A method for stability analysis and oscillation suppression of a grid-forming new energy system

Abstract

The application relates to the technical field of power system stability analysis and control, and specifically discloses a network-constructing new energy system stability analysis and oscillation suppression method, which comprises the following steps: in a unified coordinate system, a full-system small-signal state space model containing network-constructing and network-following converters, a passive network and loads is established, a global state matrix is formed, and multi-time scale dynamics are represented; eigenvalue decomposition is performed on the global state matrix, oscillation modes are divided according to frequencies, and ultra-low-frequency to high-frequency oscillation modes are identified; by establishing the full-system small-signal state space model in the unified coordinate system and combining two-stage screening of eigenvalue decomposition and participation factors, the application can accurately trace the ultra-low-frequency, low-frequency, subsynchronous and high-frequency oscillation modes of different time scales to specific key excitation equipment and core control links, overcomes the problems that in traditional methods, the mapping between oscillation modes and physical equipment is fuzzy, and it is difficult to lock the instability source, and provides clear targeting targets for oscillation suppression.

Description

Technical Field

[0001] This invention relates to the field of power system stability analysis and control technology, and in particular to a method for stability analysis and oscillation suppression of a grid-type new energy system. Background Technology

[0002] With the large-scale integration of grid-connected converters as new grid support equipment, the power system exhibits characteristics of low inertia, weak damping, and dynamic interaction across multiple time scales. Grid-connected control endows converters with voltage source characteristics, and its control loop includes a power synchronization loop, a virtual impedance loop, and a voltage and current dual loop, forming a complex dynamic coupling relationship with grid impedance and other converters. At the same time, grid-connected converters, passive networks, and load dynamics further increase the oscillation risk of the system. Accurately identifying oscillation modes in different frequency ranges and clarifying their relationship with physical equipment and control links is the foundation for ensuring the safe and stable operation of grid-connected new energy systems.

[0003] In existing multi-machine heterogeneous grid-connected new energy systems, the fuzzy mapping between oscillation modes at different time scales and specific equipment and control links is difficult to decouple, making it impossible to quickly pinpoint the source of instability. Moreover, the nonlinear coupling effect and sensitivity of various control parameters on system stability vary significantly, and conventional parameter adjustments can easily induce instability in neighboring modes or deteriorate dynamic response performance. Furthermore, when the actual system operates in a weak grid or multi-condition switching scenario, the oscillation suppression effect will significantly degrade or even trigger new instability risks due to model simplification or operating point shift. Therefore, a stability analysis and oscillation suppression method for grid-connected new energy systems is proposed to address the aforementioned problems. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, the present invention provides a method for stability analysis and oscillation suppression of a grid-type new energy system, which can effectively solve the problems involved in the prior art.

[0005] The objective of this invention can be achieved through the following technical solution: This invention provides a method for stability analysis and oscillation suppression of a grid-type new energy system, comprising the following steps:

[0006] Step S1: Under a unified coordinate system, establish a small-signal state-space model of the entire system, including grid-type and grid-type converters, passive networks, and loads, to form a global state matrix, representing the dynamics at multiple time scales. The dynamics at multiple time scales are fully coupled to avoid analysis bias caused by model fragmentation.

[0007] Step S2: Perform eigenvalue decomposition on the global state matrix, classify oscillation modes according to frequency, identify oscillation modes from ultra-low frequency to high frequency, combine with participation factor analysis to reveal the coupling and interaction mechanism between dynamics at different scales, identify the dominant oscillation mode, accurately classify broadband oscillation modes, and lock in the key modes that play a decisive role in stability.

[0008] Step S3: Based on the two-level screening of participation factors, the dominant oscillation mode is accurately traced to the key equipment and core control links. The modal sensitivity method is used to evaluate the partial derivative sensitivity of system stability to key parameters. The source of oscillation is quickly located to specific equipment and control links, overcoming the fuzzy problem of traditional tracing.

[0009] Step S4: Eliminate redundant parameters by sorting them according to the partial derivative sensitivity. With the goal of maximizing the minimum damping ratio, construct an optimization problem with stability margin and dynamic performance constraints. Precisely select optimization variables to avoid blindly adjusting parameters and causing instability in neighboring modes.

[0010] Step S5: The optimal control parameters are solved by combining particle swarm optimization algorithm with sequential quadratic programming. Its robustness under various typical scenarios is verified. The combination of global search and local fine optimization ensures that the parameters are effective under multiple operating conditions.

[0011] Step S6: Substitute the optimal control parameters into the electromagnetic transient simulation model of the grid-type new energy system, inject targeted disturbances for time-domain verification, evaluate the damping ratio improvement rate and oscillation amplitude suppression effect, and close the theoretical analysis and engineering verification loop to ensure that the suppression scheme is practically operable.

[0012] Preferably, step S1 specifically includes:

[0013] Under a unified dq rotating coordinate system, linearized control models for grid-type converters and grid-following converters are established respectively. The linearized control models at least cover the dynamic equations of the power synchronization loop, outer loop controller, inner loop controller and output filter, effectively eliminating the coupling error introduced by coordinate transformation and ensuring phase consistency of multi-machine models.

[0014] A passive network linearization model incorporating the equivalent impedances of transformers, transmission lines, and power grids is established, along with a load linearization dynamic model reflecting the coupling characteristics of load power and voltage frequency. This model fully incorporates network impedance and load dynamics, overcoming the stability assessment bias caused by static load assumptions.

[0015] A linearized control model combining converters, networks, and loads is established. Algebraic variables are eliminated through the voltage and current connection relationships of system nodes, forming a global state matrix expressed by a global state vector. This achieves a standardized expression of the entire system's state space and is directly compatible with linear system analysis toolchains.

[0016] Preferably, step S2 specifically includes:

[0017] Numerical eigenvalue decomposition is performed on the global state matrix to obtain the real and imaginary parts of each eigenvalue. The real part represents the mode decay rate, and the imaginary part represents the oscillation angular frequency. Eigenvalue decomposition can quantify the stability margin and oscillation frequency of each oscillation mode.

