A method and related apparatus for estimating the inertia and damping of a power system

Abstract

This invention discloses a method and related apparatus for estimating the inertia and damping of a power system. Based on disturbance data of the power system, this invention obtains variables reflecting inertia characteristics during the inertia-dominated phase and variables reflecting damping characteristics during the damping-dominated phase. It employs artificial intelligence technology to obtain initial inertia and damping, and optimizes the initial inertia and damping by combining the measured active power of the power system. Based on the optimization results, simulations under the same disturbance excitation are compared to determine the final inertia and damping. This method is applicable to power systems with synchronous generators, power electronic interface power supplies, and high proportions of renewable energy access. It enables online identification and dynamic estimation of the inertia and damping characteristics of the power system under disturbance or operational fluctuation conditions, thereby supporting power system frequency stability analysis, inertia assessment, and adaptive optimization of frequency regulation control parameters.

Description

Technical Field

[0001] This invention relates to a method and related device for estimating the inertia and damping of a power system, belonging to the field of power system automation and intelligent control technology. Background Technology

[0002] Inertia and damping of a power system are key indicators reflecting its frequency stability. Inertia determines the initial frequency change rate when the system is subjected to disturbances, while damping affects the attenuation rate and recovery capability of the system's frequency deviation. In traditional power systems, inertia support is mainly provided by synchronous generators, and their equivalent inertia constant can be calculated from unit parameters and rotor energy. However, with the integration of a high proportion of power electronic interface power sources (such as photovoltaic power generation and wind power converters), the system's equivalent inertia shows a significant decrease and becomes dynamically variable, making the real-time and accurate estimation of inertia and damping a crucial foundation for maintaining the safe and stable operation of the power grid. Existing methods for estimating inertia and damping mainly include the small disturbance linearization method, power change response method, step disturbance method, and energy balance method. However, with the expansion of system scale and the increase in renewable energy penetration, these methods are difficult to accurately estimate inertia and damping under complex topologies and multi-disturbance scenarios. Summary of the Invention

[0003] This invention provides a method for estimating the inertia and damping of a power system, which solves the problems disclosed in the background art.

[0004] According to one aspect of this application, a method for estimating the inertia and damping of a power system is provided, comprising:

[0005] Based on the disturbance data of the power system, variables reflecting the characteristics of inertia at each moment of the inertia-dominated stage and variables reflecting the characteristics of damping at each moment of the damping-dominated stage are obtained; among them, the inertia-dominated stage and the damping-dominated stage are two sequential stages in the disturbance stage.

[0006] Input all variables into the estimation model to obtain the initial inertia and damping of the power system;

[0007] The active power of the power system is calculated using the initial inertia and damping. Based on the calculated active power and the measured active power, the initial inertia and damping are optimized to obtain the optimal inertia and damping.

[0008] A power system model is constructed using optimal inertia and damping. Simulations are performed on the power system model and the actual power system under the same disturbance excitation. If the simulation comparison results meet the preset requirements, the optimal inertia and optimal damping are the estimated results.

[0009] Furthermore, the disturbance data includes the frequency and active power at each moment during the disturbance phase;

[0010] Before acquiring the variables, the process also includes a step of dividing the disturbance phase into an inertia-dominated phase and a damping-dominated phase based on the power system disturbance data. This step includes:

[0011] Based on the first derivative of frequency, the first derivative of frequency deviation, and the active power change rate, the disturbance initiation time, the power system stabilization time, and the inflection point of the active power change rate are determined respectively. The period from the disturbance initiation time to the inflection point of the active power change rate is defined as the inertia-dominated period, and the period from the inflection point of the active power change rate to the power system stabilization time is defined as the damping-dominated period. Among these, the frequency deviation is the difference between the frequency and the rated frequency of the power system.

[0012] Furthermore, the disturbance data includes the frequency and active power at each moment during the disturbance phase;

[0013] Based on the power system disturbance data, variables reflecting inertia characteristics at each moment of the inertia-dominated phase and variables reflecting damping characteristics at each moment of the damping-dominated phase are obtained, including:

[0014] From the disturbance data of the power system, obtain the frequency and active power at each moment of the inertia-dominated phase, the frequency and active power at each moment of the previous period, and the frequency and active power at each moment of the damping-dominated phase; where the previous period is a period of duration L that is located before and adjacent to the inertia-dominated phase, and L is a preset value.

[0015] Based on the frequency deviation at each moment, the rate of change of frequency between adjacent moments, and the active power at each moment, variables reflecting the characteristics of inertia at each moment in the inertia-dominated stage and variables reflecting the characteristics of damping at each moment in the damping-dominated stage are constructed; where the frequency deviation is the difference between the frequency and the rated frequency of the power system.

