New energy grid-connected system inertia online identification method based on small signal injection

Abstract

The invention discloses a new energy grid-connected system inertia online identification method based on small signal injection, which relates to the technical field of signal identification, and comprises the following steps: injecting a small signal with a predetermined entropy feature into a power system by selecting an injection point to obtain an actual injection signal; recording injection timestamp data and acquiring electrical data of the pre-selected measurement points after injection; according to the preprocessed electrical data and the actual injection signal, calculating a signal entropy sequence of each pre-selected measurement point signal and mutual information amount of each measurement point signal and the injection signal, and calculating a bidirectional entropy flow matrix of each pre-selected measurement point; calculating an information entropy flow between measurement points by using a bidirectional entropy flow measurement method, and constructing an entropy flow network; extracting system inertia parameters based on entropy flow network features; and verifying and correcting the identification result. The initial inertia parameter is corrected through the entropy flow compensation algorithm, the recognition precision is remarkably improved, noise interference is effectively resisted, and the high-precision online recognition degree of the inertia of the new energy grid-connected system is achieved.

Description

Technical Field

[0001] This invention relates to the field of signal recognition technology, and in particular to an online inertia recognition method for new energy grid-connected systems based on small signal injection. Background Technology

[0002] Against the backdrop of energy transition and the construction of new power systems, the scale of new energy power generation is rapidly expanding. Solar and wind power, with their advantages of being clean and renewable, are gradually occupying an important position in the power supply system. However, the grid connection of new energy sources has brought significant changes to the inertia characteristics of the power system. The proportion of traditional synchronous generators is decreasing, the overall system inertia is reduced, seriously threatening the frequency stability of the power system, causing increased frequency fluctuations, and affecting the reliability and stability of power supply. Therefore, inertia identification of new energy grid-connected systems is crucial.

[0003] Current inertia identification methods have many limitations. For example, existing small-signal methods utilize random environmental perturbations or artificially injected pseudo-random sequences, combined with algorithms such as Kalman filtering and wavelet analysis to estimate inertia. However, the spatiotemporal coupling phenomenon between the wideband noise generated by power electronic converters and the injected signal significantly affects the identification accuracy. Specifically, the injected signal undergoes spatiotemporal distortion due to dynamic changes in the topology during propagation within the system, and simultaneously interacts complexly with noise generated by nonlinear power electronic devices in the frequency domain, making it difficult for traditional frequency domain analysis methods to effectively separate the injected signal response from the background noise. Because existing small-signal methods do not fully consider the nonlinear distortion of the signal transmission path caused by new energy power electronic devices, the inertia estimation error is high, resulting in low identification accuracy. Summary of the Invention

[0004] To address the issues of high estimation errors and low identification accuracy in existing technologies for inertia, this invention provides an online inertia identification method for new energy grid-connected systems based on small-signal injection. By correcting the initial inertia parameters through an entropy flow compensation algorithm, the method significantly improves identification accuracy, effectively resists noise interference, and achieves high-precision online inertia identification for new energy grid-connected systems. The specific technical solution is as follows: This invention provides an online inertia identification method for new energy grid-connected systems based on small-signal injection, comprising: By selecting an injection point, a small signal with predetermined entropy characteristics is injected into the power system to obtain the actual injected signal, and the injection timestamp data is recorded. Based on the injection timestamp data, electrical data of pre-selected measurement points are collected after injection. The electrical data is then pre-processed with median filtering and time alignment. The pre-selected measurement points are selected based on entropy flow sensitivity analysis. Based on the preprocessed electrical data and the actual injected signal, calculate the signal entropy sequence of each preselected measurement point signal and the mutual information between each measurement point signal and the injected signal, and calculate the bidirectional entropy flow matrix of each preselected measurement point. Based on the bidirectional entropy flow matrix, a system entropy flow network is constructed to identify response features related to the injected signal and obtain key response path data. Based on the bidirectional entropy flow matrix, effective entropy flow data is extracted using entropy flow statistical analysis and threshold generation methods. Based on the critical response path data and effective entropy flow data, calculate the preliminary inertia parameters; The initial inertia parameters are compensated using an entropy flow compensation algorithm to form corrected inertia parameters. The inertia identification results are then corrected based on the corrected inertia parameters, and the inertia identification results are output.

[0005] Preferably, injecting a small signal with predetermined entropy characteristics into the power system by selecting an injection point includes: Obtain the basic electrical parameters of the power system and construct a composite small signal with predetermined entropy characteristics based on the principle of maximizing information entropy; The electrical operating status data of the power system is acquired, and the composite small signal is optimized in combination with the basic electrical parameters to obtain the optimized injection amplitude. The composite small signal with optimized amplitude and predetermined entropy characteristics is injected into the power system through a selected injection point to form the actual injection signal.

[0006] Preferably, the acquisition of the basic electrical parameters of the power system, and the construction of a composite small signal with predetermined entropy characteristics based on the principle of maximizing information entropy, includes: Calculate the system time characteristic parameters based on the aforementioned basic electrical parameters, and perform entropy maximization optimization based on the set signal basic period to form entropy optimization distribution parameters; Based on the entropy optimization distribution parameters, a pseudo-random sequence is generated, and the periodic component is removed through autocorrelation analysis to form a low-autocorrelation random sequence. The low autocorrelation random sequence is subjected to spectral optimization to form an entropy-enhanced signal.

[0007] Preferably, the electrical data collected from the pre-selected measurement points after injection based on the injection timestamp data includes: Based on the selected injection point, the location of the pre-selected measurement point is determined through entropy flow sensitivity analysis, and a list of measurement point locations is generated. Based on the list of measurement point locations and the injected timestamp data, electrical data of each measurement point is collected synchronously within a defined time window.

[0008] Preferably, the pre-selected measurement points are selected based on entropy flow sensitivity analysis, including: Based on the topology of the power system and the selected injection point information, a system structure diagram is established to obtain a network topology model. The sensitivity of each node to small signal response is calculated using the entropy flow sensitivity analysis method, generating node sensitivity data; Based on the node sensitivity data, key measurement points are selected as pre-selected measurement points.

[0009] Preferably, the step of calculating the signal entropy sequence of each pre-selected measurement point signal and the mutual information between each measurement point signal and the injected signal based on the pre-processed electrical data and the actual injected signal, and calculating the bidirectional entropy flow matrix of each pre-selected measurement point includes: Based on the preprocessed electrical data and the actual injected signal, the data is processed in segments using the sliding time window method. The Shannon entropy calculation method is applied to the measured signal within each time window to obtain the signal entropy value sequence of each measurement point. By combining the Kullback-Leibler divergence calculation method, the mutual information between the signal at each measurement point and the injected signal is calculated to obtain the mutual information data; Based on the signal entropy sequence and mutual information data, the forward entropy flow and the forward entropy flow are calculated for each pair of measurement points using the transfer entropy calculation method, resulting in a bidirectional entropy flow matrix.