[0018] Based on the imaginary part of the eigenvalues, the oscillation modes are classified into ultra-low frequency mode, low frequency oscillation mode, subsynchronous supersynchronous oscillation mode and high frequency oscillation mode, covering a dynamic spectrum of multiple time scales from electromechanical to electromagnetic. Frequency classification can achieve accurate identification and positioning of wide-band oscillation modes.

[0019] The patterns corresponding to several eigenvalues ​​whose real part is closest to the imaginary axis of the complex plane and whose sum of participation factors is the highest are identified as the dominant oscillation modes. The identification of dominant oscillation modes focuses on the key oscillation risks that have the greatest impact on system stability.

[0020] Preferably, step S2 further includes:

[0021] Calculate the standard participation factor of each state variable for each oscillation mode, quantify the correlation strength between state variables and oscillation modes, and directly identify key state variables that are strongly correlated with oscillations through the participation factor values.

[0022] Based on the participation factor values, a mapping relationship is established between the oscillation mode and the state variables of a specific physical device and the corresponding control loop, so as to achieve accurate physical tracing of the oscillation mode to the specific device and control loop.

[0023] Based on the results of the participation factor analysis, the coupling interaction paths and dominant incentive directions between dynamic processes at different time scales are revealed, and the coupling relationships and main sources of incentives among dynamic processes at multiple time scales are clarified.

[0024] Preferably, step S3 specifically includes:

[0025] For each dominant oscillation mode, the sum of the participation factors of all state variables within each device is calculated, and the devices with the highest sum are selected as key excitation devices to achieve rapid and accurate identification of the oscillation source and avoid missing key contributing devices.

[0026] Within the key excitation equipment, the state variables corresponding to each control link are sorted by participation factors, and the control link to which the state variable with the highest value belongs is selected as the core participation link. The responsibility for oscillation is precisely assigned to the specific control loop, thereby improving the targeting of parameter adjustment.

[0027] The system outputs structured localization results, which include oscillation mode frequency, damping ratio, key excitation device identifiers, and names of core participating components, providing clear and traceable physical basis for subsequent parameter optimization.

[0028] Preferably, step S3 further includes:

[0029] Using the eigenvalues ​​of the dominant oscillation mode as the dependent variable and the grid short-circuit ratio, equipment active power output, virtual inertia, and virtual damping coefficient as independent variables, the partial derivative sensitivity of the eigenvalues ​​with respect to their respective variables is calculated to accurately quantify the direction and intensity of each parameter's control over stability.

[0030] Based on the sign and magnitude of the partial derivative sensitivity, we can determine whether changes in key parameters promote or inhibit the stability of the oscillation mode, and quickly identify the risks and benefits of parameter adjustments.

[0031] Using the selected key parameters as the x-axis and the real and imaginary parts of the eigenvalues ​​as the y-axis, an eigenvalue locus diagram is plotted to visually present the path of the influence of parameter changes on system stability and clearly reveal the evolution trajectory of system stability under parameter changes.

[0032] Preferably, step S4 specifically includes:

[0033] The adjustable parameters of the core participating components are defined as the set of optimization parameters. These adjustable parameters include virtual inertia, virtual damping coefficient, proportional gain and integral coefficient of the proportional-integral controller, and virtual impedance parameters, which effectively limit the optimization range and avoid interference from irrelevant parameters.

[0034] With the objective function of maximizing the minimum damping ratio of the system, a parameter optimization problem is constructed. The minimum damping ratio is directly related to the attenuation capability of the weakest oscillation mode of the system, directly improving the damping of the weakest mode and preventing the system from becoming unstable due to local instability.

[0035] Set optimization constraints, including ensuring that the real parts of all eigenvalues ​​are less than the preset stability margin threshold, and that key response quantities meet dynamic performance limit requirements under time-domain step disturbances, balancing stability margin and dynamic quality to avoid slow response or overshoot.

[0036] Preferably, step S5 specifically includes:

[0037] By sorting all candidate parameters using partial derivative sensitivity, redundant parameters with sensitivity close to zero are eliminated, and the parameters with the highest sensitivity are selected as the initial variable set for co-optimization, which effectively reduces the computational dimension and improves optimization efficiency.

[0038] Within the reduced parameter space, the particle swarm optimization algorithm is called to perform a global search. In each iteration, the global state matrix is ​​recalculated and the objective function and constraints are evaluated to achieve global optimization and avoid getting trapped in local optima.

[0039] Using the near-optimal solution output by particle swarm optimization as the initial value, the sequential quadratic programming algorithm is called to perform a local fine search to obtain the optimal set of control parameters that satisfy the constraints, thereby improving the local convergence accuracy and ensuring the optimality of the solution.

[0040] Preferably, step S5 further includes:

[0041] Multiple typical operating scenarios, including weak grids, strong grids, and different active power output levels, are selected to build a wide-condition verification case library to ensure that the optimization parameters cover the typical range of changes in grid strength and power dispatch in actual operation.

[0042] The optimized control parameters were substituted into the global state matrix of each typical scenario, and eigenvalue decomposition and dominant oscillation mode identification were performed again to verify the consistency and effectiveness of the optimized parameters for oscillation mode identification under different operating conditions.

[0043] By comparing the minimum damping ratio and the real distribution of eigenvalues ​​of the system under different scenarios, it is verified that the optimized parameters can maintain the preset stability margin under different operating conditions, ensuring that the system still has sufficient stability margin when the operating conditions change or the operating point deviates.

[0044] Preferably, step S6 specifically includes:

[0045] The optimized control parameters are applied to the electromagnetic transient simulation model that strictly corresponds to the small-signal model of the whole system, maintaining the consistency of model topology and parameters, ensuring that theoretical analysis and time-domain verification are aligned under the same model benchmark, and eliminating evaluation bias caused by model differences;

[0046] Injecting targeted perturbations that match the characteristics of the dominant oscillation mode into the electromagnetic transient simulation model, such perturbations include active power steps or voltage drops, accurately excites the target oscillation mode, and avoids interference from unrelated dynamics on the evaluation of the suppression effect.

