it difficult to effectively separate the contribution of each node's independent disturbance to the frequency response from the measurement data. Furthermore, the prevalent white noise measurement errors in actual measurement data severely interfere with the estimation accuracy of statistical characteristics such as correlation functions, thus affecting the accuracy of parameter identification results. Therefore, a solution is urgently needed to address these issues. Summary of the Invention

[0004] This invention provides a method and apparatus for identifying parameters of a power system node frequency response model, in order to overcome the deficiencies in the prior art.

[0005] This invention provides a method for identifying parameters of a power system node frequency response model, comprising the following steps: Acquire operational data from multiple nodes of the power system under stable operating conditions; Based on the operational data of the multiple nodes, an independent perturbation estimation sequence is constructed for each of the multiple nodes; wherein, the independent perturbation estimation sequence is obtained by minimizing the residual correlation between different nodes; Based on the independent perturbation estimation sequence, an equivalent input correlation function and an output correlation function are constructed for node frequency response model identification; Based on the equivalent input correlation function and the output correlation function, an optimization objective function is established to identify the parameters of the node frequency response model, and the parameters of the node frequency response model are obtained by solving the optimization objective function.

[0006] According to the present invention, a method for identifying parameters of a power system node frequency response model includes acquiring operating data of multiple nodes of the power system under stable operating conditions, comprising: The raw operating data of multiple nodes of the power system under stable operating conditions are collected from the wide-area measurement system; wherein, the raw operating data includes the active power measurement value, voltage measurement value and frequency measurement value of each node; The active power measurement, voltage measurement, and frequency measurement values ​​of each node are preprocessed to obtain the active power fluctuation, voltage fluctuation, and frequency fluctuation of each node, which are used as the operating data.

[0007] According to a method for identifying parameters of a power system node frequency response model provided by the present invention, the preprocessing of the active power measurement, voltage measurement, and frequency measurement of each node to obtain the active power fluctuation, voltage fluctuation, and frequency fluctuation of each node includes: The active power measurement, voltage measurement, and frequency measurement of each node are respectively processed to remove the mean, and the corresponding mean-removed sequence is obtained. The mean-free sequence is normalized by dividing it by the corresponding rated value to obtain the active power fluctuation, voltage fluctuation, and frequency fluctuation of each node.

[0008] According to a method for identifying parameters of a power system node frequency response model provided by the present invention, the step of constructing an independent disturbance estimation sequence for each of the plurality of nodes based on the operating data of the plurality of nodes includes: For each node, a linear decomposition model containing coefficients to be optimized is constructed based on the active power fluctuation, voltage fluctuation, and frequency fluctuation of the node; wherein, the linear decomposition model is used to extract the common component driven by voltage and frequency from the active power fluctuation of the node. Based on the linear decomposition model, the residual sequence of each node is constructed; The coefficients to be optimized are solved by minimizing the second-order correlation of the residual sequences between different nodes. Based on the optimized coefficients to be optimized, an independent perturbation estimation sequence is constructed for each node.

[0009] According to the present invention, a method for identifying parameters of a power system node frequency response model includes constructing a residual sequence for each node based on the linear decomposition model, comprising: For each node, based on the voltage fluctuation and the corresponding voltage response coefficient, and the frequency fluctuation and the corresponding frequency response coefficient, the common component driven by both voltage and frequency is extracted from the active power fluctuation of the node to obtain the residual sequence of each node. The voltage response coefficient is used to characterize the sensitivity of node active power changes to voltage fluctuations, and the frequency response coefficient is used to characterize the sensitivity of node active power changes to frequency fluctuations.

[0010] According to the present invention, a method for identifying parameters of a power system node frequency response model includes optimizing the coefficients to be optimized by minimizing the second-order correlation of residual sequences between different nodes. Calculate the residual cross-correlation function of any two different nodes within a set lag window; The squares of the residual cross-correlation functions between all node pairs within the set lag window are summed to construct a global decorrelation objective function; The optimization coefficients are obtained by numerical optimization methods to minimize the global decorrelation objective function.

[0011] According to the present invention, a method for identifying parameters of a power system node frequency response model includes constructing an equivalent input correlation function and an output correlation function for node frequency response model identification based on the independent disturbance estimation sequence, comprising: Based on the active power fluctuation of the target node and the aggregated amount of the independent perturbation estimation sequences of all other nodes besides the target node, the equivalent input correlation function is constructed. The output correlation function is constructed based on the active power fluctuation and frequency fluctuation of the target node.

[0012] According to a method for identifying parameters of a power system node frequency response model provided by the present invention, the equivalent input correlation function is constructed based on the active power fluctuation of the target node and the aggregated amount of the independent disturbance estimation sequences of all other nodes besides the target node, including: The equivalent input correlation function is obtained by adding the product of the active power autocorrelation function of the target node and the cross-correlation function of the target proportional coefficient and the aggregation amount. The target proportional coefficient is a parameter to be optimized, used to characterize the equivalent contribution of power disturbances of other nodes to the frequency response of the target node.

[0013] According to a method for identifying parameters of a power system node frequency response model provided by the present invention, the step of establishing an optimization objective function for identifying the parameters of the node frequency response model based on the equivalent input correlation function and the output correlation function, and solving for the parameters of the node frequency response model based on the optimization objective function, includes: Based on the preset model order, construct the node frequency response model transfer function containing the parameters to be identified; The transfer function is discretized to generate the corresponding unit impulse response sequence; Based on the unit impulse response sequence and the equivalent input correlation function, construct the output correlation function predicted by the model; A parameter optimization objective function is constructed with the goal of minimizing the error between the output correlation function predicted by the model and the actual output correlation function. Solve the objective function for parameter optimization to obtain the parameters to be identified in the node frequency response model.

[0014] According to the present invention, a method for identifying parameters of a power system node frequency response model includes solving the objective function for parameter optimization to obtain the parameters to be identified in the node frequency response model, comprising: The particle swarm optimization algorithm is used to perform global optimization on the parameters to be identified in order to obtain the optimal parameter vector that minimizes the objective function of the parameter optimization.

[0015] According to the present invention, a method for identifying parameters of a power system node frequency response model, after obtaining the parameters of the node frequency response model based on the optimization objective function, the method further includes: The discrete transfer function of the node frequency response model is generated based on the parameters; Apply a unit step input to the discrete transfer function and calculate the corresponding unit step response sequence; The unit step response sequence is compared with the actual response from the simulation platform or measured data to verify the accuracy of the model identification results.