[0010] Preferably, the step of constructing a system entropy flow network based on the bidirectional entropy flow matrix, identifying response features related to the injected signal, and obtaining key response path data includes: Based on the bidirectional entropy flow matrix and network topology model, an entropy flow network is constructed with pre-selected measurement points as nodes and entropy flow as edge weights, thus obtaining the entropy flow network model. A network feature extraction algorithm is used to extract key path features from the entropy flow network model to obtain network feature parameters; Based on the network feature parameters, the most relevant response path to the injected signal in the system is identified through entropy flow tracing, and key response path data is obtained.

[0011] Preferably, the step of extracting effective entropy flow data based on the bidirectional entropy flow matrix, using entropy flow statistical analysis and threshold generation methods, includes: Based on the bidirectional entropy flow matrix, the distribution characteristics of the entropy flow values ​​are calculated using the entropy flow statistical analysis method to obtain entropy flow statistical characteristic data. Based on the entropy flow statistical characteristic data, the optimal entropy flow identification threshold under the current system state is calculated using a threshold generation method to obtain the entropy flow threshold parameter; The entropy flow threshold parameter is applied to the bidirectional entropy flow matrix to filter out effective entropy flow relationships that exceed the threshold, thereby obtaining effective entropy flow data.

[0012] Preferably, the calculation of preliminary inertia parameters based on the critical response path data and effective entropy flow data includes: Based on the effective entropy flow data and critical response path data, calculate the equivalent inertia characteristic parameters of the system; Based on the equivalent inertia characteristic parameters, the compensation value of the inertia parameters is calculated using the entropy flow compensation algorithm; The mapping relationship between entropy flow characteristics and inertia parameters is established through the inertia feature mapping model, and an inertia feature mapping table is obtained. Based on the preset information transmission efficiency parameters and inertia characteristic mapping table, the inertia parameters of the system are calculated, and preliminary inertia parameters are obtained. The inertia parameters include the equivalent inertia constant and the damping coefficient.

[0013] Preferably, the initial inertia parameters are compensated using an entropy flow compensation algorithm to form corrected inertia parameters, including: Based on the preliminary inertia parameters and entropy flow network model, analyze the nonlinear influencing factors in the system and generate nonlinear impact assessment data; The inertia compensation parameters are obtained by calculating the compensation value of the inertia parameters based on the nonlinear impact assessment data using the entropy flow compensation algorithm.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention discloses an online inertia identification method for new energy grid-connected systems based on small-signal injection. By injecting a small signal with predetermined entropy characteristics at a selected injection point, the method performs median filtering and time alignment preprocessing on the collected electrical data. It then calculates the signal entropy sequence, mutual information, and bidirectional entropy flow matrix to mine internal system features, accurately construct the system entropy flow network, and identify critical response paths. Effective entropy flow data is extracted based on entropy flow statistical analysis and threshold generation methods, which removes noise interference and improves identification accuracy. An entropy flow compensation algorithm is used to compensate the initial inertia parameters, further optimizing the inertia identification results and achieving high-precision online inertia identification, providing strong support for the stable operation of power systems. Attached Figure Description

[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0016] Figure 1 This is a flowchart of an online inertia identification method for a new energy grid-connected system based on small signal injection, according to the present invention.

[0017] Figure 2This is a flowchart of an embodiment of the online inertia identification method for a new energy grid-connected system based on small signal injection according to the present invention.

[0018] Figure 3 This is a flowchart of another embodiment of the online inertia identification method for a new energy grid-connected system based on small signal injection according to the present invention.

[0019] Figure 4 This is a flowchart of another embodiment of the online inertia identification method for a new energy grid-connected system based on small signal injection according to the present invention.

[0020] Figure 5 This is a flowchart of another embodiment of the online inertia identification method for a new energy grid-connected system based on small signal injection according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be understood that, when used in this specification, the terms “comprising” and “including” indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0023] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should also be further understood that the term "and / or" as used in this specification refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations.

[0025] Please refer to the following examples. Figures 1 to 5 .

[0026] like Figure 1 As shown, this invention provides an online inertia identification method for new energy grid-connected systems based on small-signal injection, comprising: Step S1: By selecting an injection point, inject a small signal with predetermined entropy characteristics into the power system to obtain the actual injected signal and record the injection timestamp data. Step S1 in this embodiment specifically includes the following steps: S11, Design composite feature small signal.

[0027] By constructing a composite small signal containing predetermined entropy characteristics, it is made to possess significant information entropy properties. Basic parameters such as the system's rated frequency, rated power, and system topology are read, and basic signal characteristics are designed based on the principle of maximizing information entropy. Combining Shannon entropy theory, a pseudo-random sequence satisfying the entropy value is constructed as the signal's basic sequence. By applying information entropy constraints and adjusting the signal's spectral distribution, an entropy-enhanced signal is obtained.

[0028] A composite small signal containing predetermined entropy characteristics is constructed to give it significant information entropy properties. By applying the principle of maximizing information entropy, a pseudo-random sequence satisfying the entropy value is constructed. Compared to traditional single-frequency or fixed-waveform small signals, the entropy-enhanced signal is richer in information capacity, with its entropy value approaching the theoretical maximum. This allows it to carry sufficient information even with relatively small injection amplitudes, enhancing its recognition capability in noisy environments.

[0029] S12. The amplitude of the injected signal is dynamically adjusted according to the system's operating status.

[0030] The system reads operational status data, including current load level and renewable energy output, and calculates an estimated noise level under the current system conditions, based on fundamental system parameters. Based on this noise level estimate, an initial signal injection amplitude is set, yielding the initial amplitude parameters. Then, by comprehensively analyzing system stability constraints and signal detectability requirements, the amplitude is adjusted to obtain the final optimized injection amplitude.

[0031] S13. Inject the designed small signal into the system through the selected injection point.

[0032] By analyzing the system topology, the optimal injection point location is selected, and injection location information is generated. The entropy-enhancing signal is combined with the optimized injection amplitude to generate the signal to be injected. Using the control interface of the new energy converter or a dedicated small-signal injection device as the optimal injection point location, the signal to be injected into the power system, forming the actual injected signal. The injection time and duration are recorded, generating injection timestamp data.

[0033] Step S2: Based on the injection timestamp data, collect electrical data of the pre-selected measurement points after injection, and perform median filtering and time alignment preprocessing on the electrical data. The pre-selected measurement points are selected based on entropy flow sensitivity analysis. Step S2 in this embodiment specifically includes the following steps: S21. Determine the location of the pre-selected measurement point based on entropy flow sensitivity analysis.

[0034] By reading the system topology and injection location information, a system structure diagram is established, resulting in a network topology model. Using entropy flow sensitivity analysis, the sensitivity of each node to small signal responses is calculated, generating node sensitivity data. Based on this data, the top N (N=5) nodes by sensitivity are selected as key measurement points (pre-selected measurement points), generating a list of measurement point locations.

[0035] S22. Acquire time synchronization data of key physical quantities in the system.

[0036] Deploy PMUs (Phasor Measurement Units) or high-precision data acquisition devices at the locations specified in the measurement point location list, and set the sampling frequency and synchronization clock source. Read the injected timestamp data, determine the acquisition time window, and synchronously acquire physical quantities such as frequency, active power, and reactive power at each measurement point within the determined time window to obtain the raw measurement dataset.

[0037] S23. Preprocess the collected raw data.

