Online comprehensive stability control method and system for new power system network type converter

Abstract

This invention discloses a novel online integrated stability control method and system for grid-connected power system converters. The method includes: real-time acquisition of operating status information of multiple grid-connected converters; online calculation of coordination control quantities for coordinating the control modes of each grid-connected converter based on the operating status information; and online adjustment of the grid-connected control parameters of the corresponding grid-connected converters according to the coordination control quantities. This invention also includes a novel online integrated stability control system for grid-connected power system converters, used to implement the above method. This invention achieves online coordination and dynamic parameter adjustment of the control modes of multiple grid-connected converters, effectively improving the overall stable operation capability of the new power system.

Description

Technical Field

[0001] This invention relates to the field of power system automation and control technology, specifically to a novel online integrated stability control method and system for grid-type converters in power systems. Background Technology

[0002] With the continuous grid connection of high proportions of renewable energy sources, represented by wind power and photovoltaics, and the increasing penetration rate of power electronic equipment, the form and operating characteristics of the power system are undergoing fundamental changes, resulting in a new type of power system characterized by a high degree of power electronics in all aspects of power generation, grid, load, and storage. This system presents new challenges such as reduced inertia, weakened disturbance resistance, and complex dynamic characteristics, and its voltage and frequency stability support methods are fundamentally different from those of traditional synchronous generator-dominated systems.

[0003] Against this backdrop, grid-connected converter technology has emerged. Unlike traditional grid-following converters that passively follow the grid voltage and frequency, grid-connected converters, by simulating the operating mechanism of synchronous generators (such as virtual synchronous machine technology) or adopting other active control strategies, can autonomously establish and adjust the amplitude and frequency of their output voltage, thereby providing the necessary voltage and frequency support for the grid. They have become one of the core devices in new power systems that replace the stabilization function of traditional synchronous machines.

[0004] Currently, control strategies for single-unit grid-connected converters, such as virtual synchronous machine control and droop control, have been extensively studied and gradually applied. However, when multiple grid-connected converters operate in the same region, these control strategies based on single-unit designs often operate independently. Due to the lack of a system-level coordination mechanism, the control objectives of different converters may conflict with each other. Their output voltage, frequency, or power support may not only fail to form a synergy but may also weaken each other, or even cause unfavorable interactions between controllers, leading to instability phenomena such as wideband oscillations in the system.

[0005] While existing technologies have explored multi-converter coordination, they largely focus on offline parameter tuning or simple master-slave control based on fixed rules. They lack a control mechanism capable of online, real-time sensing of the system's overall operating status and dynamically and comprehensively coordinating the allocation of control objectives and parameters for each grid-type converter. This lack of a system-level online coordination mechanism prevents grid-type converter clusters from adaptively responding to continuous changes in system operating conditions, hindering their ability to fully leverage their collective potential for stable support of new power systems and posing a significant threat to the safe and stable operation of the system.

[0006] Therefore, how to design an online, system-level integrated stability control method to achieve coordinated optimization of the control performance of multiple grid-connected converters and avoid control conflicts and cluster oscillations has become a key technical problem that urgently needs to be solved. Summary of the Invention

[0007] This application provides a novel online integrated stability control method and system for grid-type converters in power systems, which at least addresses the problems existing in the prior art.

[0008] A first aspect of this application provides a novel online integrated stability control method for grid-type converters in power systems, characterized by comprising the following steps: S1. Real-time acquisition of operating status information of multiple grid-connected converters; S2. Based on the operating status information, calculate online the coordination control quantity used to coordinate the control modes of various grid-type converters; S3. Adjust the grid-type control parameters of the corresponding grid-type converter online according to the coordinated control quantity.

[0009] This application establishes a system-level online dynamic control framework. Through a closed-loop process of real-time data acquisition, online decision-making, and dynamic adjustment, it achieves for the first time the coordinated operation of multiple independently operating grid-type converters as an organic whole. This breaks through the limitations of traditional single-unit fixed parameter control or offline coordination, enabling the converter cluster to respond online and adaptively to changes in system operating conditions. This fundamentally improves the dynamic coordination capability and overall stability level of the new power system in dealing with complex disturbances and suppressing cluster oscillations.

[0010] In this embodiment of the application, step S1 involves the collection of operating status information through the fusion of multi-source heterogeneous information. Specifically, this includes: acquiring the instantaneous values ​​and fundamental components of the three-phase voltage, three-phase current, and output power at the output point of each grid-type converter through local voltage and current sensors; acquiring the grid common connection point voltage, frequency, and phase angle data containing grid connection point information of at least two grid-type converters through a wide-area measurement system or data concentrator; and acquiring key control variables of other grid-type converters in adjacent or preset clusters through communication network interaction. The key control variables include virtual power angle, virtual frequency, active power reference value, and reactive power reference value.

[0011] This application integrates three types of multi-source heterogeneous information: local electrical quantities of the converter, wide-area synchronization quantities of the power grid, and internal key control variables. This application constructs a panoramic and synchronous system dynamic perception capability, providing a complete data foundation that includes both external electrical performance and internal control intentions for subsequent coordinated decision-making. It significantly improves the system's ability to accurately identify potential stability risks, especially oscillations caused by controller interactions, in the early stages, achieving a leap from local measurement to global perception.

[0012] In this embodiment of the application, step S1 further includes an operation status analysis and feature extraction step: performing time-domain and frequency-domain analysis on the collected operation status information to extract feature quantities used to characterize the system's stable state; the feature quantities include at least: the estimated frequency and damping ratio of the system's dominant oscillation mode, the amplitude and phase deviation vector between the output voltage of each grid-type converter and the voltage at the grid's point of common coupling, and the equivalent parameters of the coupling impedance matrix or admittance matrix reflecting the active-reactive power interaction strength among the grid-type converter clusters.

[0013] This application extracts massive amounts of raw data into characteristic quantities with clear physical meaning, such as the damping ratio of the system's dominant oscillation mode, the voltage synchronization deviation vector, and the coupling admittance parameter. This enables a shift from qualitative judgment to quantitative assessment of the system's stability status, allowing for early quantitative warning of instability risks and clearly indicating weak links in stability and unfavorable interaction paths between equipment. This provides crucial decision-making basis for subsequent implementation of precise and targeted coordinated control.

[0014] In this embodiment, step S2, the online calculation of the coordinated control quantity specifically includes the following sub-steps: S21: Based on the operating status information, assess the margin of the current system operating point relative to the preset stability boundary, and identify the key grid-type converters that play a dominant supporting role in system stability and have the potential to cause oscillation risks; S22: According to the preset coordinated strategy model containing multi-objective optimization functions, calculate the control function type and intensity weight allocation scheme that should be prioritized for each key grid-type converter; S23: Quantify the function type and intensity weight into a specific set of control parameter adjustment instructions to form the coordinated control quantity.

[0015] This application transforms the abstract stability problem into a structured online optimization problem through a decision-making process of assessing margin, identifying key equipment, and optimizing allocation. This achieves a qualitative leap from risk perception to generating precise control, enabling limited control resources to be scientifically prioritized and allocated to the key converters that have the greatest impact on system stability. It also clarifies the specific functional directions and strengths that need to be strengthened, greatly improving the pertinence and efficiency of coordinated control.

[0016] In this embodiment, the multi-objective optimization function in the preset coordination strategy model aims to enhance the overall small-disturbance stability of the system, improve the transient voltage / frequency recovery process, and optimize the power distribution among the grid-type converter clusters. It uses the physical limits, control bandwidth, and network security constraints of each grid-type converter as constraints and employs online rolling optimization or optimal control algorithms based on real-time linearization models to solve the problem, thereby dynamically generating an allocation scheme for functional types and intensity weights.

[0017] This application clarifies that the coordination strategy model focuses on multi-objective collaborative optimization, such as enhancing small-disturbance stability, improving transient performance, and optimizing power allocation. It employs advanced algorithms such as online rolling optimization to solve the problem, ensuring that the generated coordinated control quantity can improve system damping and suppress oscillations while taking into account the quality of transient voltage and frequency recovery and operational economy. This achieves the optimal or suboptimal balance of the overall system performance and avoids secondary problems that may be caused by single-objective control.