[0016] The present invention also provides a power system node frequency response model parameter identification device, comprising the following modules: The acquisition module is used to acquire operating data of multiple nodes in the power system under stable operating conditions. A construction module is used to construct an independent perturbation estimation sequence for each of the multiple nodes based on the operational data of the multiple nodes; wherein the independent perturbation estimation sequence is obtained by minimizing the residual correlation between different nodes; A construction module is used to construct equivalent input correlation functions and output correlation functions for node frequency response model identification based on the independent perturbation estimation sequence; The identification module is used to establish an optimization objective function for identifying the parameters of the node frequency response model based on the equivalent input correlation function and the output correlation function, and to solve for the parameters of the node frequency response model based on the optimization objective function.

[0017] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the power system node frequency response model parameter identification method as described above.

[0018] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the power system node frequency response model parameter identification method as described above.

[0019] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the power system node frequency response model parameter identification method as described above.

[0020] This invention provides a method and apparatus for identifying parameters of a power system node frequency response model. The method involves acquiring operational data from multiple nodes of a power system under stable operating conditions; constructing an independent disturbance estimation sequence for each node based on the operational data; wherein the independent disturbance estimation sequence is obtained by minimizing the residual correlation between different nodes; constructing an equivalent input correlation function and an output correlation function for identifying the node frequency response model based on the independent disturbance estimation sequence; establishing an optimization objective function for identifying the parameters of the node frequency response model based on the equivalent input correlation function and the output correlation function; and solving the optimization objective function to obtain the parameters of the node frequency response model. Therefore, this invention utilizes random fluctuation data under stable power system operation conditions to achieve parameter identification without the need for artificial disturbances, demonstrating good engineering applicability. Simultaneously, by constructing independent disturbance estimation sequences and minimizing residual correlations between different nodes, it effectively solves the problem of difficulty in distinguishing disturbance sources caused by the coupling of disturbances between multiple nodes, thus enabling the statistical separation of the independent disturbance contributions of each node from the operational data. Furthermore, by constructing equivalent input and output correlation functions and establishing an optimization objective function for solution, the correlation function method can accurately identify node frequency response model parameters under random disturbance conditions, thereby achieving accurate modeling of the system's frequency dynamic characteristics without relying on external excitations. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the method for identifying parameters of a power system node frequency response model provided by the present invention.

[0023] Figure 2 This is a schematic diagram of the structure of the power system node frequency response model parameter identification device provided by the present invention.

[0024] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] The following is combined Figures 1-3 This invention describes a method and apparatus for identifying parameters of a power system node frequency response model.

[0027] Figure 1 This is a flowchart illustrating the power system node frequency response model parameter identification method provided by the present invention, as shown below. Figure 1 As shown, the method includes the following: Step 100: Obtain the operating data of multiple nodes of the power system under stable operating conditions.

[0028] Specifically, operational data refers to the electrical measurement information of multiple nodes collected in real time by a wide-area measurement system when the power system is in steady-state operation and there is no artificial disturbance. Specifically, it includes the active power measurement value, voltage measurement value, and frequency measurement value of each node. After preprocessing, these raw measurement data yield the corresponding active power fluctuation, voltage fluctuation, and frequency fluctuation, which are used to characterize the dynamic response characteristics of the system under the influence of factors such as random load changes and fluctuations in new energy output.

[0029] Step 200: Based on the operational data of the multiple nodes, construct an independent perturbation estimation sequence for each of the multiple nodes; wherein, the independent perturbation estimation sequence is obtained by minimizing the residual correlation between different nodes.

[0030] Specifically, the independent disturbance estimation sequence refers to the residual signal obtained by removing the common component driven by voltage and frequency from the active power fluctuation of a node through statistical separation techniques. This sequence approximately represents the independent random disturbance source within the node in a statistical sense. Its construction process minimizes the second-order correlation between the residual sequences of different nodes by optimizing a set of undetermined coefficients, thereby achieving effective separation of independent disturbances of each node under the condition of mutual coupling of disturbances in multiple nodes.

[0031] Step 300: Based on the independent perturbation estimation sequence, construct the equivalent input correlation function and output correlation function for node frequency response model identification.

[0032] Specifically, the equivalent input correlation function is a statistical characteristic function constructed based on the active power fluctuation of the target node and the aggregate of the independent disturbance estimation sequences of all other nodes. It is used to comprehensively reflect the joint excitation effect of the target node's own power fluctuation and other node disturbances on the target node's frequency response. The output correlation function is a cross-correlation function constructed based on the active power fluctuation of the target node and the frequency fluctuation of the node. It is used to characterize the statistical correlation between power disturbance and frequency response.

[0033] Step 400: Based on the equivalent input correlation function and the output correlation function, establish an optimization objective function for identifying the parameters of the node frequency response model, and solve for the parameters of the node frequency response model based on the optimization objective function.

[0034] Specifically, the optimization objective function is a mathematical expression established based on the linear relationship between the equivalent input correlation function and the output correlation function. By minimizing the error between the model-predicted output correlation function and the measured output correlation function, the parameters of the node frequency response model to be identified can optimally fit the input and output statistical characteristics of the actual system, thereby achieving accurate estimation of model parameters such as the node frequency response transfer function.

[0035] Taking a node in a real power grid as an example, firstly, 30 minutes of active power, voltage, and frequency time series data are collected from the phasor measurement unit (PMU) of the node, with a sampling interval of 0.1 seconds, totaling 18,000 sampling points. After the collected raw data is de-meaned and normalized, the active power fluctuation, voltage fluctuation, and frequency fluctuation of the node are obtained. For multiple nodes including this node, a linear decomposition model is constructed and the voltage response coefficient and frequency response coefficient of each node are optimized to minimize the sum of squares of the cross-correlation function between the residual sequences of different nodes within the lag window, thereby obtaining the independent disturbance estimation sequence of each node. Based on this, for the target node, an equivalent input correlation function is constructed based on its active power autocorrelation function and the aggregate quantity and cross-correlation function of the independent disturbance estimation sequences of other nodes; at the same time, an output correlation function is also constructed. Subsequently, an optimization objective function based on the equivalent input correlation function and the output correlation function is established. The node frequency response transfer function parameters that minimize the error between the model's predicted output correlation function and the measured output correlation function are solved by the particle swarm optimization algorithm. Finally, an NFR model that can accurately describe the dynamic relationship between the node's power perturbation and frequency response is obtained.

