Inertia active support demand evaluation method for new energy power system

Abstract

The invention relates to the field of power system simulation, in particular to an inertia active support demand evaluation method for a new energy power system, which comprises the following steps: S1, constructing a power grid model; s2, calculating a system impedance matrix and a short-circuit ratio; s3, constructing an observability Gramer matrix; s4, guiding an analytical solution of the Lyapunov equation; and S5, evaluating the inertia active support demand considering the space difference. The method breaks through the space-time decoupling limitation of a traditional method, forms a multi-scale analysis system covering electromagnetic transient-electromechanical transient-medium and long term dynamic, is especially suitable for a complex power grid scene containing high-proportion new energy access, and provides full-process quantitative decision support for system planning and operation control.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power system simulation, in particular to a kind of inertia active support demand evaluation method of new energy power system. BACKGROUND

[0002] With the large-scale access of new energy, the inertia characteristics of power system have changed significantly. The inertia of traditional synchronous generators is gradually replaced by the virtual inertia of new energy stations, which weakens the system's ability to resist frequency changes and makes frequency stability problems increasingly prominent. Therefore, accurately evaluating the inertia active support demand of new energy power system is of great significance to ensure system frequency safety. However, existing technologies still have many limitations in inertia active support demand evaluation. For example, the parameter observability problem leads to errors in some line parameters, affecting the accuracy of the evaluation results; the lack of dynamic adaptability makes it difficult to accurately simulate the harmonic coupling effect caused by new energy units connected through power electronic interfaces, resulting in deviations between the model and the actual system; the data validity problem makes it difficult to fully consider the voltage-dependent characteristics of loads, resulting in errors in equivalent parameter calculation; and the time-space decoupling limitation makes it difficult for traditional methods to consider electromagnetic transient, electromechanical transient and long-term dynamic processes simultaneously, resulting in incomplete and inaccurate evaluation results.

[0003] To address these limitations, the present application proposes an inertia active support demand evaluation method suitable for complex power grid scenarios with high proportion of new energy access. SUMMARY

[0004] To address the above problems, a new energy power system inertia active support demand evaluation method is provided, which divides the power grid into multiple control areas, quantifies the inertia level of each area by constructing an inertia distribution matrix, and constructs a comprehensive evaluation index.

[0005] To solve the problems of existing technologies, the present application provides a new energy power system inertia active support demand evaluation method, comprising the following steps: S1: Construct a power grid model, integrate synchronous machine detailed model and new energy unit equivalent model, and realize the time-space characteristic representation of AC-DC hybrid system through hierarchical modeling strategy; S2: Calculate system impedance matrix and short circuit ratio, adopt white box analysis and black box data driven fusion method, combine load dynamic characteristic compensation technology, realize accurate calculation of short circuit ratio index; S3: Construct observability gram matrix, design norm solving algorithm based on state space model, recursively calculate energy-frequency distribution characteristics through discrete state space model, and quantify the suppression ability of system to random power fluctuation; S4: Guiding the analytical solution of Lyapunov equation, establishing the explicit expression of stability criterion, evaluating system stability; S5: Evaluate the inertia active support demand considering spatial differences, build a spatial inertia distribution matrix, integrate the electrical distance weighting coefficient and dynamic correlation factor, quantify the inertia level of each region, and realize the accurate evaluation of regionalized inertia support demand.

[0006] Preferably, step S1 comprises: 1) Data acquisition and preprocessing: In the data acquisition stage, high-precision synchronized phasor measurement units (PMU) are deployed to obtain real-time voltage and current phasor data of each node in the power grid, and environmental parameters such as wind speed and light intensity from the weather station are integrated to form a multi-source heterogeneous data set. In the data preprocessing stage, the original measurement values are filtered to eliminate noise interference, and the synchronization of different source data is ensured through time alignment algorithm; 2) Hierarchical mixed model construction: A hierarchical modeling strategy is used to construct detailed synchronous generator models, power transmission and distribution network models, and equivalent models of new energy units; 3) Mixed parameter identification strategy: In the first stage, historical operation data are used to calibrate the static parameters such as generator inertia time constant and line impedance through genetic algorithm; in the second stage, an online identification algorithm is deployed to update the load model parameters through extended Kalman filter based on real-time measurement data; 4) Model verification and closed-loop test: The test environment is built, the constructed power grid model is deployed on the simulation platform through OPC protocol and real-time digital simulator (RTDS); precision evaluation and reconstruction, comparison of the dynamic response trajectories of the actual system and the constructed power grid model, and evaluation of the precision of the power grid model using correlation coefficient method, when the correlation coefficient is less than 0.95, the model reconstruction process is triggered.