[0018] In this embodiment, step S3, adjusting the network control parameters online according to the coordinated control quantity, is specifically implemented as follows: For network converters assigned to enhance voltage support, priority is given to adjusting the reference value of their outer loop voltage controller, the reactive power-voltage droop coefficient in the droop coefficient, or the inductive component in the virtual impedance; for network converters assigned to enhance frequency support and inertia response, priority is given to adjusting the virtual inertia, virtual damping coefficient, or active power-frequency droop coefficient in their virtual synchronous machine control; the adjustment is integrated into the converter's local control loop in a superimposed incremental or smooth switching manner.

[0019] This application establishes a clear and direct mapping relationship from abstract decisions on function type and intensity weight to specific control parameter adjustment instructions, and specifies a safe parameter integration method to ensure that upper-level optimization decisions can be accurately and losslessly transformed into lower-level executable control actions. At the same time, by superimposing incremental or smooth switching mechanisms, it avoids secondary impacts on the equipment itself and the power grid caused by parameter mutations, and ensures the stability and reliability of the control process.

[0020] In this embodiment, the method is executed nested across multiple time scales: the first time scale is from milliseconds to seconds, where the local controllers of each grid-type converter perform internal control based on the current fixed parameters; the second time scale is from ten seconds to minutes, where steps S1 to S3 are executed to complete a round of system-level operational status assessment, coordination control quantity calculation, and online adjustment of control parameters, and the updated parameters are sent to the local controllers; the third time scale is above minutes, used for self-learning and adaptive correction of the weight coefficients and stability boundary criteria in the coordination strategy model.

[0021] This application achieves the division of labor and collaboration between rapid equipment response, online system optimization, and strategy self-learning by nesting the execution of complete control tasks at three time scales: millisecond, second-minute, and above minute. It scientifically balances the speed of control, the complexity of optimization, and the adaptability of strategy, enabling the system to cope with instantaneous disturbances, optimize electromechanical dynamics, and evolve over the long term, forming a control system with hierarchical time intelligence and sustainable improvement.

[0022] In this embodiment of the application, the method further includes an online self-updating step for the coordination strategy model: continuously collecting system dynamic response data before and after each execution of steps S1 to S3 to construct a sample dataset; using the sample dataset, periodically or trigger-based updating of key parameters in the coordination strategy model through parameter identification, reinforcement learning, or neural network training methods, including the weight coefficients of the optimization objective function, the threshold in the stability margin assessment, and the parameters of the dynamic equivalent model of the grid-type converter.

[0023] This application endows the coordination strategy model with the ability to perform online self-updates using historical control data, enabling the entire control system to break free from its dependence on fixed prior knowledge and the accuracy of the initial model. By continuously learning from actual operating experience, it constantly corrects its internal model parameters and decision preferences, thereby adaptively tracking the slow changes in power grid structure, equipment characteristics, and operating modes, significantly improving the long-term applicability and robustness of the method.

[0024] In this embodiment, the process of adjusting the grid-type control parameters online is constrained by the safe operating range of parameters determined based on the thermal stress of the converter semiconductor devices, the DC bus voltage fluctuation range, the output current harmonic distortion rate, and the grid connection standards specified in the grid guidelines. Before adjusting the parameters, it is first verified whether the adjustment target value indicated by the coordinated control quantity is within the safe operating range of the corresponding converter. If not, the range boundary value or the safe value calculated according to the preset rules is used as the replacement.

[0025] This application adds a safety constraint verification and correction step based on the physical limits of the equipment and grid connection standards to the online parameter adjustment process. In the control channel that pursues the optimal performance of the system, an inherent safety defense is embedded to ensure that no control command will cause the converter to overload, overheat or exceed the power quality standard. This fundamentally solves the safety and reliability problem in the engineering application of advanced algorithms and is a key guarantee for the industrial implementation of this method.

[0026] Another aspect of this application provides a novel online integrated stability control system for grid-type converters in power systems, comprising: a status information aggregation and processing unit for executing step S1; a coordination decision and calculation unit for executing step S2; and an instruction generation and communication unit for executing instruction generation and issuance in step S3.

[0027] In the above embodiments, the system protects the hardware architecture and software module entities that implement the method. Through the collaborative work of each module, efficient, reliable, and scalable physical support is provided for the implementation of the method. Attached Figure Description

[0028] Figure 1A flowchart illustrating a novel online integrated stability control method for grid-type converters in a power system, provided as an embodiment of this application; Figure 2 A schematic diagram of a novel online integrated stability control system for grid-type converters in a power system provided in this application embodiment; Figure 3 A schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0029] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0030] To make the purpose, technical solution, and advantages of this application clearer, the following will be described in conjunction with the appendix. Figure 1-3 The following is an explanation using specific examples.

[0031] Please refer to Figure 1 , Figure 1 This application provides a novel online integrated stability control method for grid-type converters in power systems, characterized by the following steps: S1. Real-time acquisition of operating status information of multiple grid-connected converters; S2. Based on the operating status information, calculate online the coordination control quantity used to coordinate the control modes of various grid-type converters; S3. Adjust the grid-type control parameters of the corresponding grid-type converter online according to the coordinated control quantity.

[0032] It is understood that this application relates to key technologies for maintaining stability in new power systems with power electronics. The new power system is a modern energy and power system that uses a high proportion of renewable energy (such as wind power and photovoltaics) as the main primary energy source, uses a high proportion of power electronic equipment (such as converters and flexible transmission devices) as the key power conversion and control interface, and has the ability to deeply integrate and flexibly interact with sources, grids, loads and storage to achieve the goals of clean, low-carbon, safe, efficient and flexible intelligent operation.

[0033] A grid-connected converter is a grid-connected converter capable of autonomously establishing and regulating its output voltage and frequency. Its behavior resembles a voltage source, and its core function is to provide active voltage and frequency support to the power grid, distinguishing it from conventional grid-connected converters that merely follow the grid's state. Online integrated stability control refers to the dynamic coordination of the operating states of multiple such converters through continuous monitoring, evaluation, and intervention, without system shutdown, to achieve rapid recovery and long-term stable operation of the system after disturbances. Operating status information includes instantaneous voltage and current values ​​and power information collected locally from the converters, synchronous data such as voltage, frequency, and phase angle at the point of common coupling obtained from the power grid's wide-area measurement system, and internal control variables of adjacent converters, such as virtual power angle and power reference values, obtained through communication network interaction. The coordinated control quantity is a decision instruction calculated by a strategy model containing a multi-objective optimization function based on real-time analysis of the above information. This instruction clarifies which key converters in the system need to be prioritized for adjustment, and what control functions and intensities should be adjusted. Network control parameters refer to the specific adjustable variables inside the converter used to implement the network control strategy. For example, the droop coefficient or virtual impedance value used to adjust the reactive voltage characteristics, and the virtual inertia and virtual damping coefficient used to simulate the inertia and damping characteristics of the synchronous machine. According to the instructions of the coordinated control quantity, these parameters will be adjusted online in the form of superimposed increments or smooth switching, thereby optimizing the overall dynamic performance of the system without interrupting the operation of the equipment.

[0034] This application constructs a system-level panoramic dynamic perception capability through real-time acquisition and fusion of multi-source heterogeneous information. This enables the control system to accurately identify key converters that cause broadband oscillations or insufficient stability margins, and their interaction modes. This transforms stability control from a passive mode based on local measurements and offline analysis to an active prevention mode based on global state and real-time assessment, fundamentally improving the ability to understand and warn of complex stability problems in new power systems. Secondly, based on an online optimization model, coordinated control quantities are dynamically generated, achieving precise and coordinated allocation of control modes for multiple grid-connected converters. This transforms the originally independently operating converter cluster with fixed parameters into an organic whole with dynamically reconfigurable parameters and online assignable functions. The system can adaptively enhance the voltage support or frequency inertia response of specific converters according to real-time operating conditions, optimize the dynamic interaction between clusters, and thus significantly enhance the system's small-disturbance stability at the global level, effectively suppressing broadband oscillations and improving the voltage and frequency recovery process after transient disturbances. Finally, by safely and smoothly adjusting key control parameters online across multiple time scales, this method achieves a dynamic balance between stable regulation and safe equipment operation. Regulation commands are executed while considering the converter's own physical limits and the safety constraints of grid connection standards, ensuring that any parameter adjustment will not cause equipment overload or power quality issues. This closed-loop process enables the system to continuously maintain optimal stable operation without interrupting power supply or relying on large-scale generator and load shedding, significantly improving the power supply reliability and operational resilience of high-proportion renewable energy grid-connected systems.