[0036] The above describes the steps of the power system node frequency response model parameter identification method provided by this invention. As can be seen from the above description, the power system node frequency response model parameter identification method provided by this invention involves: acquiring operating data of multiple nodes in a power system under stable operating conditions; constructing an independent disturbance estimation sequence for each of the multiple nodes based on the operating data; wherein the independent disturbance estimation sequence is obtained by minimizing the residual correlation between different nodes; constructing an equivalent input correlation function and an output correlation function for node frequency response model identification based on the independent disturbance estimation sequence; establishing an optimization objective function for identifying the node frequency response model parameters based on the equivalent input correlation function and the output correlation function; and solving for the node frequency response model parameters based on the optimization objective function. Therefore, this invention utilizes random fluctuation data under stable power system operation conditions to achieve parameter identification without the need for artificial disturbances, demonstrating good engineering applicability. Simultaneously, by constructing independent disturbance estimation sequences and minimizing residual correlations between different nodes, it effectively solves the problem of difficulty in distinguishing disturbance sources caused by the coupling of disturbances between multiple nodes, thus enabling the statistical separation of the independent disturbance contributions of each node from the operational data. Furthermore, by constructing equivalent input and output correlation functions and establishing an optimization objective function for solution, the correlation function method can accurately identify node frequency response model parameters under random disturbance conditions, thereby achieving accurate modeling of the system's frequency dynamic characteristics without relying on external excitations.

[0037] Based on the above embodiments, in this embodiment, step 100, acquiring operational data of multiple nodes of the power system under stable operating conditions, includes: Step 110: Collect raw operating data of multiple nodes of the power system under stable operating conditions from the wide-area measurement system; wherein, the raw operating data includes the active power measurement value, voltage measurement value and frequency measurement value of each node.

[0038] Step 120: Preprocess the active power measurement value, voltage measurement value and frequency measurement value of each node to obtain the active power fluctuation value, voltage fluctuation value and frequency fluctuation value of each node, which are used as the operation data.

[0039] Step 120 specifically includes: Step 121: Perform mean-reduction processing on the active power measurement value, voltage measurement value and frequency measurement value of each node to obtain the corresponding mean-reduction sequence.

[0040] Step 122: Divide the mean-removed sequence by the corresponding rated value to normalize it, and obtain the active power fluctuation, voltage fluctuation and frequency fluctuation of each node.

[0041] Specifically, the raw operating data of multiple nodes of the power system under stable operating conditions are first collected from a wide-area measurement system. This wide-area measurement system includes phasor measurement units (PMUs) or data acquisition and monitoring systems (SCADA) deployed at each node, which can record the electrical measurement information of each node in real time. The raw operating data collected specifically includes the active power measurement value, voltage measurement value, and frequency measurement value of each node. These measurement values ​​reflect the real-time operating status of the system under the influence of factors such as random load fluctuations and changes in the output of new energy sources.

[0042] Subsequently, the collected raw operating data is preprocessed to extract the fluctuation components for subsequent modeling and analysis. Specifically, the active power, voltage, and frequency measurements of each node are de-meaned, i.e., the time average value is subtracted from the original time series to eliminate the influence of the steady-state operating point, resulting in a de-meaned series reflecting the dynamic fluctuation components. Each de-meaned series is then divided by its corresponding rated value for per-unit processing to eliminate the dimensional influence caused by differences in voltage levels and power benchmarks between different nodes, making the data uniformly comparable. Finally, the active power fluctuation, voltage fluctuation, and frequency fluctuation of each node are obtained, which serve as the operating data used in subsequent steps to construct independent disturbance estimation sequences and construct correlation functions.

[0043] In one embodiment, under stable power system operation conditions, operational data of multiple nodes are acquired from a power system simulation platform or the PMU or SCADA system of the actual power grid via a wide-area measurement system. This data includes: node active power fluctuation data, node voltage fluctuation data, and node frequency fluctuation data. The active power quantity measurement is ,node The voltage measurement is ,node Frequency measurement After removing the mean and normalizing using the corresponding nominal values, the fluctuation amount is obtained. , and .

[0044] Sampling interval is The number of sampling points is In practical applications, the data acquisition window is generally between 20 and 40 minutes, with a sampling interval of about 0.1 seconds, which can ensure the stability of the frequency response identification results.

[0045] The power system node frequency response model parameter identification method provided in this embodiment can eliminate the influence of steady-state operating points and dimensional differences by collecting raw operating data of multiple nodes from a wide-area measurement system and performing de-meaning and per-unit preprocessing. This provides standardized and comparable input data for subsequent high-precision cross-node decorrelation processing and frequency response model identification.

[0046] Based on the above embodiments, in this embodiment, step 200 constructs an independent perturbation estimation sequence for each of the plurality of nodes based on the operational data of the plurality of nodes, including: Step 210: For each node, construct a linear decomposition model containing coefficients to be optimized based on the active power fluctuation, voltage fluctuation, and frequency fluctuation of the node; wherein, the linear decomposition model is used to extract the common component driven by voltage and frequency from the active power fluctuation of the node.

[0047] Step 220: Based on the linear decomposition model, construct the residual sequence for each node.

[0048] Step 220 specifically includes: Step 221: For each node, based on the voltage fluctuation and the corresponding voltage response coefficient, and the frequency fluctuation and the corresponding frequency response coefficient, the common component driven by voltage and frequency is extracted from the active power fluctuation of the node to obtain the residual sequence of each node. The voltage response coefficient is used to characterize the sensitivity of node active power changes to voltage fluctuations, and the frequency response coefficient is used to characterize the sensitivity of node active power changes to frequency fluctuations.

[0049] Step 230: Optimize the coefficients to be optimized by minimizing the second-order correlation of the residual sequences between different nodes.

[0050] Step 230 specifically includes: Step 231: Calculate the residual cross-correlation function of any two different nodes within a set lag window.

[0051] Step 232: Accumulate the squares of the residual cross-correlation functions between all node pairs within the set lag window to construct a global decorrelation objective function.

[0052] Step 233: Solve for the coefficients to be optimized by numerical optimization methods to minimize the global decorrelation objective function.

[0053] Step 240: Based on the optimized coefficients to be optimized, construct an independent perturbation estimation sequence for each node.