[0007] Preferably, the detailed synchronous generator model is a five-order detailed model considering the rotor dynamic process of the synchronous generator.

[0008] Preferably, when constructing the power transmission and distribution network model, the π-type equivalent circuit is used to describe the distributed parameter characteristics of AC transmission lines, and the average value model based on voltage source converter (VSC) is used to reflect the fast power control capability of DC lines.

[0009] Preferably, when constructing the equivalent model of new energy units, the equivalent wind speed-power curve is used for wind farms combined with the wake effect correction model, and the dynamic output model considering cloud shading is used for photovoltaic power stations.

[0010] Preferably, step S2 comprises: The white box analysis method branch first constructs a node admittance matrix based on prior knowledge, and then introduces sensitivity analysis to optimize the node admittance matrix; The black-box data-driven branch firstly measures and acquires real-time data under various working conditions, and then constructs a linear equation parameter matrix; Short-circuit ratio calculation and precision verification: impedance matrix and short-circuit ratio are solved based on the results of the white-box analytical branch and the black-box data-driven branch.

[0011] Preferably, in the white-box analytical branch, the real-time data acquired in the black-box data-driven branch are introduced to perform parameter correction on the node admittance matrix; In the black-box data-driven branch, the load priori knowledge in the white-box analytical branch is introduced, and the real-time data acquired under various working conditions are combined to construct a voltage-power characteristic surface constraint parameter search space.

[0012] Preferably, in the step S3, the observability of the system state is judged by calculating the singular value decomposition result of the Gram matrix; when the minimum singular value is greater than a preset threshold, it is determined that the system satisfies the observability condition, otherwise the sensor configuration needs to be adjusted or the measurement dimension needs to be increased.

[0013] Preferably, in the step S3, the H2 norm solution needs to substitute the eigenvalue of the Gram matrix into the integral expression, and the energy gain of the system to the random input is calculated by the numerical integral method.

[0014] Preferably, in the step S4, firstly, based on a specific scenario, the specific inertia problem and boundary are determined; then, P(m) is defined as the solution of the Lyapunov equation, and the cost function is converted to represent as a function f(m) of the inertia vector and an inertia allocation vector mi; finally, the explicit gradient Δf(m) of f(m) is calculated.

[0015] Preferably, the step S5 includes: Space inertia modeling: the power grid is divided into multiple control areas, each area is configured with independent inertia resources, and the inertia level of each area is quantified by constructing an inertia distribution matrix; Demand assessment: a comprehensive assessment index is constructed, which integrates frequency deviation, inertia reserve and spatial distribution characteristics, and quantifies the inertia demand of the system by weighted summation.

[0016] A data processing device, comprising: a memory for storing a computer program; a processor for executing the computer program to implement the steps of the inertia active support demand assessment method of a new energy power system.

[0017] A computer readable storage medium, the computer readable storage medium stores a computer program, the computer program is executed by a processor to implement the steps of the inertia active support demand assessment method of a new energy power system.

[0018] The beneficial effects of the present application compared to the prior art are: breaking through the time-space decoupling limit of traditional methods, forming a multi-scale analysis system covering electromagnetic transient-mechanical transient-medium and long-term dynamics, especially suitable for complex power grid scenarios with high proportion of new energy access, providing full-process quantitative decision support for system planning and operation control. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a flowchart of a new energy power system inertia active support demand evaluation method of the present application.

[0020] Figure 2 is a flowchart of step S2 in a new energy power system inertia active support demand evaluation method of the present application.

[0021] Figure 3 is a flowchart of the explicit gradient in step S4 in a new energy power system inertia active support demand evaluation method of the present application. DETAILED DESCRIPTION

[0022] In order to further understand the features, technical means and specific purposes and functions achieved by the present application, the present application will be described in further detail below with reference to the drawings and specific embodiments.