[0035] In this embodiment of the application, step S1 involves the collection of operating status information through the fusion of multi-source heterogeneous information. Specifically, this includes: acquiring the instantaneous values ​​and fundamental components of the three-phase voltage, three-phase current, and output power at the output point of each grid-type converter through local voltage and current sensors; acquiring the grid common connection point voltage, frequency, and phase angle data containing grid connection point information of at least two grid-type converters through a wide-area measurement system or data concentrator; and acquiring key control variables of other grid-type converters in adjacent or preset clusters through communication network interaction. The key control variables include virtual power angle, virtual frequency, active power reference value, and reactive power reference value.

[0036] The core of the fusion acquisition of operating status information in step S1 of this application embodiment lies in integrating three types of data from different sources and with different properties to construct a comprehensive and synchronous panoramic view of the system dynamics. The fundamental component refers to the power frequency sinusoidal component extracted from the instantaneous voltage and current waveforms through digital filtering or signal decomposition algorithms; this is the basis for analyzing the system's steady-state and low-frequency dynamics. The wide-area measurement system is a high-precision networked monitoring system based on synchronous phasor measurement units, capable of providing voltage phasor and frequency data with a unified time scale to multiple geographically dispersed measurement points with millisecond-level accuracy. The data concentrator is an alternative or supplementary solution when a wide-area measurement system is unavailable, responsible for rapidly aggregating and time-aligning key data from multiple converter local controllers within a specific area. Virtual power angle and virtual frequency are core state variables generated by simulating the rotor motion equation of a synchronous generator in the internal control algorithm of a grid-type converter. They represent the phase angle and rotational speed of the virtual electromotive force constructed by its internal control, respectively, and directly determine the converter's power output and synchronization characteristics. The active power reference value and reactive power reference value are the power control target values ​​set by the upper control loop of the converter according to the scheduling instructions or local strategies, reflecting the current operating setpoint of the equipment.

[0037] The implementation principle of step S1 in this application embodiment is based on a hierarchical, distributed data fusion architecture. At the data source level, the local controller of each grid-type converter first performs high-speed sampling and analog-to-digital conversion on the raw signals output by its voltage and current sensors. For the acquisition of instantaneous values ​​and fundamental components, coordinate transformation based on instantaneous reactive power theory is usually adopted. For example, the three-phase AC quantities are converted into DC quantities in a rotating coordinate system through the existing Clarke transform and Park transform, thereby conveniently separating the fundamental component and calculating instantaneous active and reactive power. These locally calculated electrical quantities, as well as key control variables such as virtual power angle, virtual frequency, and power reference values ​​updated in real time by the internal control algorithm, are encapsulated into time-stamped data packets. At the data aggregation level, the global information of the grid common connection point is directly obtained through the phasor measurement unit of the wide-area measurement system, or actively polled and collected from multiple converters within its jurisdiction by the data concentrator deployed in the substation and completed time synchronization processing. Finally, at the coordination control layer, via a high-speed communication network, adjacent or intra-control converters, or all converters and the upper-level coordination controller, exchange the aforementioned key control variable data packets at preset intervals. The coordination controller utilizes state estimation or data assimilation techniques to fuse and verify all multi-source heterogeneous data within a unified mathematical model framework, eliminating bad data and reconstructing a synchronized and consistent dynamic state sequence for all key nodes in the entire system. This provides a high-quality, highly reliable input dataset for subsequent stability assessments and coordination decisions.

[0038] This embodiment takes a grid-connected cluster in a certain region, which includes five grid-connected photovoltaic power stations (denoted as PV1 to PV5) and one grid-connected energy storage power station (denoted as ESS), as an example to explain in detail the specific implementation process of multi-source heterogeneous information fusion and collection in step S1. The cluster is connected to the regional main grid through a 110kV line, and its common connection point is located at the bus of the 110kV substation.

[0039] The local controllers of the grid-connected converters in each photovoltaic power station and energy storage power station synchronously sample the voltage and current sensor signals at the output point at a frequency of 10,000 times per second. The sampled data is sent to a digital signal processor, where a digital low-pass filter first filters out high-frequency noise above the switching frequency, and then an algorithm based on instantaneous reactive power theory is used for data processing. Specifically, the three-phase voltages (ua, ub, uc) and currents (ia, ib, ic) are transformed using Clarke transform (α-β transform) and Park transform (dq transform) to convert them to a dq coordinate system that rotates synchronously with the grid fundamental frequency.

[0040] In the dq coordinate system, the DC values ​​(ud, uq, id, iq) of voltage and current are mappings to the fundamental components and can be directly used to calculate instantaneous active and reactive power. Simultaneously, the controller records the sequence of instantaneous three-phase voltage and current values ​​before the transformation for subsequent waveform analysis. These calculated instantaneous values ​​of fundamental voltage, current, active power, and reactive power, along with the statistical characteristics (such as peak values) of the original sampling sequence, are packaged at an update rate of 100 times per second to form a local electrical quantity data packet. Each data packet carries a timestamp generated by a local high-precision clock.

[0041] Within the 110kV substation with cluster access, a phasor measurement unit (PMU) of a wide-area measurement system is installed. This PMU accurately measures the three-phase voltage and current of the point of common coupling bus at a rate of 50 frames per second, and calculates the amplitude, frequency, and absolute phase angle (phase relative to UTC zero point) of the voltage fundamental positive sequence component based on a global satellite synchronous clock signal. This voltage amplitude, frequency, and phase angle data, with high-precision unified time scales (accuracy better than 1 microsecond), is uploaded in real-time to the coordination device deployed in the regional control center via the power dispatch data network at a rate of 50 times per second. As a backup, a data concentrator is also installed in the substation. This data concentrator actively polls the critical data of PV1-PV5 and the local controllers within the ESS station at a frequency of 100 times per second via a high-speed local area network, and uses the network time protocol for time alignment to form a synchronous dataset containing the voltage, power, and internal status (such as virtual power angle) of each station's grid connection point, serving as a supplement or verification of the PMU data.

[0042] A communication connection based on a standard (such as IEC 61850) manufacturing message specification is established between the coordinating unit and the local controllers of all grid-type converter sites within the cluster. The coordinating unit broadcasts status information request commands to each site at a frequency of 10 times per second. Upon receiving the request, each site's local controller immediately packages and replies with key status variables from its current internal control loop. These variables include: Virtual power angle: calculated in real-time based on the rotor motion equations simulated in the virtual synchronous machine algorithm; Virtual frequency: obtained from the derivative of the virtual power angle or the output of the phase-locked loop; Active power reference values ​​and reactive power reference values: derived from the power plant-level energy management system or local power regulation commands. These control variable data packets also include precise local timestamps to ensure accurate time alignment and integration with PMU data and local electrical quantity data on the coordinating unit side.

[0043] The coordination unit of the regional control center includes a state information aggregation and processing unit. This unit, acting as the data fusion center, performs the following operations: Time synchronization and alignment: Using the timestamps of each received data packet and the absolute timestamp of the PMU data, all data from different sources and with different transmission delays are interpolated or aligned to a unified time series (e.g., aligned to every integer 10 milliseconds). Data verification and cleaning: Through consistency checks (such as Kirchhoff's laws at the verification node) and reasonable range filtering, bad data caused by communication interference or equipment malfunctions is removed. State estimation and reconstruction: Based on highly reliable PMU data and combined with network topology parameters, state estimation algorithms such as weighted least squares are used to estimate the state of some nodes that cannot be directly measured (such as the power flow of some photovoltaic power plant's collection lines). Panoramic database generation: Finally, for each aligned time point, a unified dynamic information covering the entire cluster is generated. This dynamic information includes: the fundamental voltage / current components and power at each station's outlet, the voltage amplitude / frequency / phase angle at the PCC point, and the virtual power angle, virtual frequency, and power reference values ​​within each station.