[0054] Specifically, for each node, a linear decomposition model is first constructed based on the active power fluctuation, voltage fluctuation, and frequency fluctuation of that node. This model expresses the active power fluctuation of the node as the sum of the product of voltage fluctuation and voltage response coefficient, the product of frequency fluctuation and frequency response coefficient, and a residual term. The voltage response coefficient and frequency response coefficient are the coefficients to be optimized, which respectively characterize the sensitivity of the node's active power change to voltage fluctuation and frequency fluctuation, reflecting the voltage and frequency characteristics of the load. The role of this linear decomposition model is to extract the common component driven by voltage and frequency from the active power fluctuation, so that the remaining residual term is closer to the independent random disturbance source inside the node in a physical sense.

[0055] Step 220 constructs the residual sequence of each node based on the linear decomposition model, which is specifically implemented through step 221: For each node, the product of voltage fluctuation and corresponding voltage response coefficient is subtracted from its active power fluctuation, and then the product of frequency fluctuation and corresponding frequency response coefficient is subtracted to obtain the residual sequence of the node. The residual sequence contains the independent random disturbance components inside the node as well as modeling errors and measurement noise.

[0056] Since the independent perturbations within different nodes should be physically uncorrelated, step 230 optimizes the solution of the coefficients to be optimized by minimizing the second-order correlation of the residual sequences between different nodes. Specifically, this is achieved through steps 231 to 233: calculating the residual cross-correlation function of any two different nodes within a set lag window. This cross-correlation function measures the degree of linear correlation between the two residual sequences under different time delays. The squares of the residual cross-correlation functions between all pairs of nodes within the set lag window are summed to construct a global decorrelation objective function. The smaller the value of this objective function, the weaker the statistical correlation between the residual sequences of each node, i.e., the closer it is to the physically uncorrelated independent perturbation sources. The coefficients to be optimized that minimize the global decorrelation objective function are solved using numerical optimization methods, such as gradient descent or Newton's method.

[0057] Finally, based on the optimized coefficients to be optimized, the residual sequence of each node is recalculated according to the expression in step 221. These optimized residual sequences are the final independent disturbance estimation sequences. They statistically approximate the independent random disturbance sources within each node, providing effective input excitation information for the identification of subsequent node frequency response models.

[0058] In one embodiment, a node frequency response model is first constructed. Engineering operation and control practices often focus more on local-scale frequency characteristics, i.e., node-specific frequency characteristics. The dynamic impact of power fluctuations on the node's frequency measurement is considered. To address this, a node frequency response (NFR) model is introduced to describe the equivalent dynamic relationship between the node's port input-output pairs.

[0059] At the node scale, nodes Active power fluctuations are used as input, and nodes The frequency response characteristics are used as output, and the node frequency response is defined as: (1) in: Represents a node Laplace transform of active power perturbation, Represents a node frequency deviation, Represents a node The node frequency response transfer function.

[0060] Because power disturbances from different nodes propagate and couple with each other in the power grid, directly identifying them using raw power data makes it difficult to distinguish the sources of disturbance. This embodiment achieves statistical separation of node-independent disturbances by constructing cross-node decorrelation processing.

[0061] Considering the voltage and frequency sensitivity of the load, the node Port active power fluctuation can be represented as: (2) In the formula, , Representing nodes respectively The deviation of voltage and frequency relative to their steady-state operating point. For nodes The internal disturbance source, , For nodes voltage and frequency coefficients These include nonlinear terms, modeling errors, and measurement noise.

[0062] Because internal disturbances act on the active power of each port through voltage and frequency feedback. , respectively with There is endogeneity between them, so we directly treat equation (2) as an exogenous regression model and solve it using the least squares method. , This will lead to a deviation between the physical interpretation of the parameters and the residuals. Equation (2) can be viewed as a linear statistical decomposition form to allow for... By stripping away the common component driven by both voltage and frequency, a statistically independent node perturbation estimate is constructed: (3) In the formula, The set of parameters to be optimized. Indicates in the parameter set Below, the nodes obtained after removing common components. The equivalent independent perturbation (residual) estimation sequence.

[0063] because , Physically, this can be understood as independent perturbations within a node; therefore, a reasonable goal is to achieve this through appropriate selection. This makes the residual sequence of each node... The second-order correlation between them is minimal, thus statistically they are close to uncorrelated node sources.

[0064] Therefore, a residual correlation function is introduced: (4) In the formula, Represents a node residual sequence Time lag The value to be taken below.

[0065] Select a symmetrical hysteresis window Define a global decorrelation objective function: (5) In the formula, Focusing on cross-node residual correlation, The smaller the value, the closer the residual sequence is to the uncorrelated source nodes in a statistical sense. In practical implementation, equation (5) is discretized into a finite sum form: (6) The optimal parameters were then solved using numerical optimization methods. .get Then, the node-independent perturbation estimation sequence can be constructed according to equation (3). .

[0066] The power system node frequency response model parameter identification method provided in this embodiment can statistically separate the independent random disturbance sources within each node from the coupled multi-node measurement data by constructing a linear decomposition model and minimizing the second-order correlation of the residual sequences of different nodes. This provides accurate input excitation information for subsequent frequency response model identification and effectively solves the technical problem of difficulty in distinguishing the superimposed disturbances of multiple nodes.

[0067] Based on the above embodiments, in this embodiment, step 300 constructs an equivalent input correlation function and an output correlation function for node frequency response model identification based on the independent perturbation estimation sequence, including: Step 310: Based on the active power fluctuation of the target node and the aggregated amount of the independent perturbation estimation sequences of all other nodes besides the target node, construct the equivalent input correlation function.

[0068] Step 310 specifically includes: Step 311: Add the product of the active power autocorrelation function of the target node and the cross-correlation function of the target proportional coefficient and the aggregation amount to obtain the equivalent input correlation function; The target proportional coefficient is a parameter to be optimized, used to characterize the equivalent contribution of power disturbances of other nodes to the frequency response of the target node.

[0069] Step 320: Construct the output correlation function based on the active power fluctuation and frequency fluctuation of the target node.

[0070] Specifically, firstly, based on the active power fluctuation of the target node and the aggregate of the independent disturbance estimation sequences of all other nodes besides the target node, an equivalent input correlation function is constructed. This aggregate is a comprehensive sequence obtained by summing the independent disturbance estimation sequences of all other nodes, used to characterize the overall impact of all other random disturbance sources in the system on the target node. Step 311 further refines the construction method of the equivalent input correlation function, that is, adding the active power autocorrelation function of the target node to the product of a target proportionality coefficient and the cross-correlation function of the above aggregate, wherein the active power autocorrelation function of the target node describes the node itself. The statistical self-similarity of power fluctuations under different time delays, and the cross-correlation function between the aggregated amount and the active power fluctuation of the target node describe the statistical influence of other node disturbances on the power fluctuation of the target node. The target proportional coefficient is a parameter to be optimized, which is used to quantify the equivalent contribution of other node power disturbances to the frequency response of the target node. Its value will be determined in subsequent steps in conjunction with the node frequency response model parameters. Through this construction method, the equivalent input correlation function integrates the excitation information of both the target node's own power disturbance and the coupled disturbances of other nodes, and can more comprehensively reflect the effective input excitation acting on the frequency response of the target node.