[0038] The original measurement dataset is read, and a mean-mode filtering method is applied to remove obvious outliers, resulting in a filtered dataset. Time alignment and interpolation are then performed on the filtered dataset to ensure strict temporal synchronization of all measurement points, generating an aligned dataset. The system's baseline trend and fluctuation components are then separated from the aligned dataset to obtain a preprocessed dataset.

[0039] Step S3: Based on the preprocessed electrical data and the actual injected signal, calculate the signal entropy sequence of each preselected measurement point signal, as well as the mutual information between each measurement point signal and the injected signal, and calculate the bidirectional entropy flow matrix of each preselected measurement point; Step S3 in this embodiment specifically includes the following steps: S31. Calculate the entropy characteristic index of the signal at each measurement point.

[0040] By reading the preprocessed dataset and the actual injected signal, the data is processed in segments using the sliding time window method. The Shannon entropy calculation method is applied to the measured signal within each time window to obtain the signal entropy sequence for each measurement point. Combined with the Kullback-Leibler divergence calculation method, the mutual information between the signal at each measurement point and the injected signal is calculated to obtain the mutual information data.

[0041] S32. Calculate the bidirectional entropy flow relationship between measurement points.

[0042] By reading the signal entropy sequence and mutual information data, a time-series relationship diagram between measurement points is established, yielding point-to-point time-series relationship data. Applying the transfer entropy calculation method, the forward entropy flow and transfer entropy flow are calculated for each pair of measurement points, resulting in a bidirectional entropy flow matrix.

[0043] In addition, the bidirectional entropy flow matrix is ​​normalized to generate a normalized bidirectional entropy flow matrix.

[0044] Step S4: Based on the bidirectional entropy flow matrix, construct the system entropy flow network, identify the response features related to the injected signal, and obtain key response path data; In step S4 of this embodiment, an entropy flow network model is constructed by reading the normalized bidirectional entropy flow matrix and the network topology model, with measurement points as nodes and entropy flow as edge weights. Using a network feature extraction algorithm, key path features, including path transmission efficiency and information attenuation rate, are extracted from the entropy flow network model to obtain network feature parameters. Based on these network feature parameters, entropy flow tracing is used to identify the response path in the system most relevant to the injected signal, thus obtaining key response path data.

[0045] Furthermore, by considering the forward entropy flow from the injection point to the measurement point and the propagation entropy flow from the measurement point to other measurement points, a complete system information transmission network is established, enabling the system to distinguish the injected signal response from background noise even in high-noise environments. In traditional methods, signal extraction mainly relies on time-frequency domain features. In high-proportion renewable energy grid-connected systems, the wideband noise generated by power electronic equipment causes traditional methods to fail. However, in this embodiment, by treating the information itself as the object of analysis, focusing only on the causal relationships and information transmission characteristics between signals, and not relying on the absolute amplitude or spectral characteristics of the signals, high recognition accuracy is maintained even under extremely low signal-to-noise ratio conditions.

[0046] Step S5: Based on the bidirectional entropy flow matrix, extract effective entropy flow data using entropy flow statistical analysis and threshold generation methods; Step S5 in this embodiment specifically includes the following steps: S51. Determine the effective signal recognition threshold based on the entropy flow distribution characteristics.

[0047] The normalized bidirectional entropy flow matrix is ​​read, and entropy flow statistical analysis methods are applied to calculate the distribution characteristics of entropy flow values, obtaining entropy flow statistical characteristic data. Based on the principle of maximizing the signal-to-noise ratio, the optimal entropy flow identification threshold under the current system state is calculated using a threshold generation method, obtaining the entropy flow threshold parameter. This entropy flow threshold parameter is then applied to the normalized bidirectional entropy flow matrix to filter out effective entropy flow relationships exceeding the threshold, obtaining effective entropy flow data.

[0048] S52. Extract system inertia characteristic parameters from the effective entropy flow relationship.

[0049] By reading effective entropy flow data and critical response path data, and applying the inertia characteristic mapping model, a mapping relationship between entropy flow characteristics and inertia parameters is established, resulting in an inertia characteristic mapping table. Information propagation speed and attenuation characteristics are extracted from the critical response path data, and propagation characteristic analysis methods are applied to calculate the system's information transmission efficiency, obtaining information transmission efficiency parameters. Based on the information transmission efficiency parameters and the inertia characteristic mapping table, the system's equivalent inertia characteristic parameters, including the equivalent inertia constant and damping coefficient, are calculated, yielding preliminary inertia parameters.

[0050] Step S6: Calculate the preliminary inertia parameters based on the key response path data and effective entropy flow data; In step S6 of this embodiment, inertia characteristic compensation is performed to address nonlinear effects.

[0051] By reading the preliminary inertia parameters and the entropy flow network model, the nonlinear influencing factors in the system are analyzed, and nonlinear impact assessment data is generated. Using the entropy flow compensation algorithm, based on the nonlinear impact assessment data, the compensation values ​​of the inertia parameters are calculated, yielding the inertia compensation parameters. The preliminary inertia parameters are combined with the inertia compensation parameters to generate the compensated corrected inertia parameters.

[0052] In this step, nonlinear influencing factors in the system are analyzed and a compensation model is established to compensate for inertia parameters based on topological characteristics and nonlinear responses. In power systems with a high proportion of renewable energy connected to the grid, the extensive integration of power electronic devices introduces significant nonlinear characteristics. Traditional inertia identification methods based on linear system theory are insufficient to accurately reflect the actual dynamic characteristics of the system. This innovation establishes a complete nonlinear compensation mechanism by analyzing the types and intensities of nonlinear sources and their sensitivity to inertia identification, combined with the topological characteristics of entropy flow networks. This compensation mechanism effectively solves the parameter identification deviation problems caused by the switching nonlinearity of power electronic devices, the nonlinearity of load dynamic characteristics, and the non-uniformity of the system topology.

[0053] Step S7: The initial inertia parameters are compensated using the entropy flow compensation algorithm to form the corrected inertia parameters. The inertia identification results are then corrected based on the corrected inertia parameters, and the inertia identification results are output.

[0054] In step S7 of this embodiment, a system response model is constructed by reading the corrected inertia parameters, the preprocessed dataset, and the actual injected signal to obtain theoretical response data. The theoretical response data is compared with the actual measured system response, and the fitting error and residual distribution characteristics are calculated to obtain fitting error data. Based on the fitting error data and entropy flow statistical characteristics data, the reliability index of the inertia estimation result is calculated to obtain the reliability assessment result.

[0055] By reading the corrected inertia parameters and reliability assessment results from multiple historical periods, a time-series inertia database is established to obtain historical inertia data. Through reliability-weighted fusion, the inertia parameters are weighted and averaged according to the reliability assessment results of each period to obtain fused inertia parameters. The temporal trend characteristics of the fused inertia parameters are analyzed to identify the variation patterns of system inertia and obtain inertia variation trend data.

[0056] This study integrates inertia estimation results from multiple time periods and analyzes their temporal trend characteristics. A weighted average of the inertia parameters is then applied based on the reliability assessment results. The inertia characteristics of power systems exhibit time-varying properties, especially in systems with a high proportion of renewable energy sources. With fluctuations in photovoltaic and wind power output, system inertia can change significantly within a short period. A single measurement is insufficient to fully reflect the dynamic changes in system inertia characteristics.