[0047] By comparing the waveforms of key response quantities under the same disturbance before and after optimization, the improvement of damping ratio and oscillation amplitude suppression rate are evaluated, and the oscillation suppression effect is comprehensively judged. The actual improvement of oscillation attenuation capability by parameter optimization is intuitively quantified.

[0048] Compared with the prior art, the beneficial effects of the present invention are:

[0049] 1. This method for stability analysis and oscillation suppression of a grid-type new energy system establishes a small-signal state-space model of the entire system under a unified coordinate system. By combining eigenvalue decomposition and two-level screening of participation factors, it can accurately trace the ultra-low frequency, low frequency, subsynchronous and high frequency oscillation modes at different time scales to specific key excitation equipment and core control links. This overcomes the problems of fuzzy mapping between oscillation modes and physical equipment and difficulty in locking the source of instability in traditional methods, and provides a clear target for oscillation suppression.

[0050] 2. This method for stability analysis and oscillation suppression of a grid-type new energy system calculates the partial derivative sensitivity of key eigenvalues ​​with respect to candidate parameters based on the results of the participation factor positioning. After eliminating redundant parameters, it constructs a constrained optimization problem with the goal of maximizing the minimum damping ratio. This limits parameter adjustment to the core link with the highest efficiency in controlling the dominant oscillation mode, effectively avoiding the drawbacks of conventional parameter tuning that induce instability of neighboring modes or deterioration of dynamic response performance due to nonlinear coupling.

[0051] 3. This method for stability analysis and oscillation suppression of a grid-connected new energy system constructs a wide-condition verification case library covering weak grids, strong grids, and different active power output levels. The optimized control parameters are substituted into each typical operating scenario for eigenvalue reanalysis to ensure that the minimum damping ratio of the system meets the stability margin requirements under all preset operating conditions. This solves the problem of oscillation suppression effect degradation or even new instability risks when the actual system operates in a weak grid or multi-condition switching scenario. Attached Figure Description

[0052] Figure 1 This is a schematic diagram illustrating the workflow of a stability analysis and oscillation suppression method for a grid-type new energy system according to the present invention.

[0053] Figure 2 This is a schematic diagram of the method flow for stability analysis and oscillation suppression of a grid-type new energy system according to the present invention. Detailed Implementation

[0054] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0055] Example 1, please refer to Figure 1 , Figure 2 This invention provides a technical solution: a method for stability analysis and oscillation suppression of a grid-type new energy system, comprising the following steps:

[0056] Step S1: Under a unified coordinate system, establish a small-signal state-space model of the entire system, including grid-type and grid-following converters, passive networks, and loads, forming a global state matrix to characterize multi-time-scale dynamics. The multi-time-scale dynamics are fully coupled, avoiding analysis bias caused by model fragmentation. Under a unified dq rotating coordinate system, establish linearized control models for grid-type and grid-following converters respectively. The linearized control model should at least cover the dynamic equations of the power synchronization loop, outer loop controller, inner loop controller, and output filter, effectively eliminating coupling errors introduced by coordinate transformation and ensuring phase consistency of multi-machine models. Establish a linearized passive network model including transformers, transmission lines, and equivalent grid impedances, and establish a linearized dynamic model of the load reflecting the coupling characteristics of load power and voltage frequency, fully incorporating network impedance and load dynamics to overcome stability assessment bias caused by static load assumptions. Combine the linearized control models of converters, networks, and loads, eliminate algebraic variables through the voltage and current connection relationships of system nodes, and form a global state matrix expressed by a global state vector, realizing a standardized expression of the entire system state space and directly adapting to the linear system analysis toolchain.

[0057] It should be noted that, in the dq rotating coordinate system, the linearized control model uses the system's common reference angular frequency as a reference to ensure that the dynamic equations of all converters are aligned within the same coordinate frame. The modeling comprehensively covers the dynamic equations of the power synchronization loop, outer loop controller, inner loop controller, and output filter, respectively reflecting the converter's behavior at different time scales such as inertial response, voltage regulation, current tracking, and harmonic attenuation. The linearization of each control loop revolves around the same steady-state operating point, and its incremental form state equation is obtained through small-signal perturbation approximation. A linearized dynamic model of the passive network and load is established. The passive network linearization model includes the transformer, transmission line, and equivalent impedance of the grid, and uses lumped parameter or distributed parameter line models to describe its voltage drop and current distribution. The load model is linearized in the dq coordinate system, employing a dynamic load structure that reflects the coupling characteristics of power, voltage, and frequency. All algebraic and differential equations of the network and load are expressed in the same coordinate system and share the same reference angular frequency with the converter model. Through node voltage equations and current continuity conditions, the electrical quantities at the physical connection points of each component are algebraically coupled. A node admittance matrix is ​​constructed based on the system topology, and the injected current of each converter is expressed as a linear combination of state variables. These are substituted into the network equations to solve for the algebraic expressions of the node voltages. Subsequently, the algebraic relationships are substituted back into the differential equations of the converter and load to eliminate intermediate voltage and current variables, ultimately yielding a standard form containing only the global state vector and its derivatives. ,in, The global state vector represents the incremental offset of the state variables of all dynamic components of the system (including grid-type converters, grid-connected converters, dynamic loads, etc.) near their steady-state operating point. The increment vector of the rate of change of the state variables represents the global state vector. The increment of the first derivative of each state variable with respect to time, i.e., the perturbation component of the rate of change of the state variable; The system state matrix is ​​a constant square matrix that fully encapsulates all linearized information of the system from millisecond-level electromagnetic transients to second-level electromechanical dynamics.