[0044] Example scenario: One afternoon, the weather changed from sunny to cloudy, and the irradiance of PV2, PV3, and PV4 power plants fluctuated asynchronously. The fusion database of the coordination device shows: PMU data indicate that the voltage frequency at the PCC point exhibits a periodic, minute fluctuation of approximately 1.8 Hz between 49.98 Hz and 50.02 Hz.

[0045] The local power fluctuations reported by each photovoltaic station were all within the allowable range. However, after fusion and comparative analysis of their virtual power angle data, it was found that the virtual power angle fluctuations of PV3 and PV4 exhibited nearly opposite characteristics, and the fluctuation frequency was consistent with the main components of the PCC frequency fluctuation.

[0046] Meanwhile, the control variables obtained from the ESS station interaction show that its active power reference value is being rapidly adjusted in feedforward control mode to smooth power fluctuations at the PCC point, and its adjustment bandwidth covers the frequency band around 1.8Hz.

[0047] The coordinating unit, through fusion analysis of these three layers of information, concluded that the network-type control of PV3 and PV4, due to their similar parameters, generated unfavorable synchronous interaction under power fluctuation excitation. Simultaneously, the rapid power regulation dynamics of the ESS station unintentionally coupled with this interaction mode, forming a positive feedback loop and jointly exciting a potential negatively damped oscillation mode with a frequency of approximately 1.8 Hz. This early and accurate diagnosis provided crucial decision-making basis for the subsequent step S2, which involved targeted calculation of coordinated control quantities (e.g., differentiated adjustment of the virtual damping coefficients of PV3 and PV4, and optimization of the ESS control bandwidth).

[0048] This application overcomes the limitations of traditional stability control, which suffers from a single information dimension and spatiotemporal asynchrony. By integrating three layers of information—local electrical quantities, wide-area synchronization quantities, and internal control states—it achieves, for the first time within a new power system framework, a holographic, real-time, and synchronous perception of grid-connected converter clusters, encompassing both external electrical behavior and internal control motivations. This provides an invaluable data foundation for accurately diagnosing complex oscillation sources and understanding the dynamic interaction mechanisms between devices, enabling subsequent coordinated control decisions to be based on a profound and comprehensive understanding of the system, thereby significantly improving the accuracy and foresight of stability control.

[0049] In this embodiment of the application, step S1 further includes an operation status analysis and feature extraction step: performing time-domain and frequency-domain analysis on the collected operation status information to extract feature quantities used to characterize the system's stable state; the feature quantities include at least: the estimated frequency and damping ratio of the system's dominant oscillation mode, the amplitude and phase deviation vector between the output voltage of each grid-type converter and the voltage at the grid's point of common coupling, and the equivalent parameters of the coupling impedance matrix or admittance matrix reflecting the active-reactive power interaction strength among the grid-type converter clusters.

[0050] In this embodiment, the operational status analysis and feature extraction steps aim to transform the multi-dimensional raw data obtained through fusion acquisition into quantitative characteristic indicators that can be directly used for stability assessment and decision-making. The dominant oscillation mode of the system refers to the specific frequency oscillation component with the largest amplitude or longest duration in the dynamic response, identified by analyzing the system's measured signals. The damping ratio estimate is a quantitative description of the rate of decay of this oscillation mode; a lower damping ratio indicates that the oscillation is more difficult to quell. The amplitude and phase deviation vector is a set, with each element corresponding to a grid-connected converter, calculating the instantaneous difference in amplitude and phase between the fundamental voltage component at its outlet and the fundamental voltage component at the grid's point of common coupling. The equivalent parameters of the coupling impedance matrix or admittance matrix are simplified mathematical model parameters used to characterize the strength and nature of the mutual influence of active and reactive power changes between any two converters in the context of multiple grid-connected converters coexisting and coupled through the grid.

[0051] Its implementation principle is based on signal processing and system identification theory. In time-domain analysis, the system calculates in real-time the amplitude and phase deviations of the output voltage of each grid-connected converter from the grid's point of common coupling voltage, forming a dynamically changing deviation vector. This vector directly reflects the synchronization tightness between each converter and the main grid; large or rapidly increasing deviations usually indicate a risk of loss of synchronization or severe power oscillations. In frequency-domain analysis, the system continuously monitors the power or frequency signals of key channels and employs identification methods such as the improved Proni algorithm or random subspace-based methods. These methods, by fitting observation data over a period of time, can estimate online the precise frequencies of one or more dominant oscillation modes in the system and their corresponding damping ratios. For obtaining coupling parameters, an online recursive least-squares algorithm is typically used. This algorithm utilizes real-time acquired data on small perturbations in the output voltage and changes in injected current of multiple converters to dynamically identify and update an equivalent admittance matrix model that approximates the dynamic power interaction relationship between them. The parameters of this model reflect the coupling strength between the devices.

[0052] Continuing with the previous example of a cluster comprising several photovoltaic (PV) power plants and one energy storage power plant, the analysis process commenced after integrating the collected voltage, current, virtual power angle, and point-of-combination (PCC) phasor data from each plant. Time-domain analysis revealed that the voltage phase deviation of PV plant No. 3 relative to the PCC voltage exhibited periodic oscillations. Frequency-domain analysis identified the frequency signal at the PCC, extracting a negatively damped oscillation mode with a frequency of 2.5 Hz and a damping ratio of only 0.3%. Furthermore, by analyzing the correspondence between recent minor fluctuations in output voltage and current changes at each plant, the online identification algorithm estimated the equivalent coupling admittance matrix of the cluster. This revealed an abnormally increased equivalent transfer conductance parameter between the energy storage power plant and PV plant No. 3, indicating an excessively strong sensitive link in the active power interaction between the two. These extracted features collectively point to a risk of a low-damped oscillation at a frequency of 2.5 Hz caused by the control interaction between a specific PV plant and the energy storage plant.

[0053] This embodiment continuously tracks the changing trends of the damping ratio, voltage synchronization deviation, and coupling parameters between devices in the dominant oscillation mode. This method can provide early and accurate quantitative warnings of system instability risks and clearly indicate weak links and interaction paths, thus providing crucial direction and basis for the targeted calculation of coordinated control quantities in subsequent steps.

[0054] In this embodiment of the application, step S2, the online calculation of the coordinated control quantity specifically includes the following sub-steps: S21: Based on the operating status information, assess the margin of the current system operating point relative to the preset stability boundary, and identify the key grid-type converters that play a dominant supporting role in system stability and have the potential to cause oscillation risks; S22: According to the preset coordinated strategy model containing multi-objective optimization functions, calculate the control function type and intensity weight allocation scheme that should be prioritized for each key grid-type converter; S23: Quantify the function type and intensity weight into a specific set of control parameter adjustment instructions to form the coordinated control quantity.

[0055] In this embodiment, step S2, which involves online calculation of the coordinated control quantity, is essentially an intelligent decision-making process based on real-time situational awareness. The preset stability boundary refers to a series of mathematical conditions or surfaces pre-defined in the system parameter space or state space through offline analysis or historical learning to distinguish between stable and unstable critical states. The stability margin is an indicator that quantitatively assesses the distance of the current operating point from this boundary, such as the minimum damping ratio calculated through eigenvalue analysis or the transient stability margin calculated using the energy function method. Networked converters that play a dominant supporting role in system stability typically refer to those devices that contribute the most to maintaining system frequency and voltage levels by providing virtual inertia or dynamic reactive power support; while networked converters that potentially trigger oscillation risks refer to those whose control dynamics interact adversely with the network or other device dynamics, and are identified as significantly reducing the damping ratio of specific oscillation modes. Both are collectively referred to as key networked converters. The coordination strategy model is a mathematical framework with embedded decision-making logic. Its core multi-objective optimization function aims to simultaneously optimize multiple, sometimes conflicting, stability indicators. Control function type refers to the specific capability direction that needs to be strengthened, such as voltage support or frequency inertia response; strength weight allocation scheme is the quantitative allocation of the priority and specific contribution of each key device to be adjusted. The final control parameter adjustment instruction set maps the abstract function type and weight scheme into numerical modification commands for specific parameters such as virtual inertia or droop coefficient that can be directly executed by the converter local controller.