[0071] Step 320 constructs an output correlation function based on the active power fluctuation and frequency fluctuation of the target node. That is, it calculates the cross-correlation function of the two sequences under different time delays. This function directly describes the statistical causal relationship between the node power disturbance and the resulting node frequency response, and is the key output feature required to identify the node frequency response model.

[0072] In one embodiment, the aggregate amount of active power disturbances within other nodes is defined as: (7) For node frequency response identification, an equivalent input correlation function is constructed: (8) in, Represents a node Active power fluctuation sequence The autocorrelation function, Represents a node Active power fluctuation sequence Aggregate amount of active power disturbances within other nodes The cross-correlation function between them; To characterize the impact of power disturbances at other nodes on the nodes The proportional coefficient of the equivalent contribution of the frequency response is an unknown quantity and is optimized together with the system parameters.

[0073] The relevant output function is: (9) in, Represents a node Active power fluctuation sequence With frequency fluctuation sequence The cross-correlation function between them.

[0074] Therefore, based on the newly constructed input correlation function and output correlation function, NFR can be identified: (10) Let the target frequency response function be... Effective memory duration is Based on the input-output relationship of a linear system, the relationship between the cross-correlation function and the system impulse response can be established: (11) in, for The system's impulse response, Input-output cross-correlation function under finite sample conditions The estimated value, Input-output autocorrelation function under finite sample conditions The estimated value.

[0075] against and To estimate the correlation function, this embodiment employs an improved correlation function estimation method with unified normalization and unified summation intervals. = For example, record them separately. For the input data sequence ,remember For output data sequence Only in the lag order interval The correlation function is estimated on the above, and all lag orders are included. Below, the lower boundary index of the sample calculated by the summation of the correlation function is... The number of valid samples is .because Then the lagging interval boundary Should meet Under this constraint, the unified estimate of the sample correlation function is defined as: (12) in, for and The estimated value of the sample cross-correlation function. The above can be obtained. Similarly, we can obtain and .

[0076] against In the identification of stable operating conditions in power systems, the selection of sample size is crucial to reduce statistical errors caused by measurement noise and other factors. Often very large, at this time The upper bound is relatively large. In practical applications, a more suitable value can be selected by considering the decay characteristics of the relevant function. To improve computational efficiency. It should be large enough to include all non-zero lags with practical physical meaning in the negative half-axis of the cross-correlation function within this interval, thus avoiding the truncation of effective information.

[0077] The power system node frequency response model parameter identification method provided in this embodiment constructs an equivalent input correlation function that comprehensively considers the node's own disturbances and the coupling effects of other nodes, and builds an output correlation function that directly reflects the relationship between power disturbances and frequency response. This provides effective features that can accurately characterize the statistical characteristics of system input and output for subsequent node frequency response model identification.

[0078] Based on the above embodiments, in this embodiment, step 400 establishes an optimization objective function for identifying the parameters of the node frequency response model based on the equivalent input correlation function and the output correlation function, and solves for the parameters of the node frequency response model based on the optimization objective function, including: Step 410: Based on the preset model order, construct the node frequency response model transfer function containing the parameters to be identified.

[0079] Step 420: Discretize the transfer function to generate the corresponding unit impulse response sequence.

[0080] Step 430: Based on the unit impulse response sequence and the equivalent input correlation function, construct the output correlation function predicted by the model.

[0081] Step 440: Construct a parameter optimization objective function with the goal of minimizing the error between the output correlation function predicted by the model and the output correlation function.

[0082] Step 450: Solve the objective function for parameter optimization to obtain the parameters to be identified for the node frequency response model.

[0083] Step 450 specifically includes: Step 451: Use the particle swarm optimization algorithm to perform global optimization on the parameters to be identified, so as to obtain the optimal parameter vector that minimizes the objective function of the parameter optimization.

[0084] Specifically, firstly, a nodal frequency response model transfer function containing the parameters to be identified is constructed based on a preset model order. This transfer function is usually in the form of a rational fraction, and the coefficients of the polynomials in the numerator and denominator are the parameters to be identified. The selection of the model order needs to balance the fitting accuracy and model complexity, and can be determined based on prior knowledge of the dynamic characteristics of the system or through the order identification method.

[0085] Step 420 discretizes the transfer function by using numerical methods such as bilinear transform to convert the continuous-time transfer function into a discrete-time transfer function. Then, a unit impulse response sequence of finite length is generated by inverse Z-transform. This sequence describes the output response of the system under unit impulse excitation and is a key bridge connecting the transfer function parameters and the prediction of the relevant function domain model.

[0086] Step 430 constructs the output correlation function predicted by the model based on the unit impulse response sequence and the equivalent input correlation function constructed in step 300. Specifically, this is achieved by performing a convolution operation between the unit impulse response sequence and the equivalent input correlation function. This predicted output correlation function characterizes the statistical relationship between the input and output expected by the model under the current parameters to be identified.

[0087] Step 440 aims to minimize the sum of squared errors between the model-predicted output correlation function and the measured output correlation function constructed in step 300, and constructs a parameter optimization objective function. This objective function quantifies the degree of matching between the model's statistical response and the actual system's statistical characteristics.

[0088] Step 450 solves the objective function of the parameter optimization to obtain the parameters to be identified in the node frequency response model. Specifically, in step 451, the particle swarm optimization algorithm is used to perform global optimization of the parameters to be identified. The particle swarm optimization algorithm searches for the optimal parameter vector in the parameter space by simulating the foraging behavior of a flock of birds. This algorithm does not depend on the selection of initial values ​​and has good global convergence ability, and is suitable for non-convex optimization problems that may exist in this embodiment.

[0089] In one embodiment, for the system parameters to be identified Let the continuous-time transfer function of the NFR model to be identified be written as: (13) In the formula, These are the system parameters to be identified. This indicates the order of the numerator and denominator. This embodiment does not preset a specific order; instead, it uses a higher-order model for fitting.