[0058] Step S2: Perform eigenvalue decomposition on the global state matrix, classify oscillation modes according to frequency, identify oscillation modes from ultra-low frequency to high frequency, and reveal the coupling and interaction mechanism between dynamics at different scales by combining participation factor analysis, identify the dominant oscillation mode, accurately classify broadband oscillation modes, and lock in the key modes that play a decisive role in stability. Perform numerical eigenvalue decomposition on the global state matrix to obtain the real and imaginary parts of each eigenvalue, where the real part represents the mode decay rate and the imaginary part represents the oscillation angular frequency. Eigenvalue decomposition can quantify the stability margin and oscillation frequency of each oscillation mode. Based on the imaginary part of the eigenvalue, the oscillation modes are classified into ultra-low frequency mode, low frequency oscillation mode, subsynchronous and supersynchronous oscillation mode, and high frequency oscillation mode, covering a multi-timescale dynamic spectrum from electromechanical to electromagnetic. Frequency classification can achieve accurate identification and positioning of broadband oscillation modes. The modes corresponding to the eigenvalues ​​whose real part is closest to the imaginary axis of the complex plane and whose sum of participation factors is the highest are identified as dominant oscillation modes. The identification of dominant oscillation modes focuses on the key oscillation risks that have the greatest impact on system stability.

[0059] It should be noted that when performing numerical eigenvalue decomposition on the global state matrix, the standard linear algebra library is first used to perform batch calculations of eigenvalues ​​and eigenvectors. For complex conjugate eigenvalues ​​that may appear in the real matrix, real Schur decomposition or the QR algorithm is used to extract the real and imaginary parts of all eigenvalues. The real part directly corresponds to the decay rate of each oscillation mode, and its absolute value reflects the convergence speed of the mode's natural response after a disturbance. The imaginary part determines the angular frequency of the oscillation. To ensure numerical accuracy, a sparse eigenvalue solving technique is used for large-scale state matrices, calculating only a few eigenvalues ​​with the largest real parts to reduce the computational burden. A negative real part with a smaller absolute value indicates weaker damping, and the mode is more likely to generate sustained oscillations under slight disturbances. After the eigenvalue calculation is completed, the oscillation frequency corresponding to the imaginary part of each eigenvalue is classified into four levels. The ultra-low frequency mode corresponds to oscillations with frequencies below 0.1 Hz, mainly related to system-level power distribution and inertial response processes. The low frequency oscillation mode covers the 0.1 to 2 Hz frequency band, reflecting the oscillations between converters or between converters and the grid. The power angle swing characteristics; subsynchronous and supersynchronous oscillation modes are in the range of 5 to 50 Hz, originating from the negative damping resonance formed by the control loop and the grid impedance in a specific frequency band, which is the most typical instability mode of the grid-type system; high-frequency oscillation modes are higher than 50 Hz, which are related to the bandwidth of the current inner loop and the resonant peak of the filter. This classification covers the complete dynamic spectrum from electromechanical time scale to electromagnetic time scale; the dominant oscillation mode is identified by comprehensively examining the two dimensions of eigenvalue position and participation factor distribution. First, all eigenvalues ​​whose real parts are closest to the imaginary axis of the complex plane are screened out, that is, the mode with the smallest absolute value of the real part. On this basis, the participation factors of all state variables under each mode are calculated and summed. The modes corresponding to the eigenvalues ​​with the highest sum of participation factors are selected as the dominant oscillation modes. The higher the sum of participation factors, the more extensive the coupling between the oscillation mode and the internal dynamic process of the system. After its excitation, it affects more equipment and control links. Through the dual screening criteria, the few key modes that play a decisive role in the stability of the system are accurately located from multiple eigenvalues.

[0060] Furthermore, step S2 also includes: calculating the standard participation factor of each state variable for each oscillation mode, quantifying the correlation strength between the state variables and the oscillation mode, directly identifying key state variables strongly correlated with the oscillation through the participation factor values, establishing a mapping relationship between the oscillation mode and the state variables of a specific physical device and the corresponding control link based on the participation factor values, realizing accurate physical tracing from the oscillation mode to the specific device and control link, and revealing the coupling interaction path and dominant excitation direction between dynamic processes at different time scales based on the participation factor analysis results, clarifying the coupling relationship and main excitation source between dynamic processes at multiple time scales;

[0061] It should be noted that after completing eigenvalue decomposition and identifying the dominant oscillation mode, the standard participation factor of each state variable with respect to each oscillation mode is calculated to quantify the correlation strength between the state variable and the oscillation mode. Specifically, for each mode, the product of corresponding elements in its right and left eigenvectors is calculated, and the modulus value is taken as the participation factor of that state variable with respect to that mode. This value reflects the relative energy share contributed by each state variable when the mode is excited. By traversing all state variables and all modes, a complete participation factor matrix is ​​formed. The state variables are then grouped by device, with each device containing state variables corresponding to multiple control loops. For each dominant oscillation mode, the sum of the participation factors of all state variables within each device is calculated to identify... The main associated equipment of this mode is then used to rank the participation factors of each state variable within the equipment, and locate the specific control link with the highest contribution. The mapping result is output in a structured form, clearly identifying the physical source of the oscillation mode. By comparing the distribution of participation factors of multiple equipment and multiple control links under the same oscillation mode, it is determined whether the mode is dominated by the local dynamics of a single equipment or by the interactive coupling between multiple equipment. Further analysis is conducted on the overlap of participation factors between different frequency modes. If a low-frequency mode and a subsynchronous mode share multiple high participation factor state variables, it indicates that there is a strong coupling path between the two. The low-frequency oscillation may modulate the damping characteristics of the subsynchronous mode through this path. Based on this, a causal chain from the excitation source to the response end is drawn to clarify the dominant excitation direction.

[0062] Step S3: Based on two-level screening of participation factors, the dominant oscillation mode is accurately traced to key equipment and core control links. The modal sensitivity method is used to evaluate the partial derivative sensitivity of system stability to key parameters, and the oscillation source is quickly located to specific equipment and control links, overcoming the fuzzy problem of traditional tracing. For each dominant oscillation mode, the sum of participation factors of all state variables within each equipment is counted, and the equipment with the highest total is selected as the key excitation equipment, realizing the rapid and accurate locking of the oscillation source and avoiding the omission of key contributing equipment. Within the key excitation equipment, the state variables corresponding to each control link are sorted by participation factors, and the control link to which the state variable with the highest value belongs is selected as the core participating link, and the oscillation responsibility is accurately assigned to the specific control loop, improving the targeting of parameter adjustment and outputting structured positioning results. The results include oscillation mode frequency, damping ratio, key excitation equipment identification, and core participating link name, providing clear and traceable physical basis for subsequent parameter optimization.