[0023] The present application proposes a new energy power system inertia active support demand evaluation method. This method is particularly suitable for complex power grid scenarios with high proportion of new energy access, such as large-scale wind power bases, photovoltaic power station clusters, etc. The volatility and uncertainty of new energy in these scenarios pose a serious threat to system frequency safety. Through the technical solution of the present application, accurate evaluation of the inertia active support demand of the new energy power system can be realized, providing full-process quantitative decision support for system planning and operation control, effectively improving the frequency stability and anti-interference ability of the system. At the same time, the technical solution proposes a spatial inertia active support demand evaluation method, which can be applied to the analysis of provincial interconnection and other cross-regional scenarios considering regional channel constraints, providing strong protection for the safe and stable operation of the power system. As shown in the figure, it specifically includes the following steps: Figure 1 S1: Construct a power grid model, integrate synchronous machine detailed model and new energy unit equivalent model, realize time-space characteristic representation of AC-DC hybrid system through hierarchical modeling strategy; S2: Calculate the system impedance matrix and short circuit ratio, adopt white box analysis and black box data driven fusion method, combine with load dynamic characteristic compensation technology, realize accurate calculation of short circuit ratio index; S3: Construct an observability Gram matrix, design norm solving algorithm based on state space model, recursively calculate energy-frequency distribution characteristics through discretized state space model, quantify the suppression ability of the system to random power fluctuations;​ S4: Guiding the analytical solution of Lyapunov equation, establishing the explicit expression of stability criterion, evaluating the system stability; S5: Evaluating the inertia active support demand considering spatial difference, constructing the spatial inertia distribution matrix, integrating the electrical distance weighting coefficient and dynamic correlation factor, quantifying the inertia level of each region, and realizing the accurate evaluation of regionalized inertia support demand.

[0024] As a specific embodiment of the inertia active support demand evaluation method of the new energy power system of the application, the step S1 of constructing the power grid model comprises the following steps:

[0025] 1) Data acquisition and preprocessing: In the data acquisition stage, high-precision phasor measurement units (PMU) are deployed to obtain real-time voltage and current phasor data of each node (generator outlet, substation) of the power grid. At the same time, environmental parameters such as wind speed and light intensity of the weather station are integrated, specifically, a weather station is deployed in the new energy station (wind power / photovoltaic) to collect wind speed, light intensity, temperature and other data (or directly obtained from the weather station deployed at the site), and the data is transmitted to the main station after preprocessing by the edge computing node to form a multi-source heterogeneous data set. Specifically, based on IEC 61850 protocol (or other protocols), PMU measurement data and weather data are integrated to form a time series aligned multi-source heterogeneous data set (including electrical and non-electrical quantities).

[0026] In the data preprocessing stage (data cleaning and synchronization), the original measurement values are filtered to eliminate noise interference, and the synchronization of different sources of data is ensured by a time alignment algorithm. More specifically, wavelet transform is used to remove high-frequency noise from the original measurement data, and the fundamental and low-frequency components are retained for noise filtering; the time stamps of each data source are synchronized by GPS clock, with an error of ≤1 μs, to ensure the spatio-temporal consistency of cross-domain data (such as electrical and meteorological quantities) to achieve time alignment, and finally the 3σ principle is used to identify and correct outlier data points, and missing data is completed by KNN interpolation (KNN interpolation (K-Nearest Neighbors Imputation) is a similarity-based missing value filling method, which uses the feature values of the K nearest neighbors of the missing sample to estimate the missing values) for anomaly value processing.

[0027] 2) Hierarchical mixed model construction: a hierarchical modeling strategy is used to construct a synchronous generator detailed model, a power transmission and distribution network model, and an equivalent model of a new energy unit.

[0028] 2.1) The synchronous generator detailed model is a five-order detailed model considering the rotor dynamic process of the synchronous generator.

[0029] The fifth-order detailed model specifically includes: rotor motion equation: describes the balance between mechanical power and electromagnetic power (including damping winding effect). Excitation system dynamic characteristics: IEEE Type 1 exciter model is adopted, including automatic voltage regulator (AVR) link. Governor control logic: simulates the frequency-power response characteristics of steam turbine / water turbine (including primary frequency modulation function).

[0030] 2.2) Build a power transmission and distribution network model. For AC transmission lines, the distribution parameter characteristics are described by a π-type equivalent circuit. For DC lines, the average value model based on voltage source converter (VSC) is used to reflect its fast power control capability.

[0031] 2.3) Build an equivalent model of new energy units. For wind farms, an equivalent wind speed-power curve is used in combination with a wake effect correction model. For photovoltaic power stations, a dynamic output model is established considering cloud shading.

[0032] For wind farms: an aggregated equivalent model is built, which equates a single wind turbine (doubly-fed / direct-driven) to a controlled voltage source, combined with a wake effect correction coefficient (such as Jensen model) to describe the dynamic characteristics. The wind speed-power curve is mapped to the output fluctuation under turbulent wind speed through dynamic characteristics.

[0033] For photovoltaic power stations: a dynamic model considering cloud shading is introduced, which combines the maximum power point tracking (MPPT) algorithm to simulate the change of direct current side power with light intensity attenuation factor. Inverter model: adopts current control type grid-connected strategy to reflect low voltage ride through (LVRT) capability.

[0034] 3) Hybrid parameter identification strategy: First stage and second stage of hybrid parameter identification strategy.

[0035] In the first stage, historical operation data is used to calibrate the static parameters such as generator inertia time constant and line impedance offline through genetic algorithm.