[0056] Its implementation principle follows a sequential logic of evaluation-decision-mapping. First, based on the characteristic quantities extracted in step S1, such as the damping ratio of the dominant oscillation mode, voltage deviation vector, and coupling admittance matrix, the system evaluates the stability margin by comparing it with preset thresholds through online calculation. For example, if the damping ratio of the dominant mode is lower than the safety threshold, it is determined that the stability margin is insufficient. At the same time, by analyzing the magnitude of the participation factors of each device in the oscillation mode and the severity of its voltage deviation, the risk sources with the greatest negative impact on the mode and the supporting sources with the greatest potential to suppress the mode can be accurately identified. Subsequently, the preset coordination strategy model is activated. This model takes maximizing the system stability margin as its core objective, while also considering transient performance recovery and power allocation fairness, constructing an optimization problem containing multiple sub-objective functions. The constraints cover the physical limits and control bandwidth of the devices. This model typically uses an optimal control algorithm based on a real-time linearized system model, or, in more complex scenarios, a rolling time-domain optimization method for fast solution, thereby calculating the functional type to be assigned to each key converter and a quantified adjustment weight coefficient. Finally, a pre-defined parameter mapping table is used to convert the function type and weight coefficients into specific and safe control parameter adjustment values. For example, if the decision is that a device needs to enhance its frequency support function with an intensity weight of 0.7, then its virtual inertia parameter is adjusted from its current value to the target value through table lookup or linear interpolation.

[0057] Continuing with the previous example, after feature extraction identifies a 2.5 Hz low-damped oscillation and identifies the No. 3 photovoltaic station and the energy storage station as the key interaction pair, step S2 is initiated. The evaluation phase determines that the current damping ratio of 0.3% is far below the 5% safety lower limit, indicating a severely insufficient stability margin. The coordination strategy model in the decision-making phase begins calculation, with its objective function set to maximize the damping ratio of this oscillation mode while limiting the output voltage deviation of each station. After optimization, the model outputs a decision: the energy storage station is assigned to enhance frequency damping with a weight of 0.8; the No. 3 photovoltaic station is assigned to adjust dynamic coupling with a weight of 0.6. Based on this decision and the equipment's safety range, the mapping phase generates specific instructions: increasing the virtual damping coefficient of the energy storage station from 25 to 45, and fine-tuning the virtual impedance inductive component of the No. 3 photovoltaic station by 15%.

[0058] This application transforms the abstract stability problem into a structured online optimization decision-making process, achieving a qualitative leap from risk perception to the generation of precise control prescriptions. Through quantitative assessment of stability margin and precise positioning of key equipment, it ensures that control resources are allocated to the most effective links. Furthermore, by solving a multi-objective optimization model, it ensures that control measures improve stability while also considering other important system performance aspects, avoiding secondary problems caused by single-objective control, and ultimately achieving intelligent, refined, and optimized improvement of system-level stability.

[0059] In this embodiment, the multi-objective optimization function in the preset coordination strategy model aims to enhance the overall small-disturbance stability of the system, improve the transient voltage / frequency recovery process, and optimize the power distribution among the grid-type converter clusters. It uses the physical limits, control bandwidth, and network security constraints of each grid-type converter as constraints and employs online rolling optimization or optimal control algorithms based on real-time linearization models to solve the problem, thereby dynamically generating an allocation scheme for functional types and intensity weights.

[0060] In this embodiment, the core of the preset coordination strategy model is a multi-objective optimization function that systematically coordinates multiple dimensions of system stability. Enhancing the overall small-disturbance stability of the system specifically refers to adjusting control parameters to improve the damping ratio of the dominant oscillation mode identified in step S1, ensuring that the system can quickly calm down after a small disturbance without triggering amplified oscillations. Improving the transient voltage or frequency recovery process focuses on the dynamic process of key node voltages and system frequencies returning to the normal allowable range after the system encounters a large disturbance, such as after a short-circuit fault is cleared. The optimization objective is usually to minimize overshoot and settling time. Optimizing the power distribution among grid-type converter clusters aims to dynamically optimize the active and reactive power output of each converter based on its capacity and operating status, thereby improving operating economy or balancing equipment load. The physical limits in the constraints include the maximum current and power capacity of the converter semiconductor devices and the allowable fluctuation range of the DC bus voltage; the control bandwidth refers to the frequency range in which its inner and outer loop controllers can effectively respond, and the adjusted parameters must not cause the system to dynamically exceed this range; the network security constraints involve the grid connection standards' hard requirements for power quality such as harmonic distortion rate and power factor.

[0061] The underlying principle is to construct and solve a constrained multi-objective mathematical optimization problem in real time. The decision variables of this problem are a vector composed of the adjustable control parameters of all key grid-type converters. The objective function is usually composed of multiple weighted sub-functions. For example, the first sub-function is the negative value of the damping ratio of the dominant oscillation mode, minimizing it means maximizing damping; the second sub-function is the integral of the critical voltage deviation during the transient process; and the third sub-function is the sum of squares of the deviations between the actual output and the ideal economic allocation value of each converter. All these objectives need to be optimized simultaneously under constraints such as equipment physical limits. The online rolling optimization method places this problem in a sliding time window and solves it repeatedly. In each cycle, it recalculates the optimal parameter adjustment sequence for a future time domain based on the latest system state and only implements the command of the first time step, thus adapting to system changes. The optimal control algorithm based on the real-time linearization model uses the equivalent coupling model obtained from the feature extraction in step S1 to establish the linear state-space equation of the system at the current operating point. Then, it uses design methods such as linear quadratic regulators to directly calculate the state feedback gain that optimizes the comprehensive objective function. This gain implicitly contains the parameter adjustment allocation scheme. Both methods can transform complex dynamic coordination problems into mathematical problems that can be solved online.

[0062] Continuing with the previous example of a cluster with a risk of 2.5 Hz oscillation, the coordination strategy model is invoked here. Its specific optimization objectives are: the primary objective is to increase the damping ratio of this mode from 0.3% to over 5%; the secondary objective is to ensure that the voltage drop at the energy storage station's output does not exceed 15% during a fault; and the auxiliary objective is to maintain the reactive power output of each photovoltaic station according to its capacity ratio as much as possible. Constraints include the maximum instantaneous current of the energy storage converter and the allowable power factor range for each photovoltaic inverter. Using an online rolling optimization method, the model predicts the system dynamics for the next 2 seconds based on the current linearized coupling model, and the optimal parameter adjustment trajectory is obtained after solving: the virtual damping of the energy storage station needs to be significantly enhanced, while its voltage loop integral gain needs to be moderately reduced to balance dynamic response speed and overshoot; the virtual impedance of photovoltaic station No. 3 needs to be adjusted to weaken coupling at a specific frequency. The model transforms the initial adjustment amount of this trajectory into a specific strength weight allocation scheme.

[0063] This embodiment, through a rigorous multi-objective optimization framework, unifies the three objectives—improving stability, ensuring transient performance, and optimizing economic operation—that are often trade-offs in actual operation, into a single decision-making process, achieving Pareto optimization of the system's overall performance. Simultaneously, by embedding strict physical and safety constraints into the solution process, it fundamentally guarantees the feasibility and security of all control commands, ensuring that the generated coordinated control quantities are not only effective but also robust and directly applicable in engineering.

[0064] In this embodiment, step S3, adjusting the network control parameters online according to the coordinated control quantity, is specifically implemented as follows: For network converters assigned to enhance voltage support, priority is given to adjusting the reference value of their outer loop voltage controller, the reactive power-voltage droop coefficient in the droop coefficient, or the inductive component in the virtual impedance; for network converters assigned to enhance frequency support and inertia response, priority is given to adjusting the virtual inertia, virtual damping coefficient, or active power-frequency droop coefficient in their virtual synchronous machine control; the adjustment is integrated into the converter's local control loop in a superimposed incremental or smooth switching manner.