[0063] It should be noted that, based on the participation factor matrix, the state variables are grouped and statistically analyzed for each dominant oscillation mode by device. Assuming there are N converter devices in the system, for the ... One dominant oscillation pattern Calculate the first The sum of participation factors of all state variables within the device ,in For the first time in this device A pair of state variables The participation factors, for all devices according to Sort the devices from largest to smallest, and select the top-ranked device as the key activation device. If the top-ranked and second-ranked devices are... If the difference is less than 0.05, both are selected simultaneously. This screening criterion ensures that the positioning results cover the main energy distribution area of ​​the oscillation mode, avoiding the omission of key contributing sources due to similar distribution of participating factors. After identifying the key excitation device, the core control link is further located within the device. All state variables and their participating factor values ​​corresponding to the device are obtained and sorted from largest to smallest. The control link to which the state variable with the highest participating factor value belongs is selected as the core participating link. If the difference between the participating factors of the highest and second highest state variables is less than 0.08, two control links are selected simultaneously. Taking a grid-type converter as an example, its state variables include the virtual power angle and virtual angular velocity of the power synchronization loop, the proportional and integral states of the voltage outer loop, and the proportional and integral states of the current inner loop. The variable with the highest participating factor directly indicates the control level that makes the greatest contribution to the damping of the oscillation mode. The structured positioning results are output in the form of a data table. Each record corresponds to a dominant oscillation mode. The record fields include: oscillation mode number, real part of eigenvalue. Imaginary part of eigenvalues oscillation frequency Damping ratio The key stimulating equipment name and number, the core participating link name and the corresponding state variable name, and the participation factor value of the state variable;

[0064] Furthermore, step S3 also includes: taking the characteristic value of the dominant oscillation mode as the dependent variable, and the grid short-circuit ratio, equipment active power output, virtual inertia and virtual damping coefficient as independent variables, calculating the partial derivative sensitivity of the characteristic value with respect to each independent variable, accurately quantifying the direction and intensity of each parameter's control on stability, judging the promoting or inhibiting trend of each key parameter change on the stability of the oscillation mode based on the positive and negative signs and numerical values ​​of the partial derivative sensitivity, quickly identifying the risk and benefit direction of parameter adjustment, and plotting the characteristic root locus diagram with the selected key parameters as the abscissa and the real and imaginary parts of the characteristic value as the ordinate, intuitively presenting the influence path of parameter changes on system stability, and clearly revealing the evolution trajectory of system stability under parameter changes;

[0065] It should be noted that, for the identified dominant oscillation mode, its eigenvalues ​​are... As the dependent variable, The imaginary unit satisfies Used to distinguish between the real and imaginary parts of a complex number. This indicates that the eigenvalues ​​appear in the form of conjugate complex pairs, meaning that an oscillation mode simultaneously contains two eigenvalues: a positive imaginary part and a negative imaginary part. The real parts of both are the same. The grid short-circuit ratio (SCR) and the active power output of the equipment are selected. Using virtual inertia J and virtual damping coefficient D as independent variables, the partial derivative sensitivity is calculated using the numerical perturbation method. Specifically, at the system's steady-state operating point, the SCR and... Apply a small increment of +1% to J and D, changing only one independent variable each time while keeping the other parameters unchanged from the baseline. Reconstruct the state matrix of the entire system and solve for the eigenvalues ​​using the difference formula. Calculate the values ​​of each partial derivative, if A negative value indicates that increasing the virtual inertia can shift the eigenvalue of the mode to the left half-plane, thus enhancing damping; if A positive value indicates that a decrease in SCR under weak grid conditions will exacerbate the instability risk of this mode, requiring targeted adjustment of control parameters to compensate for the damping reduction caused by insufficient grid strength; when When the absolute value of is greater than 0.05, the virtual damping coefficient D has a significant ability to regulate the attenuation rate of this mode, and the negative sign indicates that increasing D can improve the damping ratio; when A positive value exceeding 0.02 indicates that an increase in current active power output will worsen the stability of the mode, and this needs to be constrained during optimization. The range of variation or the introduction of an adaptive gain scheduling strategy. For parameters with an absolute sensitivity value less than 0.005, they are judged to have a weak impact on the dominant oscillation mode and are eliminated in subsequent parameter optimization to avoid ineffective adjustments. By comparing the magnitude of the partial derivative sensitivity under different independent variables, the regulation efficiency of each parameter on stability is quantified. After completing the sensitivity evaluation, a feature root trajectory diagram is plotted with the selected key parameters as the x-axis to intuitively present the path of the influence of parameter changes on system stability. Taking the virtual inertia J as an example, in to Within the interval, samples are taken uniformly with a step size of 0.5. For each sampling point, the state matrix of the entire system is rebuilt and the eigenvalues ​​are calculated. The real part of each eigenvalue is then... As the horizontal axis, imaginary part Plotting the vertical axis on the complex plane, the lines connecting adjacent sampling points form a trajectory curve, and a stable boundary line is superimposed on the trajectory diagram. and engineering margin line Observe the location where the feature root trajectory crosses the boundary. If the trajectory crosses from the right half-plane to the left half-plane, the J value corresponding to the intersection point is the critical inertia threshold. The actual selected J should be greater than this threshold and retain at least 0.02 real part margin.

[0066] Step S4: Eliminate redundant parameters by sorting them according to the partial derivative sensitivity. With the goal of maximizing the minimum damping ratio, construct an optimization problem with stability margin and dynamic performance constraints. Precisely select optimization variables to avoid blindly adjusting parameters and causing instability in neighboring modes.

[0067] Step S5: The optimal control parameters are solved by combining particle swarm optimization algorithm with sequential quadratic programming. Its robustness under various typical scenarios is verified. The combination of global search and local fine optimization ensures that the parameters are effective under multiple operating conditions.

[0068] Step S6: Substitute the optimal control parameters into the electromagnetic transient simulation model of the grid-type new energy system, inject targeted disturbances for time-domain verification, evaluate the damping ratio improvement rate and oscillation amplitude suppression effect, and close the theoretical analysis and engineering verification loop to ensure that the suppression scheme is practically operable.