[0016] The variable s at time t t for:

[0017] ;

[0018] In the formula, the variable at time t1 , , The frequency deviation at time t1 The rate of change of frequency between time t1 and time t1-1 Let t1 be the active power.

[0019] Furthermore, the estimation model employs a multilayer perceptron neural network, with the loss function during training being:

[0020] ;

[0021] In the formula, L(θ) is the loss value, N is the total number of training samples, and H... i and D i These are the true values ​​of inertia and damping output by the multilayer perceptron neural network for the i-th training sample, respectively. These are the inertia estimate and damping estimate output by the multilayer perceptron neural network for the i-th training sample, respectively, where λ is the weighting factor. Let be the measured active power of the power system at time t. These are the inertia estimates and damping estimates output by the multilayer perceptron neural network, respectively. To adopt The active power of the power system at time t is calculated.

[0022] Furthermore, the objective of optimizing the initial inertia and damping is to minimize the active power response residual of the power system. The formula for the active power response residual of the power system is:

[0023] ;

[0024] In the formula, J(H,D) represents the active power response residual of the power system, and n represents the total number of sampling points. Let be the measured active power of the power system at time t. These are the inertia estimates and damping estimates output by the multilayer perceptron neural network, respectively. To adopt The active power of the power system at time t is calculated.

[0025] Furthermore, a power system model is constructed using optimal inertia and damping, including:

[0026] The optimal inertia and damping are mapped to the generalized active power-frequency admittance model, and the circuit equivalence relationship between inertia and damping is established. This circuit equivalence relationship is then used as the power system model, and the formula is as follows:

[0027] ;

[0028] In the formula, These are the generalized active power-frequency admittance and the Laplace operator, respectively. For the damping equivalent conductance, D * For optimal damping, H is the equivalent capacitance due to inertia. * For optimal inertia, α and β are proportionality coefficients, and S N and ω N The rated capacity and rated angular frequency of the power system.

[0029] Furthermore, the preset requirement is that the root mean square error between the simulation results of the actual power system and the simulation results of the power system model is less than or equal to a first threshold.

[0030] Furthermore, the method also includes, if the simulation comparison results do not meet the preset requirements, retraining the estimation model, obtaining the optimal inertia and damping based on the retrained estimation model, and re-performing the simulation.

[0031] According to another aspect of this application, an inertia and damping estimation device for a power system is provided, comprising:

[0032] The variable acquisition module acquires variables reflecting inertia characteristics at each moment of the inertia-dominated stage and variables reflecting damping characteristics at each moment of the damping-dominated stage, based on the disturbance data of the power system; wherein, the inertia-dominated stage and the damping-dominated stage are two sequential stages in the disturbance stage.

[0033] The estimation module inputs all variables into the estimation model to obtain the initial inertia and damping of the power system;

[0034] The optimization module uses the initial inertia and damping to calculate the active power of the power system. Based on the calculated active power and the measured active power, it optimizes the initial inertia and damping to obtain the optimal inertia and damping.

[0035] The verification module constructs a power system model using optimal inertia and damping. Under the same disturbance excitation, it simulates both the power system model and the actual power system. If the simulation comparison results meet the preset requirements, the optimal inertia and optimal damping are the estimated results.

[0036] According to another aspect of this application, a computer-readable storage medium is provided that stores one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform a method for estimating the inertia and damping of a power system.

[0037] According to another aspect of this application, a computer device is provided, including one or more processors and one or more memories, wherein one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing a method for estimating the inertia and damping of a power system.

[0038] The beneficial effects achieved by this invention are as follows: Based on the disturbance data of the power system, this invention obtains variables reflecting the characteristics of inertia in the inertia-dominated stage and variables reflecting the characteristics of damping in the damping-dominated stage. It uses artificial intelligence technology to obtain the initial inertia and damping, and combines the measured active power of the power system to optimize the initial inertia and damping. Based on the optimization results, it performs simulation comparison under the same disturbance excitation to determine the final inertia and damping. It can be applied to power systems with synchronous generators, power electronic interface power supplies, and high proportion of renewable energy access. It can realize online identification and dynamic estimation of the inertia and damping characteristics of the power system under disturbance or operational fluctuation conditions, thereby supporting power system frequency stability analysis, inertia assessment, and adaptive optimization of frequency regulation control parameters. Attached Figure Description

[0039] Figure 1 A flowchart of a method for estimating the inertia and damping of a power system;

[0040] Figure 2 This is a schematic diagram of a multilayer perceptron neural network;

[0041] Figure 3 This is a block diagram of a device for estimating the inertia and damping of a power system. Detailed Implementation

[0042] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit this application or its application or use. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0043] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application.