[0065] In this embodiment, step S3 adjusts the grid-type control parameters online based on the coordinated control quantity. Its core is to transform upper-level decisions into lower-level executable actions and ensure a smooth system transition through a secure execution mechanism. The outer-loop voltage controller, within the grid-type converter control hierarchy, is responsible for calculating and outputting current reference commands based on the difference between the target voltage and the feedback voltage. Adjusting its reference value directly changes the converter's voltage setpoint. The reactive power-voltage droop coefficient defines the static slope of the converter's output reactive power as a function of the grid connection point voltage deviation. Increasing this coefficient enhances the strength and speed of reactive power support provided by the converter during voltage fluctuations. The inductive component in the virtual impedance is an impedance value artificially introduced in the control algorithm to simulate inductor characteristics. Increasing this component enhances the converter's voltage source characteristics and affects its dynamic power coupling characteristics. For frequency and inertia support, virtual synchronous machine control is an algorithm that dynamically simulates the rotor motion equations and electromagnetic characteristics of a synchronous generator. Virtual inertia is the time constant simulating the rotor's mechanical inertia in this algorithm, affecting the converter's instantaneous power response to frequency changes; virtual damping coefficient is a parameter simulating the dissipation of oscillating energy by the damping winding; and active power-frequency droop coefficient defines the static relationship between output active power and frequency deviation. The incremental overlay method refers to directly adding or subtracting a value determined by a coordinated control quantity from the existing parameter values; the smooth switching method, within a set transition time, allows the parameters to continuously change from the current value to the target value according to a specific curve, such as a ramp or exponential curve.

[0066] Its implementation principle is based on a deep understanding of the dynamic model and control loop of grid-connected converters. The essence of adjustment is to reshape the dynamic external characteristics of the converter by changing the gain, time constant, or reference setpoint of the control loop. When it is necessary to enhance the voltage support function of a converter, increasing its outer loop voltage controller reference value can directly increase its output voltage amplitude; increasing its reactive power-voltage droop coefficient means that the same voltage drop will trigger a larger reactive current output; and increasing the inductive component of the virtual impedance is equivalent to enhancing the decoupling capability between the converter output and the grid, making it closer to an ideal voltage source, thereby improving the voltage control capability and stability of the local grid. When it is necessary to enhance frequency support and inertia response, increasing the virtual inertia will cause the converter to simulate the inertial release or kinetic energy absorption behavior of a synchronous generator when the system frequency change rate is large, providing instantaneous power support; increasing the virtual damping coefficient will enhance its ability to attenuate power or frequency oscillations; and adjusting the active power-frequency droop coefficient changes its static power allocation ratio participating in primary frequency regulation. To ensure that the adjustment process itself does not introduce shocks, the incremental method is suitable for small, gradual parameter optimization; the smooth switching method is used for larger parameter changes. By controlling the derivative of the transition process, the continuity of power and current output is guaranteed, and secondary disturbances to the power grid are avoided.

[0067] Continuing with the previous example, the coordinated control quantity has generated specific instructions: increase the virtual damping coefficient of the energy storage power station from 25 to 45, and slightly increase the inductive component of the virtual impedance of the No. 3 photovoltaic power station by 15%. During the execution of step S3, the local controller of the energy storage power station receives the adjustment instructions. Due to the large adjustment range of the virtual damping coefficient, to avoid sudden power changes, the controller adopts a smooth switching method, linearly increasing the coefficient from 25 to 45 along a ramp function within 300 milliseconds. At the same time, the local controller of the No. 3 photovoltaic power station adopts a superimposed incremental method, instantaneously increasing the inductive component of its virtual impedance by 15%. Throughout the adjustment process, the control systems of both stations continuously monitor key variables such as DC bus voltage and output current to ensure that all operations are strictly within the safe operating range. After the adjustment is completed, the damping capability of the energy storage power station against system oscillations is significantly enhanced, and the dynamic coupling characteristics between the No. 3 photovoltaic power station and the grid are reshaped. The combined effect of both effectively suppresses the previously identified 2.5 Hz low-damped oscillation.

[0068] This application establishes a precise, reliable, and secure pathway from high-level coordination and decision-making to low-level control execution. By mapping abstract control function enhancement strategies to specific, operable control parameter adjustments, and supplementing this with a secure integration mechanism that combines incremental overlays or smooth switching, it ensures that every system-level optimization decision is seamlessly and smoothly translated into device-level control behavior. This not only achieves dynamic, online optimization of system stability but also fundamentally guarantees the reliability and robustness of the control process itself, avoiding secondary risks caused by parameter mutations. This enables online integrated stability control to be safely applied to actual operating power systems.

[0069] In this embodiment, the method is executed nested across multiple time scales: the first time scale is from milliseconds to seconds, where the local controllers of each grid-type converter perform internal control based on the current fixed parameters; the second time scale is from ten seconds to minutes, where steps S1 to S3 are executed to complete a round of system-level operational status assessment, coordination control quantity calculation, and online adjustment of control parameters, and the updated parameters are sent to the local controllers; the third time scale is above minutes, used for self-learning and adaptive correction of the weight coefficients and stability boundary criteria in the coordination strategy model.

[0070] In this embodiment, the method is nested across multiple time scales. This means that the complete stability control task is decomposed into three levels with different update and execution cycles according to the speed of its dynamic process, with each level performing its own function while working closely together. The first level, millisecond to second-level internal control, is performed by the converter's local controller. Based on currently given fixed parameters optimized at a slower time scale, it executes instantaneous control algorithms such as the current inner loop and voltage outer loop to cope with instantaneous disturbances in the power grid and ensure rapid tracking of the grid-connected current. The second level, ten-second to minute-level system-level coordination, is the execution cycle of the core steps S1 to S3 of this method. It is responsible for sensing the global system situation, making optimization decisions, and updating the control parameter set issued to all converters at a slower pace than local control. The third-level self-learning, which takes more than a minute, is an offline or quasi-online process that continuously learns and improves the second-level coordination strategy model itself. Based on long-term accumulated system operation and control effect data, it makes periodic or trigger-based adjustments to meta-parameters such as weight coefficients and stability boundary criteria within the model, enabling the entire coordination strategy to adapt to long-term, slow changes in power grid structure, load characteristics, and new energy penetration.

[0071] Its implementation principle is based on the decoupling and collaborative design of the dynamic time scale of the power electronic power system. The lowest layer, the first millisecond-level control, focuses on the speed of electromagnetic transient processes and equipment-level protection. Its control cycle matches the frequency of power electronic switching, ensuring the safe and stable operation of the equipment itself at any instant and providing a reliable and fast-response execution basis for the upper-level control. The middle second layer, the ten-second-minute-level coordination, addresses system-level electromechanical transients and small-to-medium-sized disturbance dynamic processes. The frequency of system oscillation modes is usually from a few tenths of a hertz to tens of hertz, and their period and dynamic process fall exactly on the order of seconds to minutes. Therefore, performing state assessment, optimization calculation, and parameter refresh with a period of ten seconds to minutes can capture these dominant dynamic modes, provide sufficient calculation time for optimization algorithms, and avoid interfering with the stable operation of the underlying fast control due to overly frequent parameter adjustments. The top third layer, the self-learning layer above minutes, handles slow dynamics such as system operating point migration, equipment aging, and seasonal changes. It utilizes historical big data and, through parameter identification or machine learning algorithms, analyzes which coordination strategies are more effective under different operating conditions. This allows for continuous fine-tuning of the objective function weights and stability margin thresholds in the optimization model used by the second layer, enabling the entire control system to possess adaptability and evolutionary capabilities. The three layers are connected through clear interfaces: the third layer outputs the updated strategy model to the second layer; the second layer outputs the optimized fixed parameter set to the first layer; and the first layer feeds back local measurement and status information to the second layer, forming a closed-loop, time-scale-separated intelligent control system.

[0072] Continuing with the previous example of the photovoltaic and energy storage cluster, when the system experiences a 2.5 Hz oscillation, three time scales operate simultaneously. At the first level (milliseconds), each photovoltaic inverter and energy storage converter continuously runs its internal control algorithm based on current virtual inertia, damping, and other parameters, responding instantaneously to voltage and frequency changes in the grid. At the second level (ten-seconds), the coordination controller initiates a new S1 to S3 process: it collects data from the past few seconds, analyzes and confirms that the oscillation still exists and the damping is insufficient, recalculates and decides to further optimize the virtual damping coefficient of the energy storage power station from 45 to 50, and issues the new parameters. At the third level (hourly scale), the self-learning module analyzes three similar oscillation events and control records within the past month, finding that such oscillations are prone to occur when the total photovoltaic output exceeds 80% and a specific line is lightly loaded. Therefore, it automatically corrects the coordination strategy model, and in the future, once similar operating conditions are detected, even if no obvious oscillation has been observed, it will preemptively slightly increase the weighting coefficient of the frequency damping function of the energy storage power station, thereby achieving preventative control.