[0069] Example 2, as Figure 1 , Figure 2 As shown, based on Embodiment 1, the present invention provides a technical solution: Step S4 specifically includes: defining the adjustable parameters of the core participating links as an optimization parameter set. The adjustable parameters include virtual inertia, virtual damping coefficient, proportional gain and integral coefficient of the proportional-integral controller, and virtual impedance parameters, effectively limiting the optimization range and avoiding interference from irrelevant parameters. The objective function is to maximize the minimum damping ratio of the system, and a parameter optimization problem is constructed. The minimum damping ratio is directly related to the attenuation capability of the weakest oscillation mode of the system, directly improving the damping of the weakest mode and preventing the system from becoming unstable due to local instability. Optimization constraints are set, including that the real parts of all eigenvalues ​​are less than the preset stability margin threshold, and that the key response quantities under time-domain step disturbances meet the dynamic performance limit requirements, taking into account both stability margin and dynamic quality, and avoiding slow response or overshoot.

[0070] It should be noted that after identifying the key excitation equipment and core participating components, the adjustable parameters of the core participating components are uniformly incorporated into the optimization parameter set. The optimized parameter set specifically includes virtual inertia J, virtual damping coefficient D, proportional gain Kp and integral coefficient Ki of each proportional-integral controller, and virtual impedance parameters Rv and Lv. These parameters directly determine the dynamic response characteristics of the grid-type converter under multiple time scales, and participation factor analysis has confirmed that it has the highest control sensitivity to the dominant oscillation mode. The construction of the optimized parameter set is limited to the core participating components, avoiding the inclusion of irrelevant or weakly correlated parameters in the optimization space; the state matrix of the entire system... Eigenvalue decomposition is performed to obtain all eigenvalues, and the damping ratio of each oscillation mode is calculated one by one. The minimum damping ratio is taken as the optimization target to directly improve the attenuation capability of the weakest oscillation mode in the system, preventing the system from becoming unstable due to excessively low damping of individual modes. At the same time, the relative attenuation rate of each mode is taken into account to ensure that all oscillation modes have an engineering-acceptable damping level after optimization. The optimization process must simultaneously meet two types of constraints. The first type is the small-signal stability margin constraint: setting a stability margin threshold. The optimization requires that the real parts of all eigenvalues ​​satisfy the following condition. The first constraint ensures that the system has a quantifiable and measurable stability margin, which is different from the critical stability criterion that only requires the real part of all eigenvalues ​​to be <0. The second type is the time-domain dynamic performance constraint: after injecting a standard step disturbance into the electromagnetic transient simulation model, key response quantities such as bus voltage deviation, frequency deviation and output power overshoot must meet the preset limit to prevent slow response or excessive overshoot due to excessive pursuit of damping. The two types of constraints together ensure that the optimization results take into account both stability and dynamic quality.

[0071] Step S5 specifically includes: sorting all candidate parameters using partial derivative sensitivity, eliminating redundant parameters with sensitivity approaching zero, selecting several parameters with the highest sensitivity as the initial variable set for collaborative optimization, effectively reducing the computational dimension and improving optimization efficiency; in the reduced parameter space, calling the particle swarm optimization algorithm for global search, recalculating the global state matrix and evaluating the objective function and constraints in each iteration to achieve global optimization and avoid getting trapped in local optima; using the quasi-optimal solution output by particle swarm optimization as the initial value, calling the sequential quadratic programming algorithm for local fine search to obtain the optimal set of control parameters that satisfy the constraints, improving local convergence accuracy and ensuring the optimality of the solution;

[0072] It should be noted that after completing the sensitivity assessment, all candidate parameters are sorted according to the absolute value of their partial derivative sensitivity to the real part of the eigenvalue of the dominant oscillation mode. A sensitivity threshold is set, and parameters with absolute values ​​less than the threshold are considered redundant parameters with weak influence on the oscillation mode and are discarded. The 2 to 4 parameters with the highest sensitivity are selected as the initial variable set for co-optimization, compressing the high-dimensional parameter space into a low-dimensional subspace, significantly reducing the computational cost of subsequent optimization, while retaining the core parameters with the highest control efficiency and avoiding interference from irrelevant parameters in the optimization process. A particle swarm optimization algorithm is used for global search in the dimensionality-reduced parameter space. During algorithm initialization, several particles are randomly generated within the parameter feasible region, each particle representing a set of parameter vectors to be optimized. In each iteration, each... The parameters corresponding to the particles are substituted into the system state matrix, and the objective function, i.e., the minimum damping ratio of the system, is calculated through eigenvalue decomposition. The constraints of real part margin of eigenvalues ​​and dynamic performance are checked. Based on the historical best position of each particle and the global best position of the population, the velocity and position of the particles are updated to guide the population to converge to a larger region of the objective function. In each iteration step, the sequential quadratic programming approximates the original constrained optimization problem into a quadratic programming subproblem. The Hessian matrix is ​​constructed using the gradient information of the objective function and the constraints. The search direction and step size are calculated to gradually approach the exact optimal point that satisfies the KKT conditions (Karush-Kuhn-Tucker conditions). Through the synergy of the two-layer strategy, the optimal set of control parameters that satisfies the stability margin and dynamic performance constraints is obtained.