[0044] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0045] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0046] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0047] It should be noted that similar symbols and letters in the following figures represent similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0048] Furthermore, in the description of the embodiments of this application, the terms "first," "second," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance. Therefore, features defined with "first" or "second" may explicitly or implicitly include one or more features.

[0049] Artificial intelligence (AI) is a field that studies and develops theories, methods, technologies, and application systems to simulate, extend, and expand human intelligence. It attempts to understand the essence of intelligence and produce new intelligent machines that can react in a way similar to human intelligence. AI technology can encompass several major areas, including computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0050] This application provides a method for estimating the inertia and damping of a power system by integrating artificial intelligence technology. It aims to achieve rapid preliminary estimation through an AI model, and by combining physical consistency correction and generalized circuit equivalent mapping, to achieve high-precision, real-time estimation of the inertia and damping of the power system. This provides a basis for frequency stability analysis and control strategy optimization, and is applicable to the fields of power system automation and intelligent control technology. This estimation method can be executed by an estimation device, which can be a terminal device or a server. The terminal device can include, but is not limited to, mobile phones, computers, etc., as described in this application. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, big data, and artificial intelligence platforms, etc., as described in this application. Optionally, this estimation method can also be executed collaboratively by multiple electronic devices with computing power. For ease of explanation, subsequent embodiments will be described as being executed by an estimation device.

[0051] See Figure 1 , Figure 1 This is a flowchart of a method for estimating the inertia and damping of a power system according to an embodiment of this application. The estimation method can be executed by an estimation device and may include at least the following steps:

[0052] Step 1: Based on the disturbance data of the power system, obtain the variables reflecting the inertia characteristics at each moment of the inertia-dominated stage and the variables reflecting the damping characteristics at each moment of the damping-dominated stage; wherein, the inertia-dominated stage and the damping-dominated stage are two sequential stages in the disturbance stage.

[0053] It should be noted that the disturbance data can mainly include the frequency and active power at each moment of the disturbance phase, and these data can be collected in real time through a synchronous phasor measurement unit or a monitoring system.

[0054] To achieve effective estimation, the raw data needs to be preprocessed after obtaining it, such as detrending, smoothing filtering, standardization, and outlier removal, in order to improve data quality and reduce estimation errors.

[0055] Detrending can be achieved using linear regression, and the formula can be expressed as:

[0056] ;

[0057] In the formula, x(t) represents the original data (frequency or active power) at time t, a0 and a1 are the least squares fitting coefficients, representing the initial offset and linear trend of the signal, respectively. dt (t) represents the data after detrending.

[0058] Smoothing filtering can be achieved using the Savitzky-Golay method, and the formula can be expressed as:

[0059] ;

[0060] In the formula, x f (t) represents the data after smoothing and filtering, c k These are the filter window weight coefficients. This is the length of the original data acquisition window, which should be centered as much as possible on the time when the disturbance begins.

[0061] The standardized formula can be expressed as:

[0062] ;

[0063] In the formula, For standardized data, These are the data mean and standard deviation, respectively, to unify signals with different dimensions to the same scale.

[0064] To obtain variables reflecting inertia characteristics and damping characteristics, the disturbance stage needs to be divided into an inertia-dominated stage and a damping-dominated stage. The specific process is as follows:

[0065] 11) Based on the first derivative of frequency, the first derivative of frequency deviation, and the rate of change of active power, determine the disturbance start time, the power system stabilization time, and the inflection point of the rate of change of active power during the disturbance phase, respectively; where the frequency deviation is the difference between the frequency and the rated frequency of the power system.

[0066] Define the first derivative of frequency as:

[0067] ;

[0068] The frequency deviation is:

[0069] ;

[0070] In the formula, RoCoF(t) is the rate of change of frequency (Hz / s), and f(t) is the frequency. For frequency deviation, f nom The rated frequency of the power system.

[0071] If the absolute value of the rate of change of frequency at time t0 is greater than the second threshold, then time t0 can be determined as the starting point of the disturbance; if t st If the power system returns to stability at time t, then determine t. st The time is the time of stabilization; if t d The rate of change of active power increases before time t d The rate of change of active power before time t decreases, or t d The rate of change of active power before time t decreases, d If the rate of change of active power increases before time t, then t is determined to be... d The time is the inflection point of the rate of change of active power; the second threshold can be in the range of 0.02 to 0.05 Hz / s. The criterion for the power system to recover stability is: the absolute value of the first derivative of the frequency deviation is less than the third threshold for a duration exceeding Ts. The third threshold can be in the range of 0.005 to 0.01 Hz / s. Ts is the steady-state duration threshold, and the time is usually set to 2 to 5 seconds.