[0057] The system reads and integrates inertia parameters and inertia change trend data, applies a trend correction algorithm to correct the inertia estimate at the current moment, and obtains the final inertia parameters. The confidence interval and uncertainty of the final inertia parameters are calculated to generate inertia estimation accuracy data. The final inertia parameters, inertia estimation accuracy data, and reliability assessment results are combined to form a complete inertia identification result report, which is provided to the power system dispatch and control system.

[0058] This invention discloses an online inertia identification method for new energy grid-connected systems based on small-signal injection. By injecting a small signal with predetermined entropy characteristics at a selected injection point, the method performs median filtering and time alignment preprocessing on the collected electrical data. It then calculates the signal entropy sequence, mutual information, and bidirectional entropy flow matrix to mine internal system features, accurately construct the system entropy flow network, and identify critical response paths. Effective entropy flow data is extracted based on entropy flow statistical analysis and threshold generation methods, which removes noise interference and improves identification accuracy. An entropy flow compensation algorithm is used to compensate the initial inertia parameters, further optimizing the inertia identification results and achieving high-precision online inertia identification, providing strong support for the stable operation of power systems.

[0059] In one embodiment, for a typical high-proportion renewable energy grid-connected power system, the system includes conventional synchronous generator units and a large number of renewable energy power generation devices such as photovoltaic and wind power. System rated frequency. The system operates at 50Hz, with a total installed capacity of 1000MW, of which more than 60% is from renewable energy sources. In such a low-inertia system, traditional inertia identification methods face problems such as low signal-to-noise ratio and low identification accuracy. This embodiment details how to apply the method of this invention for online system inertia identification. The specific details are as follows: Step 1: Small Signal Injection Design and Implementation First, read the system's basic parameters: set the rated frequency. =50Hz, rated power =1000MW, system topology data T (including network structure with 37 nodes and 54 lines). Calculate the characteristic time constant of the system based on the system's rated inertia constant H. .

[0060] Based on the system's characteristic time constant, the fundamental period T of the signal is set. base = , is the system characteristic time constant.

[0061] Based on the principle of maximizing Shannon entropy, the optimal signal distribution characteristics are calculated. For discrete random signals, the formula for calculating Shannon entropy is: Where: H(X) is the Shannon entropy of signal X; The value taken from signal X The probability of; It is a logarithmic function with base 2.

[0062] To maximize entropy, the signal should have a uniform distribution characteristic, i.e. ≈1 / n, where n is the number of signal value types. In this embodiment, n=16 is chosen, such that H(X)_max≈ =4.

[0063] Based on the above entropy optimization distribution parameters, the initial pseudo-random sequence x(n) = (a·x(n-1) + c)modm is generated using the linear congruential method; where: x(n) is the nth value of the pseudo-random sequence; a is the multiplier; c is the increment; m is the modulus; and mod is the modulo operation. The initial seed value is x(0), generating an initial random sequence of length 1000.

[0064] Autocorrelation analysis was applied to calculate the autocorrelation function of the sequence, R(k) = E[x(n)·x(n+k)], where R(k) is the autocorrelation function, E is the expectation operator, x(n) is the nth value of the sequence, and k is the time delay. A periodic component was identified in the sequence; at k=128, R(k) = 0.32. Singular spectrum analysis was used to remove the periodic component, resulting in a low-autocorrelation random sequence, confirming that R(k) < 0.1 (k > 0).

[0065] An FFT (Fast Fourier Transform) analysis was performed on a low autocorrelation random sequence to obtain its spectral distribution F(ω). It was observed that the energy is mainly distributed between 0-10 Hz, while the system is most sensitive to the low-frequency band of 0.1 Hz-2 Hz.

[0066] ; Where: G(ω) is the spectrum optimization function; ω is the angular frequency.

[0067] By combining the sequence spectrum with the spectrum optimization function F_opt(ω)=F(ω)·G(ω), an entropy-enhanced signal in the time domain is generated through inverse FFT transformation. The final signal entropy value is verified to be 0.87, which is still within the target interval [0.8, 0.9].

[0068] The initial amplitude of the injected signal is set to 0.3% of the system's rated power, that is: P inj-initial =0.3%× =0.3%×1000MW=3MW Where: P inj-initial Inject an initial amplitude into the signal.

[0069] Based on the signal-to-noise ratio requirement SNR≥1 and the system stability constraint ΔP_max≤0.5% Adjust the injection amplitude. Calculate the signal-to-noise ratio (SNR) = P. inj - initial / P noise =3MW / 3MW=1, which just meets the requirement. Considering the current operating state and stability margin of the system, the initial injection amplitude of 3MW just meets the minimum signal-to-noise ratio requirement (SNR=1). In practice, to ensure more reliable identification, it is usually desirable to leave a margin, that is, to make the SNR slightly greater than 1. At this time, there is an optimization interval [3MW, 5MW] between the optimized injection amplitude of 3MW (just meeting the SNR) and 5MW (stability limit). The final optimized injection amplitude is P. inj_opt =3.5MW. Analyzing the system topology, the optimal injection point location is selected as node 10, which connects to a wind farm with an installed capacity of 200MW.

[0070] The entropy-enhanced signal is combined with the optimized injection amplitude to generate the signal P to be injected. inj (t)=P inj_opt ·s(t); P inj_opt To optimize the injection amplitude; s(t) is the normalized entropy enhancement signal, with a value range of [-1, 1].

[0071] The signal is injected through the converter control interface of the wind farm. The injection time is set to 10:30:00 on April 25, 2025, and the duration is 15 seconds to form an actual injected signal. The injection timestamp data is recorded.

[0072] Step 2: Multi-point collaborative data collection Read the system topology and injection location information to build a network topology model containing 37 nodes.

[0073] Using the entropy flow sensitivity analysis method, the sensitivity of each node to the small signal response is calculated: Wherein: S i The sensitivity of node i; I(Y) i ;X) represents the injected signal X and the response Y of node i. i Mutual information between them; H(Y) i Let S be the entropy value of node i's response. The calculated node sensitivity ranking is as follows: S_10 = 1.00 (injection point); S_12 = 0.83; S_15 = 0.79; S_8 = 0.71; S_22 = 0.68; S_5 = 0.64; S_31 = 0.57; S_18 = 0.53. Based on the node sensitivity data, the top 7 most sensitive nodes are selected as key measurement points: nodes 10, 12, 15, 8, 22, 5, and 31, forming a list of measurement point locations.

[0074] PMU devices were deployed at the seven locations specified in the measurement point location list, with a sampling frequency of 100Hz. Based on the injection timestamp data (April 25, 2025, 10:30:00), the acquisition time window was determined to be from 5 seconds before injection to 10 seconds after injection, i.e., 10:29:55 to 10:30:10. Within the determined time window, physical quantities such as frequency, active power, and reactive power were synchronously collected at each measurement point to obtain the original measurement dataset, which contains a data matrix of 7 nodes × 3 physical quantities × 1500 sampling points.