[0073] This application, through a scientifically nested timescale design, perfectly balances the speed of control, the globality of optimization, and the adaptability of the strategy. This architecture ensures millisecond-level response capabilities for instantaneous device-level safety, achieves second-minute-level coordinated decision-making for dynamic system-level optimization, and endows the entire system with self-learning and evolutionary capabilities exceeding minutes in the face of long-term changes. This makes the proposed online integrated stability control method not a static algorithm, but a dynamic organism with hierarchical intelligence and continuous self-improvement capabilities, thus enabling it to robustly and efficiently address the complex and ever-changing stability challenges of new power systems in the long term.

[0074] In this embodiment of the application, the method further includes an online self-updating step for the coordination strategy model: continuously collecting system dynamic response data before and after each execution of steps S1 to S3 to construct a sample dataset; using the sample dataset, periodically or trigger-based updating of key parameters in the coordination strategy model through parameter identification, reinforcement learning, or neural network training methods, including the weight coefficients of the optimization objective function, the threshold in the stability margin assessment, and the parameters of the dynamic equivalent model of the grid-type converter.

[0075] In this embodiment, the online self-updating step of the coordination strategy model endows the entire control system with the ability to continuously learn and improve from historical experience. The sample dataset is a standardized data set consciously collected and stored by the system before and after each perception, decision-making, and closed-loop control operation. It typically includes the characteristics of the disturbance event, the content of the control instructions, and the dynamic response trajectory of the system after control. Parameter identification is a data-driven method that uses input and output data to inversely deduce the values ​​of unknown parameters in a mathematical model by minimizing prediction errors. Reinforcement learning is a machine learning paradigm that allows an agent to try different control strategies in a system environment and iteratively optimize its decision logic based on the obtained reward or penalty signals. Neural network training is the process of learning the complex nonlinear mapping relationship between input features and optimal output decisions using a large amount of sample data through deep neural network models. Periodic updates refer to retraining the model at preset fixed time intervals, such as daily or weekly; triggered updates are initiated after a specific event is detected, such as when the actual effect of a control operation deviates from the expected value by a certain range. The weighting coefficients of the optimization objective function determine the relative importance balance among multiple stable sub-objectives; the threshold in the stability margin assessment is the baseline for judging whether the system is in a critical state; the parameters of the dynamic equivalent model of the grid converter are key coefficients used to predict its simplified mathematical expression for responding to external commands.

[0076] Its implementation principle is based on a closed-loop learning and improvement framework. After each control action in steps S1 to S3, the system initiates a data recording process to completely save the system characteristics before the control event, the issued coordination control quantity, and the actual dynamic response of the system within a certain period after the control, forming a labeled sample. Over time, these samples are continuously accumulated, building an increasingly rich database. When model updates are needed, the system calls the corresponding learning algorithm to mine this dataset. For example, using parameter identification methods, the relationship between the actual output power change of the converter and the virtual inertia adjustment command can be analyzed in multiple control events, thereby correcting the gain coefficient of the inertia response in its dynamic equivalent model, making the model prediction closer to reality. Using reinforcement learning methods, the entire coordination decision-making process is modeled as a Markov decision process, with the improvement of the system stability margin as a reward. Through continuous trial and error and updates, the policy network from system state to control parameter allocation scheme is directly optimized. Neural network training can be used to establish a complex mapping from high-dimensional system features to optimal weight coefficients, replacing manually set fixed formulas. Updates can be periodic to ensure the model stays up-to-date, or they can be triggered by unusual events to enable targeted and rapid improvements.

[0077] Continuing with the example of the photovoltaic and energy storage cluster, after successfully suppressing the 2.5 Hz oscillation, the system automatically recorded the event: oscillation frequency and damping ratio, identified key equipment, and issued adjustment commands. Subsequent dynamic response data showed that the damping ratio recovered to 6%, but the reactive power output fluctuation of the energy storage station slightly increased. This data was packaged into a sample and stored in the historical database. A week later, a periodic self-learning task was initiated, analyzing dozens of similar samples accumulated recently. Parameter identification revealed that, under the current grid strength, the marginal benefit of the virtual damping coefficient of the energy storage station to system damping began to decline after reaching a certain value. Reinforcement learning evaluation found that when pursuing the maximization of the damping ratio, the weight coefficient for the voltage recovery target was too low, leading to reactive power fluctuations. Therefore, a self-update step was automatically executed: it fine-tuned the weight ratio of the frequency stability and voltage stability sub-objectives in the coordination strategy model and updated the parameters regarding the effective range of virtual damping in the equivalent model of the energy storage station. When a similar operating condition occurs again, the new model will automatically generate a more balanced and effective coordinated control quantity.

[0078] This application enables the entire online integrated stability control system to break through the limitations of relying on fixed prior knowledge and human experience, evolving into an intelligent agent with continuous evolution capabilities. By learning from each control practice, the system can continuously revise its internal world model and decision preferences, making the coordination strategy increasingly accurate in aligning with the dynamic characteristics of the actual power grid. This not only significantly improves the effectiveness and adaptability of control but also greatly reduces the dependence on the accuracy of the initial model and the long-term manual maintenance costs, laying a solid technical foundation for achieving the stable operation of a fully autonomous, adaptive high-proportion renewable energy power system.

[0079] In this embodiment, the process of adjusting the grid-type control parameters online is constrained by the safe operating range of parameters determined based on the thermal stress of the converter semiconductor devices, the DC bus voltage fluctuation range, the output current harmonic distortion rate, and the grid connection standards specified in the grid guidelines. Before adjusting the parameters, it is first verified whether the adjustment target value indicated by the coordinated control quantity is within the safe operating range of the corresponding converter. If not, the range boundary value or the safe value calculated according to the preset rules is used as the replacement.

[0080] In this embodiment, the core of the safety constraint mechanism for online adjustment of grid-type control parameters is to set necessary physical and standard boundaries for dynamic optimization, ensuring that all control actions are within the acceptable safety range for both the equipment and the power grid. Semiconductor device thermal stress refers to the effect of increased junction temperature caused by current and switching losses in power switching devices such as insulated-gate bipolar transistors (IGBTs) or silicon carbide metal-oxide-semiconductor field-effect transistors (SFETs). Excessive thermal stress can shorten device lifespan or even cause failures. The DC bus voltage fluctuation range is the upper and lower limits of the allowable voltage operation of the DC-side capacitors inside the converter. Exceeding this range may damage the capacitors or cause AC-side control instability. The output current harmonic distortion rate is a percentage indicator measuring the degree to which the converter's output current waveform deviates from a sine wave. Power grid guidelines strictly limit this to prevent pollution of power grid power quality. The parameter safe operating range is a numerical range pre-calculated or determined experimentally for each adjustable control parameter, taking into account all the above limitations. For example, virtual inertia may have an upper limit determined by the equipment's heat dissipation capacity, while the droop coefficient has a lower limit determined by the tolerance of DC voltage fluctuations.

[0081] Its implementation principle is based on an online safety verification and correction logic that precedes the parameter adjustment execution stage. First, the system maintains a dynamic or static database of safe operating ranges for each grid-connected converter. This range may be dynamically fine-tuned based on factors such as equipment operating temperature and the current value of the DC bus voltage. When the coordinated control quantity calculates the adjustment target value for a specific parameter of a converter, this target value is not immediately applied but is first sent to the safety verification module. The verification module queries the safe operating range corresponding to the parameter to determine whether the target value falls within the range. If the target value is safe, it is allowed. If the target value exceeds the safety boundary, the correction logic is initiated. A simple correction strategy is to use the range boundary value for truncation. For example, if the target virtual inertia requirement is 10 seconds, but the safety upper limit is 8 seconds, then 8 seconds is ultimately used. More complex trade-off calculations may consider multiple interrelated parameters. For example, when it is necessary to increase both virtual inertia and damping coefficient to enhance damping, if the inertia has reached its upper limit, the algorithm will increase the damping coefficient proportionally within the allowable range according to preset rules, or adjust other relevant parameters appropriately, so as to approach the original control target as closely as possible under safety constraints and achieve the best trade-off between safety and control performance.