[0072] 12) The period from the start of the disturbance to the inflection point of the active power change rate is defined as the inertia-dominated period, and the period from the inflection point of the active power change rate to the power system stabilization point is defined as the damping-dominated period; that is, [t0, t...] d As the inertia-dominated phase, [t] d , t st This is the damping-dominated stage.

[0073] In the rotor motion equations of a power system, changes in active power are the fundamental driving force causing frequency fluctuations. The rate of frequency change directly reflects the power system's ability to resist changes in its motion state, i.e., the inertia characteristic. The larger the inertia, the smaller the rate of frequency change under the same power deficit. Frequency deviation reflects the resistance experienced by the system when it deviates from its rated state, i.e., the damping characteristic. The stronger the damping, the smaller the steady-state frequency deviation. Therefore, the above three parameters constitute the minimum complete set describing the dynamic process of the power system's frequency response, which can fully map the nonlinear relationship between inertia and damping. Although voltage amplitude or reactive power contains some dynamic information of the power system when considering voltage-sensitive loads, from the perspective of simplifying calculations, the above three parameters have the highest sensitivity and representativeness in the dominant loop of active-frequency control, enabling high-precision parameter identification. Therefore, in some embodiments, the process of obtaining the variables reflecting the inertia characteristic at each moment of the inertia-dominated stage and the variables reflecting the damping characteristic at each moment of the damping-dominated stage may include:

[0074] A1) Obtain the frequency and active power at each moment of the inertia-dominated phase, the frequency and active power at each moment of the previous period, and the frequency and active power at each moment of the damping-dominated phase from the disturbance data of the power system; wherein, the previous period is a period of duration L, located before and adjacent to the inertia-dominated phase, and L is a preset value, which can be 5~20.

[0075] A2) Based on the frequency deviation at each moment, the rate of change of frequency between adjacent moments, and the active power at each moment, construct variables reflecting the characteristics of inertia at each moment in the inertia-dominated stage and variables reflecting the characteristics of damping at each moment in the damping-dominated stage; where the frequency deviation is the difference between the frequency and the rated frequency of the power system.

[0076] It should be noted that the variable s at time t t It can be represented as:

[0077] ;

[0078] In the formula, the variable at time t1 , , The frequency deviation at time t1 The rate of change of frequency between time t1 and time t1-1 Let t1 be the active power.

[0079] This embodiment introduces an extended time lag to address the time-varying characteristics and memory effects of power system dynamic inertia. In novel power systems containing a high proportion of power electronic devices, the power system inertia is no longer a fixed physical constant but exhibits dynamic characteristics that change with control strategies and operating states. Furthermore, there is a physical control delay between the occurrence of power disturbances and the manifestation of frequency response. Data from a single moment can only provide a static snapshot of the power system's state and cannot fully reveal the dynamic evolution of the power system. By extending the time lag, the model can capture the long-range dependence and trajectory of frequency and active power over time, thereby more accurately identifying the nonlinear characteristics of power system inertia over time. In addition, introducing historical time-period data can act as a filter, effectively smoothing random noise in the measurement data and significantly improving the robustness and accuracy of parameter identification under complex dynamic conditions.

[0080] Step 2: Input all variables into the estimation model to obtain the initial inertia and damping of the power system.

[0081] It should be noted that the estimation model here is an artificial intelligence model, such as a multilayer perceptron neural network, long short-term memory network, gated recurrent unit, convolutional neural network, support vector machine, etc.

[0082] Because of its simple structure and high computational efficiency, the multilayer perceptron (MLP) neural network has a smaller computational load compared to the complex gating mechanisms of other networks, thus meeting the high real-time requirements of power systems for parameter identification. Secondly, the MLP has a strong ability to approximate nonlinear functions, accurately fitting the complex strong nonlinear mapping relationship between inertia, damping, and system frequency response. Finally, since the preceding steps have constructed an input feature vector containing historical information by extending the time lag, the temporal dependency has actually been transformed into spatial features. At this point, the fully connected structure of the MLP can efficiently extract the correlation between features, avoiding the risk of overfitting and ensuring generalization performance with limited training samples.