[0036] Genetic algorithm optimization: Objective function: minimize the mean square error (MSE) of simulation output (active / reactive power) and historical operation data.

[0037] Parameter range: generator inertia time constant (H ∈ [2, 8] s), line positive sequence impedance (Z ∈ [0.1, 0.5] pu).

[0038] Termination condition: iteration number ≥ 200 or MSE ≤ 1e-4.

[0039] In the second stage, an online identification algorithm is deployed based on real-time measurement data to update the load model parameters rolling using extended Kalman filter.

[0040] Extended Kalman Filter (EKF): State-space model: Treat load power as a stochastic process and establish an ARMAX model.

[0041] Observation matrix update: Real-time measurement of node voltage / power, recursively correct load model parameters (such as ZIP coefficients).

[0042] Noise covariance adjustment: Dynamically adjust Q / R matrix according to measurement noise level to avoid filter divergence.

[0043] 4) Model validation and closed-loop testing:

[0044] Model validation and closed-loop testing include test environment setup and precision evaluation and reconstruction. Test environment setup: Connect with Real-Time Digital Simulator (RTDS) through OPC protocol and deploy the constructed power grid model on the simulation platform. Precision evaluation and reconstruction: Compare the dynamic response trajectories of the actual system and the constructed power grid model, and evaluate the precision of the power grid model using the correlation coefficient method. When the correlation coefficient is less than 0.95, trigger the model reconstruction process.

[0045] As shown in Figure 2 as a specific embodiment of the inertia active support demand evaluation method of the new energy power system of the present application, the step S2 comprises:

[0046] 1) White-box analysis method branch: First, construct the node admittance matrix based on prior knowledge, and then introduce sensitivity analysis to optimize the node admittance matrix.

[0047] 2) Black-box data-driven method branch: First, measure and obtain real-time data under various working conditions, and then construct a linear equation parameter matrix.

[0048] 3) Short-circuit ratio calculation and precision verification: Solve the impedance matrix and short-circuit ratio based on the results of the white-box analysis method branch and the black-box data-driven method branch.

[0049] Further, in the white-box analysis method branch, the real-time data obtained in the black-box data-driven method branch are used to correct the parameters of the node admittance matrix; in the black-box data-driven method branch, the load prior knowledge in the white-box analysis method branch is introduced and combined with the real-time data under various working conditions to construct a voltage-power characteristic surface constraint parameter search space.

[0050] More specifically, in the field of power system analysis and control, short-circuit ratio (SCR) is a core indicator for assessing the strength and stability of the power grid, and its calculation accuracy directly depends on the accuracy of the equivalent admittance matrix of the DC feed-in node. The acquisition of this matrix is essentially an impedance identification problem of the power system, and the solutions can be summarized into two technical paths: white-box modeling analysis method based on accurate network parameters and topology structure, and black-box data-driven method relying on electromechanical transient simulation data. These two methods have significant differences in theoretical system, implementation conditions and applicable scenarios, and the analysis of their technical evolution paths and limitations has important guiding significance for engineering practice.

[0051] The white-box modeling analysis method takes the complete knowability of system internal parameters as a prerequisite and derives equivalent parameters through network equation construction. The core of this method is to establish a mathematical model that describes the electrical characteristics of the system, and the typical representative is the identification of Thevenin equivalent parameters based on the node admittance matrix. This technical route first constructs a node admittance matrix containing generators, transformers, transmission lines, etc., eliminates the influence of non-observed nodes through matrix correction techniques, and finally solves the Thevenin equivalent potential and equivalent impedance through matrix inversion operation. Its improved algorithm modifies the process by introducing a sensitivity analysis matrix, which can control the calculation error within 2% in the static load flow scenario. However, this method is extremely sensitive to the time-varying nature of system parameters, and when the system operating conditions change suddenly, the parameter identification error can exceed 15%. The multi-port equivalent function of BPA-SCCP software improves the calculation efficiency through modular modeling, but its essence is still based on static equivalent of fixed topology, which cannot adapt to the dynamic power grid environment with large-scale new energy access.

[0052] The fundamental limitations of the white-box method come from two dimensions: one is the parameter observability problem, and the measurement error of some line parameters in the actual power grid is 5%-10% due to the limitation of measurement means, and the state quantity such as transformer tap position also has the risk of information loss; the second is the lack of dynamic adaptability, and the harmonic coupling effect caused by the access of new energy units through power electronic interfaces makes the traditional fundamental frequency equivalent model have a 10%-20% amplitude deviation. The measured data of a provincial power grid shows that when the wind power penetration exceeds 25%, the deviation between the short-circuit ratio calculation result based on the white-box model and the measured value of the wide-area measurement system (WAMS) is 0.3pu, directly affecting the system stability assessment.