[0082] Continuing with the previous example of photovoltaic and energy storage clusters, assuming the coordinated control calculations show that to better suppress oscillations, the virtual inertia of the energy storage power station needs to be significantly increased from the current 5 seconds to 12 seconds. Before execution, the local controller performs a safety check. It queries the safe operating range table under the current operating conditions, which indicates that considering the current radiator temperature and output level, the safe upper limit for the virtual inertia is 8 seconds. Simultaneously, the model predicts that forcibly setting it to 12 seconds could cause DC bus voltage fluctuations to exceed the allowable range during rapid frequency changes. Therefore, the check module determines that the target value of 12 seconds is unsafe. According to the preset trade-off rules, the system initially uses the boundary value of 8 seconds as a substitute. However, for further optimization, it initiates a rapid local optimization calculation, finding that while setting the virtual inertia to 8 seconds, increasing the virtual damping coefficient from 45 to 55 still achieves a comprehensive damping effect close to the original scheme within a safe range. Therefore, the final adjustment command is safely corrected to: virtual inertia adjusted to 8 seconds, and virtual damping coefficient adjusted to 55.

[0083] This embodiment incorporates an inherently safe safety barrier into the fast control channel that pursues optimal system-level stability. By transforming the physical limits of the equipment and grid connection standards into clearly defined parameter safety ranges, and by performing mandatory online verification and intelligent trade-off corrections on all control commands, it ensures that the aggressiveness and innovation of online integrated stability control do not come at the cost of sacrificing equipment lifespan, damaging power grid quality, or causing secondary faults. This fundamentally solves the safety and reliability challenges often faced by advanced control algorithms in practical engineering applications, making this method not only intelligent and efficient but also safe and reliable, providing a key foundation for its application in industrial fields.

[0084] See Figure 2 A second aspect of this application provides a novel online integrated stability control system for grid-type converters in power systems, comprising: The status information aggregation and processing unit 21 is used to execute step S1; Coordination decision-making and calculation unit 22 is used to execute step S2; The instruction generation and communication unit 23 is used to perform instruction generation and issuance in step S3.

[0085] Please see Figure 3 , Figure 3 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 3 The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to execute the novel online integrated stability control method for grid-connected converters in power systems.

[0086] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0087] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0088] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0089] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation methods described in any embodiment of the novel power system grid-type converter online integrated stability control method provided in the embodiments of this application, or they can execute the implementation methods of the electronic devices described in the embodiments of this application, which will not be repeated here.

[0090] In another embodiment of this application, an electronic device is provided. The electronic device stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the novel online integrated stability control method for grid-type converters in the above embodiments. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in an electronic device, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. Computer-readable media can include any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0091] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0092] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0093] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0094] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or units, or it may be an electrical, mechanical, or other form of connection.

[0095] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0096] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0097] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A novel online integrated stability control method for grid-type converters in power systems, characterized in that, Including the following steps: S1. Real-time acquisition of operating status information of multiple grid-connected converters; S2. Based on the operating status information, calculate online the coordination control quantity used to coordinate the control modes of each of the grid-type converters; S3. Based on the aforementioned coordinated control quantity, adjust the grid-type control parameters of the corresponding grid-type converter online.

2. The method according to claim 1, characterized in that, In step S1, the acquisition of the operating status information is a fusion acquisition of multi-source heterogeneous information, specifically including: acquiring the instantaneous values ​​and fundamental components of the three-phase voltage, three-phase current, and output power at the output point of each grid-type converter through the local voltage and current sensors of each grid-type converter; acquiring the grid common connection point voltage, frequency, and phase angle data containing grid connection point information of at least two grid-type converters through a wide-area measurement system or data concentrator; and acquiring the key control variables of other grid-type converters in adjacent or preset clusters through communication network interaction. The key control variables include virtual power angle, virtual frequency, active power reference value, and reactive power reference value.

3. The method according to claim 1 or 2, characterized in that, Step S1 also includes a running status analysis and feature extraction step: The collected operational status information is analyzed in the time and frequency domains to extract feature quantities that characterize the system's stable state. The characteristic quantities include at least: the estimated frequency and damping ratio of the dominant oscillation mode of the system, the amplitude and phase deviation vector between the output voltage of each grid converter and the voltage at the point of common coupling of the grid, and the equivalent parameters of the coupling impedance matrix or admittance matrix reflecting the active-reactive power interaction strength among the grid converter clusters.

4. The method according to claim 1, characterized in that, In step S2, the online calculation of the coordinated control quantity specifically includes the following sub-steps: S21: Based on the operating status information, assess the margin of the current system operating point relative to the preset stability boundary, and identify the key grid-type converters that play a dominant supporting role in system stability and have the potential to cause oscillation risks; S22: Based on the preset coordination strategy model containing multi-objective optimization functions, calculate the control function types and intensity weight allocation schemes that should be prioritized for each key grid-type converter. S23: Quantify the function type and intensity weight into a specific set of control parameter adjustment instructions to form the coordinated control quantity.

5. The method according to claim 4, characterized in that, The multi-objective optimization function in the preset coordination strategy model aims to enhance the overall small-disturbance stability of the system, improve the transient voltage / frequency recovery process, and optimize the power distribution among the grid-type converter clusters. It uses the physical limits, control bandwidth, and network security constraints of each grid-type converter as constraints and employs online rolling optimization or optimal control algorithms based on real-time linearization models to solve the problem, thereby dynamically generating the allocation scheme of the function type and intensity weight.

6. The method according to claim 4 or 5, characterized in that, In step S3, the online adjustment of the grid-type control parameters based on the coordinated control quantity is specifically implemented as follows: For grid-type converters assigned to enhance voltage support, priority is given to adjusting the reference value of their outer loop voltage controller, the reactive power-voltage droop coefficient in the droop coefficient, or the inductive component in the virtual impedance; for grid-type converters assigned to enhance frequency support and inertia response, priority is given to adjusting the virtual inertia, virtual damping coefficient, or active power-frequency droop coefficient in their virtual synchronous machine control; the adjustment is integrated into the converter's local control loop in a superimposed incremental or smooth switching manner.

7. The method according to claim 1, characterized in that, The method is executed nested across multiple time scales: the first time scale is from milliseconds to seconds, where the local controllers of each grid-type converter perform internal control based on the current fixed parameters; the second time scale is from ten seconds to minutes, where steps S1 to S3 are executed to complete a round of system-level operational status assessment, coordinated control quantity calculation, and online adjustment of control parameters, and the updated parameters are sent to the local controllers; the third time scale is above minutes, used for self-learning and adaptive correction of the weight coefficients and stability boundary criteria in the coordinated strategy model.

8. The method according to claim 5, characterized in that, The method further includes an online self-updating step for the coordination strategy model: continuously collecting system dynamic response data before and after each execution of steps S1 to S3 to construct a sample dataset; using the sample dataset, periodically or trigger-based updating of key parameters in the coordination strategy model through parameter identification, reinforcement learning, or neural network training methods, the key parameters include the weight coefficients of the optimization objective function, the threshold in the stability margin assessment, and the parameters of the dynamic equivalent model of the grid-type converter.

9. The method according to claim 6, characterized in that, The process of adjusting the online grid-type control parameters is constrained by the safe operating range of parameters determined based on the thermal stress of the converter semiconductor devices, the DC bus voltage fluctuation range, the output current harmonic distortion rate, and the grid connection standards specified in the grid guidelines. Before adjusting the parameters, it is first verified whether the adjustment target value indicated by the coordinated control quantity is within the safe operating range of the corresponding converter. If not, the range boundary value or the safe value calculated according to the preset rules is used for replacement.

10. A novel online integrated stability control system for grid-type converters in power systems, used to implement the method as described in any one of claims 1 to 9, characterized in that, include: A status information aggregation and processing unit is used to execute step S1; A coordination decision-making and calculation unit is used to execute step S2; The instruction generation and communication unit is used to perform instruction generation and issuance in step S3.

Images