[0073] Step S5 also includes: selecting multiple typical operating scenarios including weak grids, strong grids and different active power output levels, constructing a wide-condition verification case library to ensure that the optimized parameters cover the typical variation range of grid strength and power dispatch in actual operation, substituting the optimized control parameters into the global state matrix under each typical scenario, re-performing eigenvalue decomposition and dominant oscillation mode identification, verifying the consistency and effectiveness of the optimized parameters for oscillation mode identification under different operating conditions, comparing the minimum damping ratio of the system and the real distribution of eigenvalues ​​under each scenario, verifying that the optimized parameters can maintain the preset stability margin under different operating conditions, and ensuring that the system still has sufficient stability margin when switching operating conditions or shifting the operating point;

[0074] It should be noted that after solving for the optimal control parameters, a typical operating scenario case library for wide-condition verification is constructed. This library fully considers the grid intensity changes and power dispatch requirements faced by grid-connected new energy systems in actual operation. Specifically, it covers weak grid scenarios, strong grid scenarios, and operating points under multiple different active power output levels. The system topology and equipment parameters of each scenario remain consistent with the original model; only the grid equivalent impedance value and the active power setpoint of each converter are changed to ensure that the verification results truly reflect the adaptability of the optimized parameters under operating condition switching conditions. Each operating scenario in the typical operating scenario case library independently stores its steady-state operating point information. The optimized control parameter set is then substituted into the small-signal state-space model of the entire system corresponding to each typical operating scenario, and the global state matrix under each scenario is re-established. Complete eigenvalue decomposition calculation is then performed. For each typical operating scenario, all oscillation modes are identified, and the corresponding values ​​for each mode are calculated. The damping ratio and the distribution of real eigenvalues ​​are analyzed, focusing on extracting the minimum damping ratio of the system and the eigenvalue information whose real part is closest to the imaginary axis in each scenario. By comparing the changes in the distribution of eigenvalues ​​before and after optimization in the same scenario, the improvement effect of the optimal parameters on the system damping characteristics under different grid strengths and different output levels is quantitatively evaluated, and it is determined whether they can effectively suppress the dominant oscillation mode. The core judgment criterion for wide-condition verification is: in all preset typical operating scenarios, the minimum damping ratio of the system after substituting the optimal control parameters is not lower than the preset stability margin threshold, and the real parts of all eigenvalues ​​are less than the corresponding margin requirements. If the verification result in a certain scenario fails to meet the above criteria, the scenario needs to be included in the additional constraints of the optimization problem, and the parameters need to be re-optimized until all operating conditions meet the requirements. Through multi-scenario cross-verification, it is ensured that the obtained optimal control parameters have good wide-condition robustness and can maintain sufficient stability margin when the actual system operating point shifts or the grid structure changes.

[0075] Step S6 specifically includes: applying the optimized control parameters to the electromagnetic transient simulation model that strictly corresponds to the small-signal model of the whole system, maintaining the consistency of model topology and parameters, ensuring that theoretical analysis and time-domain verification are aligned under the same model benchmark, eliminating evaluation deviations caused by model differences, injecting targeted disturbances that match the characteristics of the dominant oscillation mode into the electromagnetic transient simulation model, including active power step or voltage drop, accurately exciting the target oscillation mode, avoiding interference from unrelated dynamics on the evaluation of the suppression effect, comparing the waveforms of key response quantities under the same disturbance before and after optimization, evaluating the damping ratio improvement and oscillation amplitude suppression rate, comprehensively judging the oscillation suppression effect, and intuitively quantifying the actual improvement of parameter optimization on oscillation attenuation capability;

[0076] It should be noted that when applying the optimal control parameter set obtained from the optimization solution to the electromagnetic transient simulation model, it is necessary to ensure that the simulation model is consistent with the established small-signal state-space model of the entire system in terms of topology, equipment parameters, and control links. Specifically, the optimized virtual inertia, virtual damping coefficient, proportional-integral controller parameters, and virtual impedance values ​​are assigned one by one to the control module of the grid-type converter, and the steady-state operating points of each converter are re-initialized. After parameter loading, a steady-state initialization verification is performed on the electromagnetic transient model to confirm that each state variable can maintain the preset equilibrium point under undisturbed conditions. For the identified dominant oscillation mode, a targeted disturbance matching the characteristics of the mode is injected into the electromagnetic transient simulation model. The disturbance type is selected based on the physical cause of the dominant oscillation mode: for ultra-low frequency or low frequency oscillation modes with frequencies below 2 Hz, an active power step disturbance with an amplitude of 5% to 10% of the rated power is applied; for subsynchronous... For oscillation modes in the frequency bands above a certain level, voltage dips or short-term three-phase short-circuit faults at the grid connection point are used as excitation sources. The application time of all disturbances is set after the system enters steady state, and the duration matches the period of the oscillation mode to be analyzed, to ensure that the disturbance can effectively excite the dynamic response of the target mode. The waveforms of key response quantities under the same targeted disturbances before and after optimization are compared. The key response quantities include grid connection point voltage, grid-connected converter output active power, and line current. The oscillation envelope is extracted from the waveforms, the actual value of the damping ratio of each dominant oscillation mode is calculated, and compared with the damping ratio before optimization to obtain the damping ratio improvement. At the same time, the peak value of steady-state oscillation before and after optimization is measured. The oscillation amplitude suppression rate is calculated by dividing the peak value before optimization by the peak value before optimization and multiplying by 100%. The actual suppression effect of the optimized parameters on oscillation is determined by combining the two indicators of damping ratio improvement and oscillation amplitude suppression rate, forming a complete closed loop from theoretical analysis to time domain verification.

[0077] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for stability analysis and oscillation suppression of a grid-type new energy system, characterized in that, Includes the following steps: Step S1: Under a unified coordinate system, establish a small-signal state-space model of the entire system, including grid-type and grid-type converters, passive networks, and loads, to form a global state matrix that represents the dynamics at multiple time scales; Step S2: Perform eigenvalue decomposition on the global state matrix, divide the oscillation modes according to frequency, identify oscillation modes from ultra-low frequency to high frequency, and combine participation factor analysis to reveal the coupling and interaction mechanism between dynamics at different scales and identify the dominant oscillation mode; Step S3: Based on the two-level screening of participation factors, the dominant oscillation mode is accurately traced to the key equipment and core control links, and the modal sensitivity method is used to evaluate the partial derivative sensitivity of system stability to key parameters. Step S4: Eliminate redundant parameters by sorting them according to the partial derivative sensitivity, and construct an optimization problem with stability margin and dynamic performance constraints with the goal of maximizing the minimum damping ratio; Step S5: The optimal control parameters are solved by combining particle swarm optimization algorithm with sequential quadratic programming, and its robustness under various typical scenarios is verified. Step S6: Substitute the optimal control parameters into the electromagnetic transient simulation model of the grid-type new energy system, inject targeted disturbances for time-domain verification, and evaluate the damping ratio improvement rate and oscillation amplitude suppression effect.