[0053] Black-box data-driven methods take a different approach by generating port measurements from electro-mechanical transient simulations and then inverting equivalent parameters through feature extraction. This approach does not rely on system internal details and only focuses on input-output response characteristics. Typical techniques include the two-point method and time-domain simulation combined with compensation method: the two-point method uses the difference between the measured data under two different operating conditions to directly solve the equivalent impedance through a linear equation system. In a simple radial power grid with load composition, the calculation time is less than 0.1 second; the time-domain simulation compensation method calculates the open-circuit voltage and short-circuit current by constructing a global network equation. In the domestic power grid measurement, the parameter identification period can be shortened to 15 minutes.

[0054] However, the black-box method faces two technical bottlenecks: first, the data effectiveness problem, the voltage-dependent characteristics of the load cause the zero drift phenomenon of the Thevenin parameters. The measurement of a certain urban power grid shows that when the load power factor changes from 0.85 to 0.95, the equivalent reactance calculation value offset reaches 8%; second, the non-linear load modeling dilemma, existing methods mostly simplify the load as a constant impedance model, ignoring the fact that the proportion of induction motor load exceeds 60%.

[0055] The influence mechanism of the load model on the equivalent parameters is worth further exploring. The time-varying nature of the load impedance is essentially due to the strong non-linear relationship between its equivalent admittance and the node voltage. When the ZIP model is used to describe the load characteristics, the calculation deviation of the equivalent reactance is negatively correlated with the proportion of constant impedance load. In scenarios where the proportion of constant power load exceeds 40%, the calculation error of the traditional two-point method may exceed 20%. The measurement data of a certain regional power grid shows that after considering the dynamic characteristics of the load, the correction amount of the equivalent admittance matrix reaches 18% of the original value, and the correction amount of the short-circuit ratio evaluation value reaches 0.15 pu, directly affecting the determination of new energy access capacity.

[0056] To overcome the limitations of existing methods, the present invention proposes a hybrid modeling approach: embedding a data-driven module in the white-box framework to correct model parameters through real-time measurement data; introducing load prior knowledge into the black-box method to construct a voltage-power characteristic surface to constrain the parameter search space. After adopting the hybrid method in a provincial power grid pilot project, the short-circuit ratio calculation accuracy is improved by 40%, and the calculation time is controlled within 5 minutes, showing potential for engineering application value.

[0057] As a specific embodiment of the present invention, in step S3, the observability of the system state is determined by calculating the singular value decomposition result of the Gram matrix; when the minimum singular value is greater than a preset threshold, it is determined that the system satisfies the observability condition, otherwise the sensor configuration needs to be adjusted or the measurement dimension needs to be increased. The H2 norm solution needs to substitute the eigenvalue of the Gram matrix into the integral expression and calculate the energy gain of the system to the random input through numerical integration method.

[0058] More specifically, constructing an observable Gram matrix requires observability analysis. This matrix consists of the output matrix and its multiple shifts. By calculating its singular value decomposition results, the observability of the system state is determined. When the minimum singular value is greater than a preset threshold, the system is considered to meet the observability condition; otherwise, the sensor configuration needs to be adjusted or the measurement dimensions increased. The construction of the Gram matrix is ​​based on the system state-space model. A discretization method is used to convert the continuous-time equation into a difference equation form, and the observable Gram matrix is ​​calculated using a recursive algorithm. The diagonal elements of this matrix reflect the energy distribution of each state variable, while the off-diagonal elements characterize the coupling properties between states. Solving for the H2 norm requires substituting the eigenvalues ​​of the Gram matrix into the integral expression and calculating the system's energy gain to random inputs using numerical integration methods. To improve computational efficiency, a fast Fourier transform is used to convert convolution operations into frequency domain multiplication, significantly reducing computational complexity. The core process is as follows:

[0059] To simulate the spatial distribution of perturbations in the grid, we parameterize the input.

[0060] In the formula, For input, For the perturbation voltage matrix, The input-output mapping matrix is ​​established as follows: (This represents a Dirac pulse.)

[0061] ; In the formula and These are the diagonal matrices of the inertia matrix and the damping / sag coefficient, respectively. This is the network Laplace matrix (or susceptibility matrix).

[0062] Typically, linear systems Norms can be efficiently computed by solving linear Lyapunov equations. Additional linear constraints are needed to account for marginal stability.

[0063] In the formula The observable Gram matrix is ​​uniquely determined by the following Lyapunov equations and additional constraints: ; ; .