[0083] Therefore, the estimation model here uses a multilayer perceptron neural network, with the structure as follows: Figure 2 As shown, the network takes a set of input variables as input, performs regression learning on these variables, and outputs estimated values ​​for inertia and damping. The formula can be expressed as:

[0084] ;

[0085] In the formula, These are the inertia estimates and damping estimates output by the multilayer perceptron neural network, respectively. For a multilayer perceptron neural network, θ represents the parameters (weights and biases) of the multilayer perceptron neural network.

[0086] The training of a multilayer perceptron neural network uses the minimum mean square error loss function and incorporates a physical consistency regularization term. The formula can be expressed as:

[0087] ;

[0088] In the formula, L(θ) is the loss value, N is the total number of training samples, and H... i and D i These are the true values ​​of inertia and damping output by the multilayer perceptron neural network for the i-th training sample, respectively. These are the inertia estimate and damping estimate output by the multilayer perceptron neural network for the i-th training sample, respectively. λ is a weighting factor, ranging from 0.1 to 0.3, used to balance the physical constraint terms. Let be the measured active power of the power system at time t. These are the inertia and damping obtained through a multilayer perceptron neural network, respectively. To adopt The active power of the power system at time t is calculated.

[0089] The loss function used in this embodiment has the significant advantage of synergistic optimization based on data-driven and physical constraints. This function consists of two parts: the first part is the mean square error of parameter prediction, which directly guides the neural network to approximate the true inertia and damping labels, ensuring the basic accuracy of the identification results; the second part is a physical consistency regularization term, which uses the dynamic equations of the power system as prior knowledge to impose physical constraints on the output of the neural network. The benefit of introducing this physical regularization term is that it forces the solution sought by the neural network not only to be numerically close to the true value, but also to be able to physically reproduce the actual power response trajectory of the power system, thereby effectively avoiding the problems of "overfitting" or outputs violating physical common sense that may occur in purely data-driven models.

[0090] Step 3: Calculate the active power of the power system using the initial inertia and damping. Based on the calculated active power and the measured active power, optimize the initial inertia and damping to obtain the optimal inertia and damping.

[0091] Step 3 can specifically be parameter correction based on the consistency of the physical model. In some embodiments, the optimization objective for the initial inertia and damping can be to minimize the active power response residual of the power system. The specific formula for the active power response residual of the power system is:

[0092] ;

[0093] In the formula, J(H,D) represents the active power response residual of the power system, n is the total number of sampling points, and is also equal to the total number of sampling times. Let be the measured active power of the power system at time t. To adopt The active power of the power system at time t is calculated.

[0094] The above function can be further optimized using a genetic algorithm within the search interval. The inner iterative solution minimizes J(H,D), yielding the optimal inertia and damping, which can be expressed as H. * and D * ;in, They are respectively The lower and upper limits, They are respectively The lower and upper limits, This represents the search range coefficient, which is usually set between 0.1 and 0.2 to define a refined search within this confidence range.

[0095] This interval uses the prediction results given by the multilayer perceptron neural network as the central benchmark, multiplied by a scaling factor. To obtain the lower limit, multiply the central benchmark by the scaling factor. By obtaining the upper limit, a local rectangular search neighborhood is formed.

[0096] The main advantage of adopting the above optimization strategy is that it achieves the complementary advantages of "data-driven initial screening" and "physical model fine calculation": First, by utilizing the fast reasoning capability of the multilayer perceptron neural network to provide a high-confidence initial solution, it avoids the problem of excessive computation time or getting trapped in local optima caused by the blind search of the global space by the genetic algorithm, and significantly improves the convergence speed; Second, by performing secondary optimization within a small range with the goal of minimizing the physical power response residual, it can effectively correct the inherent biases that the multilayer perceptron neural network may have as a model, and ensure that the parameters obtained in the end not only conform to the data rules, but also more accurately meet the physical dynamic constraints of the power system, thereby achieving higher precision parameter identification than a single model.

[0097] Step 4: Construct a power system model using the optimal inertia and damping. Under the same disturbance excitation, simulate the power system model and the actual power system respectively. If the simulation comparison results meet the preset requirements, the optimal inertia and optimal damping are the estimated results.

[0098] To verify the effectiveness of optimal inertia and damping, simulation and model consistency verification can be performed under the same disturbance excitation. Specifically, simulations are conducted under the same disturbance excitation, and the simulation results are compared.