2. The method for stability analysis and oscillation suppression of a grid-type new energy system according to claim 1, characterized in that: Step S1 specifically includes: In a unified dq rotating coordinate system, linearized control models for grid-type converters and grid-following converters are established respectively. The linearized control models at least cover the dynamic equations of the power synchronization loop, outer loop controller, inner loop controller, and output filter. A passive network linearization model incorporating the equivalent impedances of transformers, transmission lines, and power grids is established, along with a load linearization dynamic model reflecting the coupling characteristics between load power and voltage frequency. A linearized control model combining the converter, network, and load is established. Algebraic variables are eliminated through the voltage and current connection relationships of system nodes, forming a global state matrix expressed by a global state vector.

3. The method for stability analysis and oscillation suppression of a grid-type new energy system according to claim 1, characterized in that: Step S2 specifically includes: Numerical eigenvalue decomposition is performed on the global state matrix to obtain the real and imaginary parts of each eigenvalue, where the real part represents the modal decay rate and the imaginary part represents the oscillation angular frequency. Based on the imaginary part of the eigenvalues, the oscillation modes are classified into ultra-low frequency mode, low frequency oscillation mode, subsynchronous supersynchronous oscillation mode and high frequency oscillation mode, covering a multi-timescale dynamic spectrum from electromechanical to electromagnetic. The patterns corresponding to several eigenvalues ​​whose real part is closest to the imaginary axis of the complex plane and whose sum of participation factors is the highest are identified as the dominant oscillation modes.

4. The method for stability analysis and oscillation suppression of a grid-type new energy system according to claim 3, characterized in that: Step S2 further includes: Calculate the standard participation factor of each state variable with respect to each oscillation mode, and quantify the correlation strength between state variables and oscillation modes; Based on the participation factor values, a mapping relationship is established between the oscillation mode and the state variables of specific physical devices and the corresponding control loops; Based on the results of the participation factor analysis, the coupling interaction paths and dominant incentive directions among dynamic processes at different time scales are revealed.

5. The method for stability analysis and oscillation suppression of a grid-type new energy system according to claim 1, characterized in that: Step S3 specifically includes: For each dominant oscillation mode, the sum of the participation factors of all state variables within each device is calculated, and the devices with the highest sum are selected as key excitation devices. Within the key excitation equipment, the state variables corresponding to each control link are sorted by participation factors, and the control link to which the state variable with the highest value belongs is selected as the core participation link. The output is a structured localization result, which includes the oscillation mode frequency, damping ratio, key excitation device identifier, and name of the core participating link.

6. The method for stability analysis and oscillation suppression of a grid-type new energy system according to claim 5, characterized in that: Step S3 further includes: Using the eigenvalues ​​of the dominant oscillation mode as the dependent variable and the grid short-circuit ratio, equipment active power output, virtual inertia, and virtual damping coefficient as independent variables, the partial derivative sensitivity of the eigenvalues ​​with respect to their respective independent variables is calculated. Based on the sign and magnitude of the partial derivative sensitivity, determine whether the changes in each key parameter promote or inhibit the stability of the oscillation mode. Using the selected key parameters as the x-axis and the real and imaginary parts of the eigenvalues ​​as the y-axis, an eigenvalue locus diagram is plotted to visually present the path of how parameter changes affect system stability.

7. The method for stability analysis and oscillation suppression of a grid-type new energy system according to claim 1, characterized in that: Step S4 specifically includes: The adjustable parameters of the core participating components are defined as the set of optimized parameters, which include virtual inertia, virtual damping coefficient, proportional gain and integral coefficient of the proportional-integral controller, and virtual impedance parameters. With the objective function of maximizing the minimum damping ratio of the system, a parameter optimization problem is constructed, where the minimum damping ratio is directly related to the attenuation capability of the weakest oscillation mode of the system. Set optimization constraints, including that the real parts of all eigenvalues ​​are less than the preset stability margin threshold, and that the key response quantities under time-domain step perturbation meet the dynamic performance limit requirements.

8. The method for stability analysis and oscillation suppression of a grid-type new energy system according to claim 1, characterized in that: Step S5 specifically includes: All candidate parameters are sorted using partial derivative sensitivity, redundant parameters with sensitivity close to zero are eliminated, and the parameters with the highest sensitivity are selected as the initial variable set for collaborative optimization. Within the reduced parameter space, the particle swarm optimization algorithm is invoked for global search. In each iteration, the global state matrix is ​​recalculated and the objective function and constraints are evaluated. Using the near-optimal solution output by particle swarm optimization as the initial value, the sequential quadratic programming algorithm is called to perform a local fine search to obtain the optimal set of control parameters that satisfy the constraints.

9. The method for stability analysis and oscillation suppression of a grid-type new energy system according to claim 8, characterized in that: Step S5 further includes: A wide-condition verification case library was constructed by selecting multiple typical operating scenarios, including weak power grids, strong power grids, and different active power output levels. The optimal control parameters obtained through optimization are substituted into the global state matrix under each typical scenario, and eigenvalue decomposition and dominant oscillation mode identification are performed again. By comparing the minimum damping ratio and the real distribution of eigenvalues ​​of the system under different scenarios, it is verified that the optimized parameters can maintain the preset stability margin under different operating conditions.

10. The method for stability analysis and oscillation suppression of a grid-type new energy system according to claim 1, characterized in that: Step S6 specifically includes: The optimized control parameters are applied to the electromagnetic transient simulation model that strictly corresponds to the small-signal model of the whole system, so as to maintain the consistency of the model topology and parameters. Inject targeted disturbances that match the characteristics of the dominant oscillation mode into the electromagnetic transient simulation model; the disturbances include active power steps or voltage dips. By comparing the waveforms of key response quantities under the same disturbance before and after optimization, the improvement in damping ratio and the oscillation amplitude suppression rate are evaluated, and the oscillation suppression effect is comprehensively judged.

Images