[0075] A mean-mode filtering method was applied to the original measurement dataset to remove obvious outliers. The filtering window width was 5 sampling points, resulting in a 98% reduction in outliers. The filtered dataset underwent time alignment and interpolation to ensure strict temporal synchronization of all measurement points, forming an aligned dataset. The time resolution was uniformly set to 10ms. A trend separation algorithm was then applied to separate the system's baseline trend and fluctuation components from the aligned dataset. Where: x(t) is the original signal; x_trend(t) is the trend component; x_fluct(t) is the fluctuation component. Wavelet decomposition is used to separate the trend, resulting in a preprocessed dataset containing both the trend and fluctuation components.

[0076] Step 3: Bidirectional Entropy Flow Measurement and Analysis The preprocessed dataset and the actual injected signal are read, and the data is processed in segments using a sliding time window method. The time window width is set to 1.5 seconds, and the window sliding step is 0.5 seconds. The Shannon entropy calculation method is applied to the measured signal within each time window. Where: H(X) is the entropy value of signal X, in bits; p(xᵢ) is the probability of the signal taking the value xᵢ, calculated by dividing the signal value into 20 equally wide intervals. The signal entropy value sequence of each measurement point is calculated. For example, the frequency signal entropy value sequence of node 10 (injection point) is: [3.82, 3.85, 4.01, 4.32, 4.29, 4.18, 4.05, 3.91, 3.84, 3.81].

[0077] Using the Kullback-Leibler divergence calculation method, the mutual information between the measured signal and the injected signal at each point is calculated: Where: I(X;Y) is the mutual information in bits; D_KL is the KL divergence; p(x,y) is the joint probability distribution of X and Y; p(x) and p(y) are the marginal probability distributions of X and Y, respectively. The mutual information data between each measurement point and the injected signal are calculated.

[0078] By reading the signal entropy sequence and mutual information data, time-delay embedding processing is performed on the signal at each measurement point to reconstruct the state space: Where: x*(t) is the reconstructed state vector; x(t) is the original time series; τ is the time delay parameter; and m is the embedding dimension. The time delay parameter τ is determined by the first local minimum of the mutual information function. In this embodiment, when m=3, the proportion of false nearest neighbors drops below 5%, therefore m=3 is chosen.

[0079] Based on the state space after time-delay embedding processing, calculate the conditional entropy of each measurement point: Where: H(X|X⁻) is the conditional entropy of signal X, in bits; p(x*(t), x*(t-1)) is the joint probability of the state vector at adjacent time points; p(x*(t)|x*(t-1)) is the conditional probability. For example, the conditional entropy of the frequency signal at node 12 is 1.28 bits.

[0080] Calculate the transfer entropy from the injection point (Y) to each measurement point (X) to quantify the positive information flow: Where: TE_{Y→X} is the transfer entropy from Y to X, in bits; p(x*(t+τ), x*(t), y*(t)) is the joint probability distribution; p(x*(t+τ)|x*(t), y*(t)) and p(x*(t+τ)|x*(t)) are conditional probabilities.

[0081] For any two measurement points X i and Xj Calculate the transit entropy between them: Where: TE_{X i →X j} is from X i To X j The transfer entropy is calculated by integrating the transfer entropy data of all measurement point pairs into a 7×7 matrix, resulting in the transfer entropy flow matrix.

[0082] By combining significant positive entropy flow data with a transitive entropy flow matrix, a directed entropy flow network is constructed with injection points as sources and measurement points as nodes. An entropy flow synthesis algorithm is then applied to weightedly synthesize the positive entropy flow and the transitive entropy flow. Where: TE_combined(i,j) is the combined entropy flow; TE_direct(i,j) is the forward entropy flow; TE_transfer(i,j) is the transfer entropy flow; α is the weight of the forward entropy flow; β is the weight of the transfer entropy flow.

[0083] The comprehensive entropy flow data is normalized to form a 7×7 normalized bidirectional entropy flow matrix.

[0084] Step 4: Construction and Analysis of Entropy Flow Network A directed weighted network is constructed using measurement points as nodes and normalized entropy flow values ​​as edge weights, forming the initial entropy flow network model. An entropy flow threshold θ is set, and edges with entropy flow values ​​greater than θ are retained to simplify the network structure, resulting in an optimized entropy flow network model. The optimized network contains 7 nodes and 11 edges.

[0085] Calculate the basic topological characteristics of the optimized entropy flow network model: Average node degree: d avg =2·E / N=2·11 / 7=3.14; Where: E is the number of edges; N is the number of nodes.

[0086] Average clustering coefficient C avg =(1 / N)·∑C i =0.37; Where: C i Let be the clustering coefficient of node i.

[0087] Average path length: L avg =(1 / (N·(N-1)))·∑∑d(i,j)=1.86; Where d(i,j) is the shortest path length between nodes i and j.

[0088] Generate network topology feature data, including features such as node degree distribution, clustering coefficient, and path length.

[0089] Calculate the transmission efficiency for each path in the optimized entropy flow network model: Where: E(P) i ) represents path P i The transmission efficiency; TE_norm(j,k) is the normalized entropy flow value between nodes j and k; ∏ represents the chain multiplication.

[0090] Based on the transmission efficiency values, identify the most efficient transmission paths in the network (the top 20% of efficiency values): P_1: 10→12→15, efficiency = 0.745; P_2: 10→8→22, efficiency = 0.621; P_3: 10→5→31, efficiency = 0.548; form a list of efficient paths, containing 3 main paths.

[0091] Calculate the temporal correlation between each efficient path and the injected signal: Where: r(P) i S inj ) represents path P i Correlation coefficient with the injected signal; S_P i For path P i The response signal on; S inj For the injected signal; τ i The optimal time delay τ is determined through sliding window cross-correlation analysis. i ; This represents the correlation analysis function.

[0092] The comprehensive relevance score is calculated using a path ranking algorithm based on normalized mutual information. Where: SC(P) i ) represents path P i The overall score; α is the relevance weight; β is the efficiency weight; For path Efficiency indicators.

[0093] Calculate the score: The two paths with the highest scores were selected as the critical response paths of the system: P_1: 10→12→15, score = 0.805; P_2: 10→8→22, score = 0.739; thus forming the critical response path data.

[0094] Step 5: Inertia Feature Extraction Based on Entropy Flow Read the normalized bidirectional entropy flow matrix and apply entropy flow statistical analysis methods to calculate the distribution characteristics of the entropy flow values. The mean entropy flow is μ. TE =0.31, standard deviation σ TE =0.26. Based on the principle of maximizing the signal-to-noise ratio, an adaptive threshold generation algorithm is applied to calculate the optimal entropy flow recognition threshold under the current system state: Where: θ_opt is the entropy flow threshold; μ TE σ is the mean of the entropy flow; TE Let θ_opt be the standard deviation of the entropy flow; k is a coefficient, determined based on the current signal-to-noise ratio (SNR). In this example, SNR = 1, corresponding to k = 0.8. The calculated θ_opt = 0.31 + 0.8 × 0.26 = 0.52. Applying the entropy flow threshold parameter to the normalized bidirectional entropy flow matrix, effective entropy flow relationships exceeding the threshold are selected, yielding effective entropy flow data. A total of 5 pairs of nodes have entropy flow values ​​exceeding the threshold, constituting effective entropy flow relationships.