[0099] In some embodiments, a generalized circuit equivalent mapping of the power system can be performed. Specifically, the optimal inertia and damping can be mapped to a generalized active power-frequency admittance model to establish the circuit equivalent relationship between inertia and damping. The formula can be expressed as:

[0100] ;

[0101] ;

[0102] In the formula, G D For the damping equivalent conductance, D * For optimal damping, C H H is the equivalent capacitance due to inertia. * For optimal inertia, α and β are proportionality coefficients, and S N and ω N The rated capacity and rated angular frequency of the power system.

[0103] This allows us to formulate the admittance function of the power system, i.e., the power system model, which can be expressed by the formula:

[0104] ;

[0105] In the formula, These are the generalized active power-frequency admittance and the Laplace operator, respectively.

[0106] Under the same disturbance excitation u(t), the dynamic simulation of the actual power system and the power system model are performed simultaneously, and the formula can be expressed as:

[0107] ;

[0108] ;

[0109] In the formula, For the input-output mapping of a real power system, For the prediction output function of the power system model, and These represent the active power obtained from dynamic simulations of the actual power system and the power system model, respectively.

[0110] The root mean square of the simulation result error can be further calculated:

[0111] .

[0112] When the root mean square error between the simulation results of the actual power system and the simulation results of the power system model is less than or equal to the first threshold, the optimal inertia and optimal damping are determined to be the estimated results. Otherwise, the estimation model is retrained and the process proceeds to step 2, that is, the optimal inertia and damping are obtained based on the retrained estimation model, and the simulation is performed again. The first threshold can be taken as 1% to 3% of the rated power.

[0113] This embodiment maps the identified optimal inertia and damping to a generalized active power-frequency admittance model. Its core purpose is to establish an equivalent mapping relationship between mechanical rotational dynamics and electrical network topology. Through this mapping, the differential equations originally describing rotor motion are transformed into linear circuit forms in the complex frequency domain, where the inertia characteristic is equivalent to capacitance and the damping characteristic is equivalent to conductance, thus forming a parallel equivalent circuit structure.

[0114] The main advantages of this modeling approach are: 1) It achieves decoupling and linearization of the physical mechanism, enabling rapid frequency domain stability analysis of complex nonlinear oscillation processes using mature linear circuit theories (such as impedance analysis and Bode plots); 2) The active power-frequency admittance model has excellent network scalability, making it easy to embed single-machine models into the node admittance matrix of large-scale power systems, thereby enabling quantitative analysis of the interaction between the node and the external power grid and the risk of broadband oscillations.

[0115] Furthermore, the input-output relationship of a power system can be directly described using state-space equations, or numerical integration simulations can be performed directly using time-domain differential equations. However, the active power-frequency admittance model used here more intuitively reveals the physical properties of inertia and damping, and is more in line with the impedance-based stability criterion system widely used in current new power systems. It has higher computational convenience and physical interpretability in engineering applications.

[0116] It should be noted that, regarding the verification of optimal inertia and damping effectiveness, in addition to the time-domain simulation comparison method based on the same disturbance excitation used in this embodiment, a multi-event cross-validation method can also be used. That is, data from other historical disturbance events that occurred in the power system but were not used in this identification are substituted into the estimated parameters for simulation verification to test the generalization performance of the parameters under different operating conditions; or a benchmark method comparison method can be used to compare the estimation results of this application with the parameter results obtained by traditional methods based on the Proni method, extended Kalman filtering, or analytical derivation to evaluate the consistency of the results.

[0117] This embodiment preferably employs "simulation of the same disturbance response and model consistency verification," a method with significant advantages: It constructs a closed-loop verification system of "data-driven estimation + physical model verification." By substituting the parameters output by the black-box model back into the white-box model for full-process deduction, it effectively solves the potential risks of uninterpretability or "overfitting" in artificial intelligence models, ensuring that the identified inertia and damping parameters are not only numerically optimal but also realistically reproduce the actual response trajectory of the power system at the physical dynamics level. Secondly, by setting a strict first threshold for the root mean square error (e.g., 1%~3% of the rated power), it achieves a quantitative assessment and quality assurance of parameter accuracy, ensuring that only high-precision parameters can be output and used for subsequent virtual inertia control or frequency stability analysis, thereby greatly improving the safety of power system operation and the reliability of control strategies.

[0118] The above method obtains variables reflecting inertia characteristics during the inertia-dominated stage and variables reflecting damping characteristics during the damping-dominated stage based on power system disturbance data. It uses artificial intelligence technology to obtain initial inertia and damping, and optimizes the initial inertia and damping by combining the measured active power of the power system. Based on the optimization results, it performs simulation comparison under the same disturbance excitation to determine the final inertia and damping. It is applicable to power systems with synchronous generators, power electronic interface power supplies, and high proportion of renewable energy access. It can realize online identification and dynamic estimation of power system inertia and damping characteristics under disturbance or operational fluctuation conditions, thereby supporting power system frequency stability analysis, inertia assessment, and adaptive optimization of frequency regulation control parameters.