[0064] As a specific embodiment of the inertia active support demand evaluation method of the new energy power system of the present application, in step S4, first, based on a specific scenario, the specific inertia problem and the boundary are determined; then, P(m) is defined as the solution of the Lyapunov equation, and the cost function is converted to represent the function f(m) of the inertia vector and the inertia allocation vector m; and finally, the explicit gradient Δf(m) of f(m) is calculated.

[0065] The algorithm for calculating the explicit gradient Δf(m) of f(m) is recommended as follows:

[0066] 1) Algorithm core: Gradient descent method or its variants

[0067] As mentioned earlier, "an efficient algorithm is used to calculate the explicit gradient Δf(m) of f(m)", and it is emphasized that it can "eliminate large-scale nonlinear decision variables and include the equation in the gradient information". This is in line with the typical characteristics of the gradient descent method (Gradient Descent) or its improved variants (such as the conjugate gradient method and the quasi-Newton method).

[0068] Gradient calculation: Through mathematical derivation or automatic differentiation techniques, the gradient Δf(m) of the objective function f(m) with respect to the decision variable m is calculated, that is, This step converts the solution of the nonlinear equation set into a gradient optimization problem, significantly reducing the computational complexity.

[0069] Iterative optimization: Use gradient information to iteratively update decision variables m to gradually approach the optimal solution. For example, in the standard gradient descent method, the update rule is: k+1 = mk-α·Δf(m k ); where α is the learning rate, controlling the iteration step size.

[0070] 2) Potential algorithm implementation

[0071] Explicit gradient derivation: As mentioned earlier, "the gradient is derived through algebraic operations layer by layer in the calculation process". Through symbolic calculation (such as MATLAB's Symbolic Math Toolbox) or manual mathematical derivation, the analytical expression of Δf(m) is directly obtained. This approach avoids the errors of numerical differentiation and improves the calculation accuracy.

[0072] Combine the characteristics of the Lyapunov equation: In the derivation process, the symmetry or sparsity of the Lyapunov equation may be used to further optimize gradient calculation. For example, if the equation matrix is symmetric positive definite, the conjugate gradient method can be used to accelerate convergence.

[0073] Fusion with gradient information: The gradient Δf(m) is used to "eliminate large-scale nonlinear decision variables" by linearization or approximation methods, transforming nonlinear problems into linear gradient descent problems, thus simplifying the solution.

[0074] 3) Advantages of the efficient algorithm

[0075] Computational efficiency improvement: Gradient algorithm transforms the solution of nonlinear equation set into gradient optimization problem, avoiding the complex calculation of directly solving high-order nonlinear equation set, significantly improving the speed.

[0076] Engineering applicability: In new energy power system, inertia evaluation needs to process large-scale data in real time, and the iterative characteristics of gradient algorithm make it more suitable for dynamic environment and meet the real-time requirements.

[0077] Generally, most computational methods can be significantly accelerated when there is an explicit gradient. Considering the gradient Δf(m) of f(m), an important advantage is that large-scale nonlinear (with respect to decision variables) Lyapunov equation sets can be eliminated and included in the gradient information. The calculation process is shown in Figure 3 , and the specific process is as follows:

[0078] This calculation process is an iterative algorithm for solving certain optimization or estimation problems, related to signal processing, system identification, etc. The following is a complete description of the process:

[0079] 1) Initialize matrix A (0) : ; where I is the identity matrix, M, L and D are predefined matrices. A (0) is a block matrix used for subsequent iterative calculations.

[0080] 2) Initialize matrix B(0): ; where V is a predefined matrix, and V 1 / 2 represents its square root matrix.

[0081] 3) Calculate P (i) : ; where Lyap represents the solution function of Lyapunov equation. P (0) is the initial covariance matrix or other related matrix, A (i) is the matrix of the current iteration.

[0082] 4) Update the iteration counter: i <- 1, start the iteration loop.

[0083] 5) Calculate Φ: Φ <- e i e i T ; where e iis the i-th standard basis vector, and Φ is a diagonal matrix that selects specific dimensions.

[0084] 6) Update matrix A (i) : ; where A (i) is updated according to Φ and a predefined matrix.

[0085] 7) Update matrix B (i) : ; where B (i) is updated according to Φ and a predefined matrix.

[0086] 8) Recalculate P (i) : ; this step is similar to step 3), but using the updated A (i) .

[0087] 9) Calculate g i : ; where Trace denotes the trace of a matrix, and g i is a scalar that evaluates the performance or error of the current iteration.

[0088] 10) Calculate H 2i : ; where ∥⋅∥2 represents the L2 norm, and H 2i is the square of the L2 norm of g i .

[0089] 11) Update the iteration counter: i <- i + 1.