[0095] Based on effective entropy flow data and critical response path data, an inverse relationship model between entropy flow propagation speed and system inertia constant is established: = in: H is the propagation speed of entropy flow, in kilometers per second; H is the system inertia constant, in seconds. The proportionality coefficient is obtained by fitting historical data. =240; The bias constant is obtained by fitting historical data. =5.

[0096] Establish a model showing the direct proportional relationship between entropy flow attenuation rate and system damping coefficient: = Where: α TE is the entropy flow attenuation rate, in bits per kilometer; D is the system damping coefficient, dimensionless. The proportionality coefficient is obtained by fitting historical data. =0.8; The bias constant is obtained by fitting historical data. =0.02.

[0097] Combining the two models above, a complete entropy flow-inertia mapping model is constructed, forming an inertia feature mapping table.

[0098] Read the critical response path data and the time series of each measurement point, and calculate the information transmission time delay for each path. The delay is determined by the maximum value of the cross-correlation function. For example, the physical distance from node 10 to node 12 is d_{10, 12} = 25 kilometers, and the information transmission time delay is τ_{10, 12} = 25 milliseconds. Calculate the propagation speed: v_{10, 12} = d_{10, 12} / τ_{10, 12} = 25 km / 0.025 s = 1000 km / s Where: v_{10, 12} is the information propagation speed from node 10 to node 12, in kilometers per second; d_{10, 12} is the physical distance, in kilometers; τ_{10, 12} is the time delay, in seconds.

[0099] Calculate the weighted average propagation velocity using the entropy flow value of each path as the weight: v avg =(TE_1·v_1+TE_2·v_2) / (TE_1+TE_2)==970.6 kilometers / second; Where: v avg TE_1 and TE_2 represent the entropy flow values ​​of path 1 and path 2, respectively; v_1 and v_2 represent the propagation speeds of path 1 and path 2, respectively.

[0100] The relationship between entropy flow value and propagation distance in critical response path data analysis was analyzed using an exponential decay model. Where: TE(d) is the entropy flow value at a distance of d; denoted as the initial entropy flow value; α is the attenuation coefficient, in units of 1 / km; d is the propagation distance, in units of km. Fitting path P_1 yields α_1 = 0.015 / km, and fitting path P_2 yields α_2 = 0.018 / km.

[0101] Considering the topological connectivity between measurement points, calculate the equivalent attenuation coefficient of the system: αeff=(TE_1·α_1+TE_2·α_2) / (TE_1+TE_2)==0.0164 / km Where: αeff is the equivalent attenuation parameter, with units of 1 / km; TE_1 and TE_2 represent the entropy flow values ​​of path 1 and path 2, respectively.

[0102] Read the average propagation velocity parameters and velocity-inertia mapping parameters, and use the inverse function to calculate the system's equivalent inertia constant: in This is the proportionality constant, with a value of 240; v TE The propagation speed of entropy flow; The bias constant has a value of 5. Read the equivalent attenuation parameter and attenuation-damping mapping parameter, and calculate the system's equivalent damping coefficient using the inverse function: D=(α TE - ) / =(0.0164-0.02) / 0.8=-0.0036 / 0.8=-0.0045 Where: D is the system equivalent damping coefficient, dimensionless; α TE Entropy flow attenuation rate, in units of 1 / km; This is the proportionality coefficient, with a value of 0.8; This is the bias constant, with a value of 0.02.

[0103] The negative value of the damping coefficient indicates the presence of negative damping in the current system, typically caused by the control characteristics of high-proportion power electronic devices. Because these devices (such as wind power and photovoltaic inverters) interact with the grid, their response characteristics may exhibit "negative resistance" at specific frequencies, effectively resulting in negative damping. This can lead to wideband oscillations and affect system stability. Therefore, based on experience with stable system operation, the damping coefficient is corrected to D = 0.005 in this case. Combining the inertia constant and damping coefficient into a set of dynamic characteristic parameters, we obtain the initial inertia parameters: H = 0.249s, D = 0.005.

[0104] By reading the initial inertia parameters and entropy flow network model, and combining them with the system's basic parameters, a simplified dynamic model of the system is established. Where: Δω(t) is the system angular frequency deviation, in radians per second; H is the system inertia constant, in seconds; ΔP(t) is the system power deviation, in per-unit value; D is the system damping coefficient, dimensionless; t is time, in seconds; ∫ is the integral sign.

[0105] The main nonlinear sources in the system were analyzed, and three main types of nonlinear influencing factors were identified: switching nonlinearity of power electronic equipment, intensity index, etc. =0.58 (proportion of new energy installed capacity); nonlinearity of load dynamic characteristics, intensity index =0.12 (based on historical data fitting); Inhomogeneity of system topology, strength index =0.25 (calculated based on network entropy). Sensitivity analysis was used to assess the impact of various nonlinear factors on the inertia identification results, and the sensitivity coefficient was calculated. ;in: The sensitivity coefficient for the first nonlinear factor; H is the system inertia constant; The strength of the first nonlinear factor. ; A negative sensitivity coefficient indicates that these nonlinear factors can lead to lower inertia identification results.

[0106] Read the entropy flow network model and network topology data to analyze the network's heterogeneity. Calculate the standard deviation of the degree distribution σ_d = 1.25, and the clustering coefficient C. avg =0.37. Construct the topology compensation function: C topo = ·σ_d+ ·(C avg -C_ref) Where: C topo σ_d is the topology compensation coefficient; C is the standard deviation of the degree distribution; avg C_ref is the average clustering coefficient; C_ref is the reference clustering coefficient, with a value of 0.5. and These are weighting coefficients, with values ​​of 0.02 and 0.05 respectively.

[0107] Calculate the topology compensation coefficient: C topo =0.02×1.25+0.05×(0.37-0.5)=0.025-0.0065=0.0185 The initial inertia parameters are combined with the topology compensation coefficients to calculate the inertia parameters after topology compensation: H topo =H·(1+C topo = 0.249 × (1 + 0.0185) = 0.249 × 1.0185 = 0.254 Wherein: H topo Here, H represents the topology-compensated inertia parameter, in seconds; H is the initial inertia constant, in seconds; C... topo is the topology compensation coefficient, which is dimensionless.

[0108] Read the data on nonlinear factors and sensitivity assessment data, and establish a nonlinear response compensation model. Calculate the combined impact of each nonlinear factor: C nonlin = =-0.2501 Where: C nonlin The nonlinear compensation coefficient is dimensionless; s i f is the sensitivity coefficient; i This represents the intensity of the nonlinear factor. A negative value indicates that the nonlinear factor causes the inertia identification result to be too low, and it needs to be corrected upwards.

[0109] By combining the topology-compensated inertia parameters with the nonlinear compensation coefficients, the fully compensated inertia parameters are calculated: H comp =H topo / (1+C nonlin )=0.254 / (1-0.2501)=0.254 / 0.7499=0.339s Wherein: H comp The inertia parameter after nonlinear compensation, in seconds; H topo C represents the topology-compensated inertia parameter, in seconds. nonlin This is a nonlinear compensation coefficient, dimensionless.