[0119] See Figure 3 , Figure 3 This is a block diagram of an inertia and damping estimation device for a power system provided in an embodiment of this application. The embodiment is a virtual device that can be loaded and executed by a computer device, which may include the aforementioned estimation equipment. Figure 3 The apparatus may include a variable acquisition module, an estimation module, an optimization module, and a verification module, which, when used to execute the above method, can:

[0120] The variable acquisition module acquires variables reflecting inertia characteristics at each moment of the inertia-dominated stage and variables reflecting damping characteristics at each moment of the damping-dominated stage, based on the disturbance data of the power system; wherein, the inertia-dominated stage and the damping-dominated stage are two sequential stages in the disturbance stage.

[0121] The estimation module inputs all variables into the estimation model to obtain the initial inertia and damping of the power system.

[0122] The optimization module uses the initial inertia and damping to calculate the active power of the power system. Based on the calculated active power and the measured active power, it optimizes the initial inertia and damping to obtain the optimal inertia and damping.

[0123] The verification module constructs a power system model using optimal inertia and damping. Under the same disturbance excitation, it simulates both the power system model and the actual power system. If the simulation comparison results meet the preset requirements, the optimal inertia and optimal damping are the estimated results.

[0124] The aforementioned device acquires variables reflecting inertia characteristics during the inertia-dominated phase and variables reflecting damping characteristics during the damping-dominated phase based on power system disturbance data. It uses artificial intelligence technology to obtain initial inertia and damping, and optimizes the initial inertia and damping by combining the measured active power of the power system. Based on the optimization results, it performs simulation comparisons under the same disturbance excitation to determine the final inertia and damping. It is applicable to power systems with synchronous generators, power electronic interface power supplies, and high proportions of renewable energy access. It can realize online identification and dynamic estimation of power system inertia and damping characteristics under disturbance or operational fluctuation conditions, thereby supporting power system frequency stability analysis, inertia assessment, and adaptive optimization of frequency regulation control parameters.

[0125] This application also relates to a computer-readable storage medium that stores one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform a method for estimating the inertia and damping of a power system.

[0126] This application also relates to a computer device including one or more processors and one or more memories, wherein one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing a method for estimating the inertia and damping of a power system.

[0127] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0128] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0129] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0130] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0131] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A method of inertia and damping estimation for a power system, characterized by, The method comprises: From disturbance data of a power system, frequency and active power at each time of an inertia dominant stage, frequency and active power at each time of a previous time period, and frequency and active power at each time of a damping dominant stage are obtained; according to frequency deviation at each time, frequency change rate between adjacent times, and active power at each time, a variable reflecting inertia characteristics at each time of the inertia dominant stage and a variable reflecting damping characteristics at each time of the damping dominant stage are constructed; wherein the inertia dominant stage and the damping dominant stage are two stages in sequence in the disturbance stage; the disturbance data comprises frequency and active power at each time of the disturbance stage; the previous time period is one time period with a length L, located before the inertia dominant stage and adjacent, L being a preset value; the frequency deviation is a difference value of the frequency and a rated frequency of the power system; the variable s t is the variable at t1 time , , is the frequency deviation at t1 time, is the frequency change rate between t1 time and t1-1 time, is the active power at t1 time; All variables are input into the estimation model to obtain the initial inertia and damping of the power system; wherein the estimation model adopts a multi-layer perception neural network, and the loss function during training is , L(θ) is the loss value, N is the total number of training samples, H i and D i are the inertia true value and the damping true value output by the multi-layer perception neural network for the ith training sample, are the inertia estimated value and the damping estimated value output by the multi-layer perception neural network for the ith training sample, λ is a weight factor, is the measured active power of the power system at time t, are the inertia estimated value and the damping estimated value output by the multi-layer perception neural network, is the active power of the power system at time t obtained by calculation using . calculating active power of the power system by using initial inertia and damping, optimizing the initial inertia and damping according to the calculated active power and measured active power, and obtaining optimal inertia and damping; A power system model is constructed using optimal inertia and damping. Simulations are performed on both the power system model and the actual power system under the same disturbance excitation. If the simulation comparison results meet preset requirements, the optimal inertia and damping are considered the estimated results. Constructing the power system model using optimal inertia and damping includes: mapping the optimal inertia and damping to a generalized active power-frequency admittance model, establishing the circuit equivalence relationship between inertia and damping, and using this circuit equivalence relationship as the power system model. The formula is as follows: , These are the generalized active power-frequency admittance and the Laplace operator, respectively. For the damping equivalent conductance, D * For optimal damping, H is the equivalent capacitance due to inertia. * For optimal inertia, α and β are proportionality coefficients, and S N and ω N The rated capacity and rated angular frequency of the power system.