[0090] 12) Check the iteration termination condition: if i = n, the iteration ends and the flow terminates. Otherwise, return to step 5 to continue the iteration.

[0091] Summary: This flowchart iteratively updates matrices A and B, and uses the Lyapunov equation to solve matrix P, finally calculates scalar g i and H 2i , until the iteration termination condition is met. The entire process is used for system stability analysis, parameter estimation or optimization problems.

[0092] As a specific embodiment of the inertia active support demand evaluation method of the new energy power system of the present application, the step S5 comprises:

[0093] Spatial inertia modeling, dividing the power grid into multiple control areas, configuring independent inertia resources in each area, and quantifying the inertia level of each area by constructing an inertia distribution matrix.

[0094] Demand assessment, construct comprehensive evaluation index, which integrates frequency deviation, inertia reserve and spatial distribution characteristics, and quantifies the inertia demand of the system by weighted summation.

[0095] More specifically, the technical scheme of the spatial inertia active support demand assessment method contains two core modules: spatial inertia modeling and demand assessment. Spatial inertia modeling needs to divide the power grid into multiple control areas, each area is configured with independent inertia resources, and the inertia level of each area is quantified by constructing an inertia distribution matrix. The elements of the matrix reflect the inertia density distribution in unit area, considering the contribution of different resources such as generators and energy storage devices. Demand assessment needs to construct a comprehensive evaluation index, which integrates frequency deviation, inertia reserve and spatial distribution characteristics, and quantifies the inertia demand of the system by weighted summation. The core process is as follows:

[0096] Assuming that each node has a non-zero inertia coefficient, and considering the upper bound constraint at each node, the minimum demand problem of the system inertia after considering the spatial constraint is transformed into solving the minimum value of the norm through the inertia density transfer matrix. This problem can be summarized in the form of mathematical expression as follows:

[0097] ;

[0098] In the formula, , and have been given in step three. Since the Lyapunov constraint has a bilinear characteristic, it involves the product of and , and the decision variable also appears in the expression of . Therefore, this problem is usually non-convex, and the scale is also large.

[0099] The present application simplifies some parameters in a scientific and reasonable way through general lower and upper bounds. The problem is transformed into the following formula: ; In the formula, is the minimum value of , is the maximum value of , is the minimum value of , is the maximum value of . The present application takes the network topology described by the Laplace matrix as a constant factor, and decouples it from the decision variable . In addition, when is equal to (i.e. short-range error penalty of angle difference), the offset term is only related to the grid size. ​

[0100] The above formula reveals the essence of the problem, after comprehensively considering the mutual relationship between the disturbance intensity , the damping coefficient , the frequency penalty weight and the decision variable , finally the inertia active support demand index considering spatial difference is obtained.

[0101] A data processing device comprises:

[0102] a memory for storing a computer program;

[0103] a processor for implementing the steps of the inertia active support demand evaluation method of a new energy power system when the computer program is executed.

[0104] A computer readable storage medium, the computer readable storage medium has a computer program stored thereon, the computer program is executed by a processor to implement the steps of the inertia active support demand evaluation method of a new energy power system.

[0105] The above embodiments only express one or several embodiments of the present application, and the description is more specific and detailed, but it cannot be understood as a limitation on the scope of the present application. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of protection of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. An inertia active support demand evaluation method for a new energy power system, characterized by, The method comprises the following steps: S1: constructing a power grid model, integrating a synchronous generator detailed model and an equivalent model of new energy units, and realizing time-space characteristic representation of an AC-DC hybrid system through a hierarchical modeling strategy; S2: calculating a system impedance matrix and a short-circuit ratio, adopting a white-box analysis and black-box data-driven fusion method, combining a load dynamic characteristic compensation technology, and realizing accurate calculation of a short-circuit ratio index; S3: constructing an observability Gram matrix, designing a norm solving algorithm based on a state space model, recursively calculating energy-frequency distribution characteristics through a discretized state space model, and quantifying the suppression capability of the system to random power fluctuations; S4: guiding the analytical solution of Lyapunov equation, establishing an explicit expression of stability criterion, and evaluating system stability; S5: evaluating inertia active support demand considering spatial differences, constructing a spatial inertia distribution matrix, integrating electrical distance weighting coefficients and dynamic correlation factors, quantifying inertia levels of each region, and realizing accurate evaluation of regionalized inertia support demand.