[0090] Discretization is performed using a bilinear transform to obtain the discrete-time transfer function: (14) in, This indicates that the Tustin (bilinear) transform is used and the sampling period is... Continuous-discrete transformation operators.

[0091] Depend on Generate length The unit impulse response is (15) Define model predictions (16) The cost function is defined using the residual L2 norm. (17) Since this embodiment does not preset a specific order, but uses a higher-order model for fitting to obtain a model that can accurately characterize the frequency dynamics, a particle swarm optimization algorithm is used for global optimization to obtain the optimal parameter vector. .

[0092] The power system node frequency response model parameter identification method provided in this embodiment transforms the input-output relationship of the relevant function domain into a parameter optimization problem and uses the particle swarm optimization algorithm for global optimization. This method can accurately solve the node frequency response model parameters and achieve high-precision statistical modeling of the system's frequency dynamic characteristics.

[0093] Based on the above embodiments, in this embodiment, after obtaining the parameters of the node frequency response model based on the optimization objective function in step 400, the method further includes: The discrete transfer function of the node frequency response model is generated based on the parameters; Apply a unit step input to the discrete transfer function and calculate the corresponding unit step response sequence; The unit step response sequence is compared with the actual response from the simulation platform or measured data to verify the accuracy of the model identification results.

[0094] Specifically, firstly, based on the optimal parameter vector obtained from the optimization solution, a discrete transfer function of the node frequency response model is constructed. This discrete transfer function is the final mathematical model determined based on the discretization in step 420, which can accurately describe the dynamic relationship between node power disturbance and frequency response. Subsequently, a unit step input is applied to this discrete transfer function, that is, a step change in active power with a per-unit value of 1 occurs in the simulation system on the basis of steady-state operation. The corresponding unit step response sequence is obtained by recursive calculation or inverse Z-transform. This sequence describes the complete dynamic process of node frequency under step disturbance, including key dynamic indicators such as initial rate of change, maximum deviation, steady-state value, and oscillation characteristics. It can intuitively reflect the physical properties of the system such as inertia level, damping characteristics, and primary frequency regulation capability. Finally, the calculated unit step response sequence is compared with the response curve of the detailed time-domain simulation model built in the simulation platform under the same step disturbance, or compared with the real frequency response data recorded by the PMU device when the power grid actually experiences disturbance. By calculating the goodness-of-fit index between the response curves, such as root mean square error or correlation coefficient, the accuracy of the identified model is quantitatively evaluated.

[0095] In one embodiment, obtaining the optimal parameter vector Then, it can be determined by the discretized transfer function. Calculate the corresponding unit step response sequence: (18) in, The truncation length calculated for the response sequence.

[0096] This response characterizes the dynamic process of node frequencies in the system when a unit step change in load occurs, and can intuitively reflect the system's inertia, damping, and primary frequency regulation characteristics. The accuracy of the identification results can be verified by comparing the response with the actual response from the simulation platform or measured data.

[0097] In practical engineering, this method can be used to identify the frequency support capability of the power grid online, and to assist in system stability assessment and control strategy optimization.

[0098] The power system node frequency response model parameter identification method provided in this embodiment converts the identified model parameters into discrete transfer functions and calculates the unit step response. Then, it compares and verifies the results with simulation or measured data. This method can effectively evaluate the accuracy and reliability of the identified node frequency response model and ensure the credibility of the model in engineering applications.

[0099] The following describes the power system node frequency response model parameter identification device provided by the present invention. The power system node frequency response model parameter identification device described below and the power system node frequency response model parameter identification method described above can be referred to in correspondence.

[0100] Figure 2 This is a schematic diagram of the power system node frequency response model parameter identification device provided by the present invention, as shown below. Figure 2 As shown, the power system node frequency response model parameter identification device provided by the present invention includes: The acquisition module 201 is used to acquire the operating data of multiple nodes of the power system under stable operating conditions; The construction module 202 is used to construct an independent perturbation estimation sequence for each of the multiple nodes based on the operational data of the multiple nodes; wherein the independent perturbation estimation sequence is obtained by minimizing the residual correlation between different nodes; Construction module 203 is used to construct equivalent input correlation function and output correlation function for node frequency response model identification based on the independent perturbation estimation sequence; The identification module 204 is used to establish an optimization objective function for identifying the parameters of the node frequency response model based on the equivalent input correlation function and the output correlation function, and to solve for the parameters of the node frequency response model based on the optimization objective function.

[0101] The present invention provides a power system node frequency response model parameter identification device, which acquires operating data of multiple nodes of a power system under stable operating conditions; constructs an independent disturbance estimation sequence for each of the multiple nodes based on the operating data of the multiple nodes; wherein the independent disturbance estimation sequence is obtained by minimizing the residual correlation between different nodes; constructs an equivalent input correlation function and an output correlation function for node frequency response model identification based on the independent disturbance estimation sequence; establishes an optimization objective function for identifying the node frequency response model parameters based on the equivalent input correlation function and the output correlation function, and solves the optimization objective function to obtain the parameters of the node frequency response model. Therefore, this invention utilizes random fluctuation data under stable power system operation conditions to achieve parameter identification without the need for artificial disturbances, demonstrating good engineering applicability. Simultaneously, by constructing independent disturbance estimation sequences and minimizing residual correlations between different nodes, it effectively solves the problem of difficulty in distinguishing disturbance sources caused by the coupling of disturbances between multiple nodes, thus enabling the statistical separation of the independent disturbance contributions of each node from the operational data. Furthermore, by constructing equivalent input and output correlation functions and establishing an optimization objective function for solution, the correlation function method can accurately identify node frequency response model parameters under random disturbance conditions, thereby achieving accurate modeling of the system's frequency dynamic characteristics without relying on external excitations.

[0102] Based on the above embodiments, in this embodiment, the acquisition module 201 is specifically used for: The raw operating data of multiple nodes of the power system under stable operating conditions are collected from the wide-area measurement system; wherein, the raw operating data includes the active power measurement value, voltage measurement value and frequency measurement value of each node; The active power measurement, voltage measurement, and frequency measurement values ​​of each node are preprocessed to obtain the active power fluctuation, voltage fluctuation, and frequency fluctuation of each node, which are used as the operating data.

[0103] Based on the above embodiments, in this embodiment, the device further includes a preprocessing module, specifically used for: The active power measurement, voltage measurement, and frequency measurement of each node are respectively processed to remove the mean, and the corresponding mean-removed sequence is obtained. The mean-free sequence is normalized by dividing it by the corresponding rated value to obtain the active power fluctuation, voltage fluctuation, and frequency fluctuation of each node.