[0110] Based on the reliability assessment of each compensation stage, reasonable weighting coefficients are set: preliminary inertia parameter weights. =0.3 (basic reliability); topology compensation weight =0.3 (moderate correction); nonlinear compensation weight =0.4 (larger correction level); The weighted fusion algorithm is applied to calculate the final equivalent inertia parameters: H final =( H+ ·H topo + ·H comp ) / ( + + ) = 0.287s Wherein: H final This is the final value of the equivalent inertia, in seconds; w i These are the weighting coefficients; H, H topo and H comp These are the initial, topology-compensated, and nonlinear-compensated inertia parameters, respectively. The final equivalent damping parameter D is calculated using the same method. final =0.007, together with the final value of the equivalent inertia, forms the set of dynamic characteristic parameters of the system, forming the corrected inertia parameters: H=0.287s, D=0.007.

[0111] Step 6: Correction of inertia estimation results Read the corrected inertia parameters, preprocessed dataset, and actual injected signal to construct a system response model: Where: Δω(s) is the Laplace transform of the system angular frequency deviation; H is the system inertia constant; D is the system damping coefficient; ΔP(s) is the Laplace transform of the system power deviation; and s is the Laplace variable. Using the actual injected signal as input, the system frequency response predicted by the model is calculated to form theoretical response data.

[0112] Compare the theoretical response data with the actual measured system response, and calculate the fitting error: Where: RMSE is the root mean square error, in Hertz; N is the number of data points; y i This is the actual measured value; i These are the model's predicted values. The normalized root mean square error (NRMSE) = RMSE / σ_y = 0.018 / 0.072 = 0.25, where σ_y is the standard deviation of the actual measured values. An NRMSE less than 0.3 indicates good fitting accuracy. The distribution characteristics of the residuals are analyzed. The residual mean μ_r = 0.0025 Hz, and the standard deviation σ_r = 0.017 Hz, which is close to a normal distribution and passes the Kolmogorov-Smirnov normality test (p-value = 0.32 > 0.05).

[0113] Using reliability assessment methods, the reliability index of the inertia estimation results is calculated: Where: R is the reliability index, with a value range of [0, 1]; NRMSE is the normalized root mean square error; μ_r is the mean residual; σ_y is the standard deviation of the actual measurement value; D_KL(p_r||p_norm) is the KL divergence between the residual distribution and the standard normal distribution. The reliability assessment result R=0.667 indicates that the reliability of this inertia identification result is at a medium to high level.

[0114] Read the corrected inertia parameters and reliability assessment results for the current and multiple historical periods. Historical data includes the results of the previous four measurements: Time period 1 (2 days ago): =0.295s, =0.72; Time Segment 2 (1 day ago, morning): =0.275s, =0.63; Time slot 3 (afternoon 1 day ago): =0.281s, =0.68; Time Segment 4 (morning of the day): =0.283s, =0.70; Current time period: =0.287s, =0.667; A reliability-weighted fusion algorithm is applied to perform a weighted average of the inertia parameters based on the reliability assessment results for each time period: Where: H_fusion is the fusion inertia parameter; R i H represents the reliability index for the i-th time period; i Let be the inertia parameter for the i-th time period.

[0115] The time trend of the inertia parameter was analyzed. Linear regression was performed on the H value for five time periods, yielding a slope k = -0.0022 s / day, indicating a slight downward trend in system inertia. This is consistent with the recent slight increase in the proportion of new energy installations.

[0116] Based on the time trend and the current system operating status, predict the trend of system inertia change in the short term (next 12 hours): Where: H_trend(t) is the inertia trend value at time t, H_fusion is the fused inertia parameter, k is the rate of change of inertia, and t is the prediction time.

[0117] Read the fused inertia parameters and inertia change trend data, and apply a trend correction algorithm to correct the inertia estimate at the current moment: Wherein: H final H_fusion is the final inertia parameter, in seconds; w_trend is the fused inertia parameter, in seconds; k is the trend weight, with a value of 0.5; k is the rate of change of inertia, in seconds / day; Δt is the prediction time window, with a value of 0.1 days.

[0118] Calculate the confidence interval and uncertainty of the final inertia parameter. Based on statistical analysis of historical data, the standard error of the system inertia parameter is SE = 0.008s.

[0119] The 95% confidence interval is: H final ±1.96·SE=0.285±1.96×0.008=0.285±0.0157=[0.269, 0.301]s; where: SE is the standard error in seconds; 1.96 is the coefficient corresponding to the 95% confidence level.

[0120] Relative uncertainty U_r = 2·SE / H final =2×0.008 / 0.285=0.016 / 0.285=5.6%; where: U_r is the relative uncertainty; SE is the standard error; H final This is the final inertia parameter.

[0121] The final inertia parameters, inertia estimation accuracy data, and reliability assessment results are combined to form a complete inertia identification result report: System equivalent inertia constant: H=0.285s; 95% confidence interval: [0.269, 0.301]s; Relative uncertainty: 5.6%; Estimated reliability index: 0.667; System equivalent damping coefficient: D=0.007; Inertia change trend: Slightly decreasing, slope k=-0.0022s / day.

[0122] This method was compared with traditional frequency response analysis and time-domain parameter identification methods under the same test conditions. The test system was the high-proportion renewable energy grid-connected power system described in the previous embodiment. The identification accuracy and robustness of the three methods were compared by testing under different signal-to-noise ratio conditions.

[0123] The actual system inertia H_true = 0.28s was determined through multiple transient tripping tests. The identification results and relative errors of the three methods under different signal-to-noise ratios are shown in the table below: The results show that under high signal-to-noise ratio (SNR) conditions (SNR=2), all three methods achieve good recognition accuracy. However, as the SNR decreases, the method of this invention exhibits a significant advantage. Under extremely low SNR conditions (SNR=0.1), the traditional frequency response method fails to converge, and the time-domain identification method has an error approaching 30%, while the method of this invention still maintains high recognition accuracy with a relative error of only 3.2%. This result verifies that the method based on bidirectional entropy flow measurement has superior anti-interference capability and recognition accuracy in high-noise environments, and is particularly suitable for low-inertia power systems with a high proportion of renewable energy grid connection.

[0124] This embodiment effectively captures the dynamic characteristics of information transmission in the system by measuring bidirectional entropy flow and constructing an entropy flow-inertia mapping relationship, and can accurately identify the system inertia parameters even under extremely low signal-to-noise ratio conditions.

[0125] The injected signal forms a path with maximum information transmission efficiency in the system, and the information transmission characteristics along this path directly reflect the system's inertia characteristics. By measuring this information transmission efficiency, the system inertia can be identified without relying on the signal amplitude and phase. It is unaffected by the system noise amplitude, focusing only on information transmission efficiency; it can effectively distinguish the response caused by the injected signal from the system's own disturbances; it is robust to system nonlinear characteristics and topology changes; and it does not rely on prior knowledge of system parameters.

[0126] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0127] 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 or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the specification of the present invention.