2. The inertia active support demand evaluation method of a new energy power system according to claim 1, characterized in that, Step S1 comprises: 1) data acquisition and preprocessing: in the data acquisition stage, high-precision phasor measurement units (PMUs) are deployed to obtain real-time voltage and current phasor data of each node in the power grid, and environmental parameters such as wind speed and light intensity from weather stations are integrated to form a multi-source heterogeneous data set; in the data preprocessing stage, the original measurement values are filtered to eliminate noise interference, and the synchronization of different source data is ensured through a time alignment algorithm; 2) hierarchical hybrid model construction: a hierarchical modeling strategy is used to construct a synchronous generator detailed model, a power transmission and distribution network model, and an equivalent model of new energy units; 3) hybrid parameter identification strategy: in the first stage, historical operation data are used to calibrate static parameters such as generator inertia time constant and line impedance offline through a genetic algorithm; in the second stage, an online identification algorithm is deployed to update the load model parameters based on real-time measurement data using an extended Kalman filter; 4) model verification and closed-loop test: a test environment is built, the constructed power grid model is deployed on a simulation platform by connecting the real-time digital simulator (RTDS) through the OPC protocol, the accuracy of the power grid model is evaluated by comparing the dynamic response trajectories of the actual system and the constructed power grid model, and the correlation coefficient method is used to evaluate the accuracy of the power grid model, and the model reconstruction process is triggered when the correlation coefficient is less than 0.

95.

3. The inertia active support demand evaluation method of a new energy power system according to claim 2, characterized in that, The synchronous generator detailed model is a five-order detailed model considering the rotor dynamic process of the synchronous generator.

4. The inertia active support demand evaluation method of a new energy power system according to claim 2, characterized in that, When constructing the power transmission and distribution network model, the π-type equivalent circuit is used to describe the distributed parameter characteristics of the AC transmission line, and the average value model based on the voltage source converter (VSC) is used to reflect the fast power control capability of the DC line.

5. The inertia active support demand evaluation method of a new energy power system according to claim 2, characterized in that, When constructing the equivalent model of new energy units, the equivalent wind speed-power curve is used for wind farms in combination with the wake effect correction model, and the dynamic output model considering cloud shading is used for photovoltaic power stations.

6. The inertia active support demand evaluation method of a new energy power system according to claim 1, characterized in that, The step S2 comprises: The white-box analysis branch first constructs a node admittance matrix based on prior knowledge, and then optimizes the node admittance matrix through sensitivity analysis; The black-box data-driven branch firstly measures and acquires real-time data under various working conditions, and then constructs a linear equation parameter matrix; Short-circuit ratio calculation and precision verification: impedance matrix and short-circuit ratio are solved based on the results of the white-box analytical branch and the black-box data-driven branch.

7. The inertia active support demand evaluation method of a new energy power system according to claim 1, characterized in that, In the white-box analytical branch, the real-time data acquired in the black-box data-driven branch are introduced to correct the parameters of the node admittance matrix; In the black-box data-driven branch, the load priori knowledge in the white-box analytical branch is introduced, and the real-time data acquired under various working conditions are combined to construct a voltage-power characteristic surface constraint parameter search space.

8. The inertia active support demand evaluation method of a new energy power system according to claim 1, characterized in that, In the step S3, the observability of the system state is judged by calculating the singular value decomposition result of the Gram matrix; when the minimum singular value is greater than a preset threshold, it is determined that the system satisfies the observability condition, otherwise the sensor configuration needs to be adjusted or the measurement dimension needs to be increased.

9. The inertia active support demand evaluation method of a new energy power system according to claim 1, characterized in that, In the step S3, the H2 norm solution needs to substitute the eigenvalue of the Gram matrix into the integral expression, and calculate the energy gain of the system to the random input by the numerical integral method.

10. The inertia active support demand evaluation method of a new energy power system according to claim 1, characterized in that, In the step S4, first, based on a specific scenario, the specific inertia problem and the boundary are determined; then, define P(m) as the solution of the Lyapunov equation, and convert the cost function into a function f(m) of the inertia vector and an inertia allocation vector mi; finally, the algorithm calculates the explicit gradient Δf(m) of f(m).

11. The inertia active support demand evaluation method of a new energy power system according to claim 1, characterized in that, The step S5 includes: S5-1: Space inertia modeling, the power grid is divided into multiple control areas, each area is configured with independent inertia resources, the inertia distribution matrix is constructed to quantify the inertia level of each area; S5-2: Demand evaluation, a comprehensive evaluation index is constructed, which integrates frequency deviation, inertia reserve and spatial distribution characteristics, and the inertia demand of the system is quantified by weighted summation.

12. A data processing device, characterized by It includes: A memory for storing a computer program; A processor for executing the computer program to implement the steps of the inertia active support demand evaluation method of a new energy power system according to any one of claims 1 to 11.

13. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the inertia active support demand evaluation method of a new energy power system according to any one of claims 1 to 11.

Images