[0104] Based on the above embodiments, in this embodiment, the construction module 202 is specifically used for: For each node, a linear decomposition model containing coefficients to be optimized is constructed based on the active power fluctuation, voltage fluctuation, and frequency fluctuation of the node; wherein, the linear decomposition model is used to extract the common component driven by voltage and frequency from the active power fluctuation of the node. Based on the linear decomposition model, the residual sequence of each node is constructed; The coefficients to be optimized are solved by minimizing the second-order correlation of the residual sequences between different nodes. Based on the optimized coefficients to be optimized, an independent perturbation estimation sequence is constructed for each node.

[0105] Based on the above embodiments, in this embodiment, the device further includes a stripping module, specifically used for: For each node, based on the voltage fluctuation and the corresponding voltage response coefficient, and the frequency fluctuation and the corresponding frequency response coefficient, the common component driven by both voltage and frequency is extracted from the active power fluctuation of the node to obtain the residual sequence of each node. The voltage response coefficient is used to characterize the sensitivity of node active power changes to voltage fluctuations, and the frequency response coefficient is used to characterize the sensitivity of node active power changes to frequency fluctuations.

[0106] Based on the above embodiments, in this embodiment, the device further includes an optimization module, specifically used for: Calculate the residual cross-correlation function of any two different nodes within a set lag window; The squares of the residual cross-correlation functions between all node pairs within the set lag window are summed to construct a global decorrelation objective function; The optimization coefficients are obtained by numerical optimization methods to minimize the global decorrelation objective function.

[0107] Based on the above embodiments, in this embodiment, the construction module 203 is specifically used for: Based on the active power fluctuation of the target node and the aggregated amount of the independent perturbation estimation sequences of all other nodes besides the target node, the equivalent input correlation function is constructed. The output correlation function is constructed based on the active power fluctuation and frequency fluctuation of the target node.

[0108] Based on the above embodiments, in this embodiment, the construction module 203 is further used for: The equivalent input correlation function is obtained by adding the product of the active power autocorrelation function of the target node and the cross-correlation function of the target proportional coefficient and the aggregation amount. The target proportional coefficient is a parameter to be optimized, used to characterize the equivalent contribution of power disturbances of other nodes to the frequency response of the target node.

[0109] Based on the above embodiments, in this embodiment, the identification module 204 is specifically used for: Based on the preset model order, construct the node frequency response model transfer function containing the parameters to be identified; The transfer function is discretized to generate the corresponding unit impulse response sequence; Based on the unit impulse response sequence and the equivalent input correlation function, construct the output correlation function predicted by the model; A parameter optimization objective function is constructed with the goal of minimizing the error between the output correlation function predicted by the model and the actual output correlation function. Solve the objective function for parameter optimization to obtain the parameters to be identified in the node frequency response model.

[0110] Based on the above embodiments, in this embodiment, the identification module 204 is further configured to: The particle swarm optimization algorithm is used to perform global optimization on the parameters to be identified in order to obtain the optimal parameter vector that minimizes the objective function of the parameter optimization.

[0111] Based on the above embodiments, in this embodiment, the device further includes a verification module, specifically used for: After obtaining the parameters of the node frequency response model based on the optimization objective function, The discrete transfer function of the node frequency response model is generated based on the parameters; Apply a unit step input to the discrete transfer function and calculate the corresponding unit step response sequence; The unit step response sequence is compared with the actual response from the simulation platform or measured data to verify the accuracy of the model identification results.

[0112] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device can be a robot or other electronic device. This electronic device may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340. The processor 310, communication interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions from the memory 330 to execute a power system node frequency response model parameter identification method, including: Acquire operational data from multiple nodes of the power system under stable operating conditions; Based on the operational data of the multiple nodes, an independent perturbation estimation sequence is constructed for each of the multiple nodes; wherein, the independent perturbation estimation sequence is obtained by minimizing the residual correlation between different nodes; Based on the independent perturbation estimation sequence, an equivalent input correlation function and an output correlation function are constructed for node frequency response model identification; Based on the equivalent input correlation function and the output correlation function, an optimization objective function is established to identify the parameters of the node frequency response model, and the parameters of the node frequency response model are obtained by solving the optimization objective function.

[0113] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a 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.

[0114] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program, which can be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer is able to execute the power system node frequency response model parameter identification method provided by the above methods, including: Acquire operational data from multiple nodes of the power system under stable operating conditions; Based on the operational data of the multiple nodes, an independent perturbation estimation sequence is constructed for each of the multiple nodes; wherein, the independent perturbation estimation sequence is obtained by minimizing the residual correlation between different nodes; Based on the independent perturbation estimation sequence, an equivalent input correlation function and an output correlation function are constructed for node frequency response model identification; Based on the equivalent input correlation function and the output correlation function, an optimization objective function is established to identify the parameters of the node frequency response model, and the parameters of the node frequency response model are obtained by solving the optimization objective function.

[0115] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the power system node frequency response model parameter identification method provided by the above methods, including: Acquire operational data from multiple nodes of the power system under stable operating conditions; Based on the operational data of the multiple nodes, an independent perturbation estimation sequence is constructed for each of the multiple nodes; wherein, the independent perturbation estimation sequence is obtained by minimizing the residual correlation between different nodes; Based on the independent perturbation estimation sequence, an equivalent input correlation function and an output correlation function are constructed for node frequency response model identification; Based on the equivalent input correlation function and the output correlation function, an optimization objective function is established to identify the parameters of the node frequency response model, and the parameters of the node frequency response model are obtained by solving the optimization objective function.

[0116] The device embodiments described above are merely illustrative. 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0117] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention 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; and these 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 the present invention.