Claims

1. A method for online inertia identification of a new energy grid-connected system based on small-signal injection, characterized in that, include: By selecting an injection point, a small signal with predetermined entropy characteristics is injected into the power system to obtain the actual injected signal, and the injection timestamp data is recorded. Based on the injection timestamp data, electrical data of pre-selected measurement points are collected after injection. The electrical data is then pre-processed with median filtering and time alignment. The pre-selected measurement points are selected based on entropy flow sensitivity analysis. Based on the preprocessed electrical data and the actual injected signal, calculate the signal entropy sequence of each preselected measurement point signal, as well as the mutual information between each measurement point signal and the injected signal, and calculate the bidirectional entropy flow matrix of each preselected measurement point; the bidirectional entropy flow matrix includes the forward entropy flow matrix and the transitive entropy flow matrix; Based on the bidirectional entropy flow matrix, a system entropy flow network is constructed to identify response features related to the injected signal and obtain key response path data. Based on the bidirectional entropy flow matrix, effective entropy flow data is extracted using entropy flow statistical analysis and threshold generation methods. Based on the critical response path data and effective entropy flow data, calculate the preliminary inertia parameters; The initial inertia parameters are compensated using an entropy flow compensation algorithm to form corrected inertia parameters. The inertia identification results are then corrected based on the corrected inertia parameters, and the inertia identification results are output.

2. The method for online inertia identification of a new energy grid-connected system based on small-signal injection according to claim 1, characterized in that, The process of injecting a small signal with predetermined entropy characteristics into the power system by selecting an injection point includes: Obtain the basic electrical parameters of the power system and construct a composite small signal with predetermined entropy characteristics based on the principle of maximizing information entropy; The electrical operating status data of the power system is acquired, and the composite small signal is optimized in combination with the basic electrical parameters to obtain the optimized injection amplitude. The composite small signal with optimized amplitude and predetermined entropy characteristics is injected into the power system through a selected injection point to form the actual injection signal.

3. The method for online inertia identification of a new energy grid-connected system based on small-signal injection according to claim 2, characterized in that, The acquisition of the basic electrical parameters of the power system, and the construction of a composite small signal with predetermined entropy characteristics based on the principle of maximizing information entropy, includes: Calculate the system time characteristic parameters based on the aforementioned basic electrical parameters, and perform entropy maximization optimization based on the set signal basic period to form entropy optimization distribution parameters; Based on the entropy optimization distribution parameters, a pseudo-random sequence is generated, and the periodic component is removed through autocorrelation analysis to form a low-autocorrelation random sequence. The low autocorrelation random sequence is subjected to spectral optimization to form an entropy-enhanced signal.

4. The method for online inertia identification of a new energy grid-connected system based on small-signal injection according to claim 1, characterized in that, The electrical data collected from the pre-selected measurement points after injection based on the injection timestamp data includes: Based on the selected injection point, the location of the pre-selected measurement point is determined through entropy flow sensitivity analysis, and a list of measurement point locations is generated. Based on the list of measurement point locations and the injected timestamp data, electrical data of each measurement point is collected synchronously within a defined time window.

5. The method for online inertia identification of a new energy grid-connected system based on small-signal injection according to claim 4, characterized in that, The pre-selected measurement points were chosen based on entropy flow sensitivity analysis and include: Based on the topology of the power system and the selected injection point information, a system structure diagram is established to obtain a network topology model. The sensitivity of each node to small signal response is calculated using the entropy flow sensitivity analysis method, generating node sensitivity data; Based on the node sensitivity data, key measurement points are selected as pre-selected measurement points.

6. The method for online inertia identification of a new energy grid-connected system based on small-signal injection according to claim 1, characterized in that, The step of calculating the signal entropy sequence of each pre-selected measurement point signal, the mutual information between each measurement point signal and the injected signal, and calculating the bidirectional entropy flow matrix of each pre-selected measurement point based on the pre-processed electrical data and the actual injected signal includes: Based on the preprocessed electrical data and the actual injected signal, the data is processed in segments using the sliding time window method. The Shannon entropy calculation method is applied to the measured signal within each time window to obtain the signal entropy value sequence of each measurement point. By combining the Kullback-Leibler divergence calculation method, the mutual information between the signal at each measurement point and the injected signal is calculated to obtain the mutual information data; Based on the signal entropy sequence and mutual information data, the forward entropy flow and the forward entropy flow are calculated for each pair of measurement points using the transfer entropy calculation method, resulting in a bidirectional entropy flow matrix.

7. The method for online inertia identification of a new energy grid-connected system based on small-signal injection according to claim 6, characterized in that, The step of constructing a system entropy flow network based on the bidirectional entropy flow matrix, identifying response features related to the injected signal, and obtaining key response path data includes: Based on the bidirectional entropy flow matrix and network topology model, an entropy flow network is constructed with pre-selected measurement points as nodes and entropy flow as edge weights, thus obtaining the entropy flow network model. A network feature extraction algorithm is used to extract key path features from the entropy flow network model to obtain network feature parameters; Based on the network feature parameters, the most relevant response path to the injected signal in the system is identified through entropy flow tracing, and key response path data is obtained.

8. The method for online inertia identification of a new energy grid-connected system based on small-signal injection according to claim 7, characterized in that, The step of extracting effective entropy flow data based on the bidirectional entropy flow matrix, using entropy flow statistical analysis and threshold generation methods, includes: Based on the bidirectional entropy flow matrix, the distribution characteristics of the entropy flow values ​​are calculated using the entropy flow statistical analysis method to obtain entropy flow statistical characteristic data. Based on the entropy flow statistical characteristic data, the optimal entropy flow identification threshold under the current system state is calculated using a threshold generation method to obtain the entropy flow threshold parameter; The entropy flow threshold parameter is applied to the bidirectional entropy flow matrix to filter out effective entropy flow relationships that exceed the entropy flow threshold parameter, thereby obtaining effective entropy flow data.

9. The method for online inertia identification of a new energy grid-connected system based on small-signal injection according to claim 7, characterized in that, The calculation of preliminary inertia parameters based on the critical response path data and effective entropy flow data includes: Based on the effective entropy flow data and critical response path data, calculate the equivalent inertia characteristic parameters of the system; Based on the equivalent inertia characteristic parameters, the compensation value of the inertia parameters is calculated using the entropy flow compensation algorithm; The mapping relationship between entropy flow characteristics and inertia parameters is established through the inertia feature mapping model, and an inertia feature mapping table is obtained. Based on the preset information transmission efficiency parameters and inertia characteristic mapping table, the inertia parameters of the system are calculated, and preliminary inertia parameters are obtained. The inertia parameters include the equivalent inertia constant and the damping coefficient.

10. The method for online inertia identification of a new energy grid-connected system based on small-signal injection according to claim 7, characterized in that, The initial inertia parameters are compensated using an entropy flow compensation algorithm to form corrected inertia parameters, including: Based on the preliminary inertia parameters and entropy flow network model, analyze the nonlinear influencing factors in the system and generate nonlinear impact assessment data; The inertia compensation parameters are obtained by calculating the compensation value of the inertia parameters based on the nonlinear impact assessment data using the entropy flow compensation algorithm.

Images