2. The method of claim 1, wherein, Before obtaining the variables, the method further comprises a step of dividing a disturbance stage into an inertia-dominant stage and a damping-dominant stage according to disturbance data of the power system, and the step comprises: determining a disturbance starting time, a power system recovery time and an active power change rate inflection point time of the disturbance stage according to a frequency first derivative, a frequency deviation first derivative and an active power change rate, regarding a stage from the disturbance starting time to the active power change rate inflection point time as the inertia-dominant stage, and regarding a stage from the active power change rate inflection point time to the power system recovery time as the damping-dominant stage; wherein the frequency deviation is a difference between the frequency and a rated frequency of the power system.

3. The method of claim 1, wherein, The optimization target of the initial inertia and damping is to minimize a power system active power response residual, The power system active power response residual formula is: ； In the formula, J(H, D) is the power system active power response residual, and n is a total number of sampling points.

4. The method of claim 1, wherein, The preset requirement is that a root mean square error between a simulation result of an actual power system and a simulation result of a power system model is less than or equal to a first threshold value.

5. The method of claim 1, wherein, The method further comprises retraining the estimation model, obtaining the optimal inertia and damping based on the retrained estimation model, and re-performing the simulation if the simulation comparison result does not satisfy the preset requirement.

6. A device for estimating the inertia and damping of a power system, characterized in that, The method comprises: The variable acquisition module acquires the frequency and active power at each time point in the inertia dominant stage, the frequency and active power at each time point in the previous time period, and the frequency and active power at each time point in the damping dominant stage from disturbance data of the power system; constructs variables reflecting inertia characteristics at each time point in the inertia dominant stage and variables reflecting damping characteristics at each time point in the damping dominant stage according to the frequency deviation at each time point, the frequency change rate between adjacent time points, and the active power at each time point; wherein the inertia dominant stage and the damping dominant stage are two stages in sequence in the disturbance stage; the disturbance data includes the frequency and active power at each time point in the disturbance stage; the previous time period is one time period with a length L located before the inertia dominant stage and adjacent to the inertia dominant stage, and L is a preset value; the frequency deviation is the difference between the frequency and the rated frequency of the power system; the variable s t is the variable at t1 time point , , is the frequency deviation at t1 time point, is the frequency change rate between t1 time point and t1-1 time point, is the active power at t1 time point; An estimation module inputs all variables into an estimation model to obtain initial inertia and damping of the power system; wherein the estimation model adopts a multi-layer perception neural network, and a loss function during training is , L(θ) is a loss value, N is a total number of training samples, H i and D i are real inertia and damping values output by the multi-layer perception neural network for the i-th training sample, are estimated inertia and damping values output by the multi-layer perception neural network for the i-th training sample, λ is a weight factor, is a measured active power of the power system at the t-th moment, are estimated inertia and damping values output by the multi-layer perception neural network, is the active power of the power system at the t-th moment obtained by calculation using . an optimization module configured to calculate active power of the power system by using initial inertia and damping, optimize the initial inertia and damping according to the calculated active power and measured active power, and obtain optimal inertia and damping. The verification module adopts optimal inertia and damping to construct a power system model, simulates the power system model and the actual power system respectively under the same disturbance excitation, and if the simulation comparison result meets the preset requirement, the optimal inertia and the optimal damping are the estimation results; wherein, the optimal inertia and damping are adopted to construct the power system model, including: mapping the optimal inertia and damping to a generalized active power-frequency admittance model, establishing a circuit equivalent relationship of inertia and damping, and taking the circuit equivalent relationship as the power system model; the formula is , The generalized active power-frequency admittance and Laplace operator are respectively, The damping equivalent conductance is D * The optimal damping is D The inertia equivalent capacitance is H * The optimal inertia is H N The proportional coefficients are α and β, and S N The rated capacity and rated angular frequency of the power system are S and ω 7. A computer-readable storage medium, characterized in that, The computer readable storage medium stores one or more programs, and the one or more programs comprise instructions which, when executed by the computing device, cause the computing device to perform the method of any one of claims 1-5.

8. A computer device, comprising: The method comprises: one or more processors and one or more memories, one or more programs stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs comprising instructions for performing the method of any one of claims 1-5.

Images