Claims

1. A method for power system node frequency response model parameter identification, characterized in that, include: Acquire operational data from multiple nodes of the power system under stable operating conditions; Based on the operational data of the multiple nodes, an independent perturbation estimation sequence is constructed for each of the multiple nodes; wherein, the independent perturbation estimation sequence is obtained by minimizing the residual correlation between different nodes; Based on the independent perturbation estimation sequence, an equivalent input correlation function and an output correlation function are constructed for node frequency response model identification; Based on the equivalent input correlation function and the output correlation function, an optimization objective function is established to identify the parameters of the node frequency response model, and the parameters of the node frequency response model are obtained by solving the optimization objective function. The construction of equivalent input correlation functions and output correlation functions for node frequency response model identification based on the independent perturbation estimation sequence includes: Based on the active power fluctuation of the target node and the aggregated amount of the independent perturbation estimation sequences of all other nodes besides the target node, the equivalent input correlation function is constructed. Based on the active power fluctuation and frequency fluctuation of the target node, the output correlation function is constructed. The equivalent input correlation function is constructed based on the active power fluctuation of the target node and the aggregated amount of the independent disturbance estimation sequences of all other nodes besides the target node, including: The equivalent input correlation function is obtained by adding the product of the active power autocorrelation function of the target node and the cross-correlation function of the target proportional coefficient and the aggregation amount. The target proportional coefficient is a parameter to be optimized, used to characterize the equivalent contribution of power disturbances of other nodes to the frequency response of the target node; The step of establishing an optimization objective function for identifying the parameters of the node frequency response model based on the equivalent input correlation function and the output correlation function, and solving for the parameters of the node frequency response model based on the optimization objective function, includes: Based on the preset model order, construct the node frequency response model transfer function containing the parameters to be identified; The transfer function is discretized to generate the corresponding unit impulse response sequence; Based on the unit impulse response sequence and the equivalent input correlation function, construct the output correlation function predicted by the model; A parameter optimization objective function is constructed with the goal of minimizing the error between the output correlation function predicted by the model and the actual output correlation function. Solve the objective function for parameter optimization to obtain the parameters to be identified in the node frequency response model.

2. The method of claim 1, wherein, The acquisition of operational data from multiple nodes of the power system under stable operating conditions includes: The raw operating data of multiple nodes of the power system under stable operating conditions are collected from the wide-area measurement system; wherein, the raw operating data includes the active power measurement value, voltage measurement value and frequency measurement value of each node; The active power measurement, voltage measurement, and frequency measurement values ​​of each node are preprocessed to obtain the active power fluctuation, voltage fluctuation, and frequency fluctuation of each node, which are used as the operating data.

3. The method of claim 2, wherein, The preprocessing of the active power measurement, voltage measurement, and frequency measurement values ​​of each node to obtain the active power fluctuation, voltage fluctuation, and frequency fluctuation of each node includes: The active power measurement, voltage measurement, and frequency measurement of each node are respectively processed to remove the mean, and the corresponding mean-removed sequence is obtained. The mean-free sequence is normalized by dividing it by the corresponding rated value to obtain the active power fluctuation, voltage fluctuation, and frequency fluctuation of each node.

4. The method of claim 2, wherein, The construction of an independent perturbation estimation sequence for each of the multiple nodes based on their operational data includes: For each node, a linear decomposition model containing coefficients to be optimized is constructed based on the active power fluctuation, voltage fluctuation, and frequency fluctuation of the node; wherein, the linear decomposition model is used to extract the common component driven by voltage and frequency from the active power fluctuation of the node. Based on the linear decomposition model, the residual sequence of each node is constructed; The coefficients to be optimized are solved by minimizing the second-order correlation of the residual sequences between different nodes. Based on the optimized coefficients to be optimized, an independent perturbation estimation sequence is constructed for each node.

5. The method of claim 4, wherein, The construction of the residual sequence for each node based on the linear decomposition model includes: For each node, based on the voltage fluctuation and the corresponding voltage response coefficient, and the frequency fluctuation and the corresponding frequency response coefficient, the common component driven by both voltage and frequency is extracted from the active power fluctuation of the node to obtain the residual sequence of each node. The voltage response coefficient is used to characterize the sensitivity of node active power changes to voltage fluctuations, and the frequency response coefficient is used to characterize the sensitivity of node active power changes to frequency fluctuations.

6. The method of claim 4, wherein, The optimization of the coefficients to be optimized by minimizing the second-order correlation of the residual sequences between different nodes includes: Calculate the residual cross-correlation function of any two different nodes within a set lag window; The squares of the residual cross-correlation functions between all node pairs within the set lag window are summed to construct a global decorrelation objective function; The optimization coefficients are obtained by numerical optimization methods to minimize the global decorrelation objective function.

7. The method of claim 1, wherein, Solving the objective function for parameter optimization to obtain the parameters to be identified in the node frequency response model includes: The particle swarm optimization algorithm is used to perform global optimization on the parameters to be identified in order to obtain the optimal parameter vector that minimizes the objective function of the parameter optimization.

8. The method of claim 1-7, wherein, After obtaining the parameters of the node frequency response model based on the optimization objective function, the method further includes: The discrete transfer function of the node frequency response model is generated based on the parameters; Apply a unit step input to the discrete transfer function and calculate the corresponding unit step response sequence; The unit step response sequence is compared with the actual response from the simulation platform or measured data to verify the accuracy of the model identification results.

9. A device for identifying parameters of a power system node frequency response model, characterized in that, include: The acquisition module is used to acquire operating data of multiple nodes in the power system under stable operating conditions. A construction module is used to construct an independent perturbation estimation sequence for each of the multiple nodes based on the operational data of the multiple nodes; wherein the independent perturbation estimation sequence is obtained by minimizing the residual correlation between different nodes; A construction module is used to construct equivalent input correlation functions and output correlation functions for node frequency response model identification based on the independent perturbation estimation sequence; The identification module is used to establish an optimization objective function for identifying the parameters of the node frequency response model based on the equivalent input correlation function and the output correlation function, and to solve for the parameters of the node frequency response model based on the optimization objective function. The construction module is specifically used for: Based on the active power fluctuation of the target node and the aggregated amount of the independent perturbation estimation sequences of all other nodes besides the target node, the equivalent input correlation function is constructed. Based on the active power fluctuation and frequency fluctuation of the target node, the output correlation function is constructed. The construction module is further used for: The equivalent input correlation function is obtained by adding the product of the active power autocorrelation function of the target node and the cross-correlation function of the target proportional coefficient and the aggregation amount. The target proportional coefficient is a parameter to be optimized, used to characterize the equivalent contribution of power disturbances of other nodes to the frequency response of the target node; The identification module is specifically used for: Based on the preset model order, construct the node frequency response model transfer function containing the parameters to be identified; The transfer function is discretized to generate the corresponding unit impulse response sequence; Based on the unit impulse response sequence and the equivalent input correlation function, construct the output correlation function predicted by the model; A parameter optimization objective function is constructed with the goal of minimizing the error between the output correlation function predicted by the model and the actual output correlation function. Solve the objective function for parameter optimization to obtain the parameters to be identified in the node frequency response model.

Images