New energy field station event-driven broadband oscillation suppression control method, system, equipment and medium

Abstract

The invention discloses a new energy station event-driven broadband oscillation suppression control method, system, equipment and medium, and belongs to the technical field of power systems and automation thereof, and the method comprises the steps: monitoring the operation quantity of a station in real time, sensing an abnormal trigger event signal, identifying an oscillation frequency band, outputting frequency band characteristics, and forming an optimal damping vector. And injecting the virtual damping signal according to the frequency band priority, carrying out oscillation control, judging the current oscillation control effect, entering a safety mode when detecting that communication is abnormal or overtime, and carrying out parameter rollback. The invention constructs a multi-layer intelligent control system integrating real-time event triggering, spectrum identification, self-learning optimization, coordinated injection and knowledge evolution. The method shows high stability and high robustness under various typical working conditions, the safety margin and the electric energy quality of a new energy grid-connected system can be improved, and a reproducible and extensible technical path is provided for autonomous operation of a large-scale new energy station in the future.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power systems and its automation technology, in particular to a new energy station event-driven wide frequency oscillation suppression control method, system, device and medium. BACKGROUND

[0002] The most widely used is the fixed parameter type virtual damping and virtual inertia control strategy. The advantage of this design is clear structure and easy to implement, but the parameters are often determined at the design stage and cannot be adjusted in real time according to the running state or external disturbance. When the system topology, load or grid impedance changes, the fixed parameters may cause the control characteristics to mismatch the actual needs, and the damping effect will be weakened, and even the oscillation may be amplified under certain working conditions.

[0003] Another idea is the damping method combining adaptive control with signal detection. It has better flexibility in theory, but in practical application, it is limited by the delay of the detection link, the uncertainty of the criterion threshold and the sampling synchronization error, etc. The response lag or false triggering problem often occurs. Especially when the oscillation frequency drifts quickly or the oscillation amplitude is small, the algorithm stability decreases significantly.

[0004] In recent years, some scholars have proposed event-triggered oscillation detection and local response mechanism, trying to trigger local control actions through rapid identification of abnormal events to reduce the calculation burden of continuous sampling. However, there are still deficiencies in system modeling and control coordination. For example, there is a lack of unified trigger management mechanism when multiple events occur simultaneously, leading to response conflict or control failure; for example, the control strategy is based on experience adjustment, and lacks targeted design of damping characteristics in different frequency bands.

[0005] There are still many deficiencies in the existing scheme that cannot be ignored: first, the real-time performance of oscillation detection and identification is insufficient. Most existing algorithms rely on continuous sampling and offline frequency domain operation, such as detection structures based on FFT or wavelet analysis. This kind of method is highly sensitive to data length and window size, and in actual application, the calculation accuracy and response time often need to be compromised. When the system disturbance is frequent or the oscillation frequency drifts quickly, the algorithm output often lags behind the physical process, causing delay in triggering damping injection. Some engineering projects reduce the sampling window to shorten the detection time, but at the expense of frequency resolution, which increases the probability of misjudgment. The resulting time effectiveness contradiction has become an important bottleneck restricting the rapid response of existing schemes.

[0006] Secondly, the virtual damping parameter lacks dynamic matching capability. At present, the damping or inertia parameters of most controllers are fixedly configured in the design stage. Even in some improved schemes, the parameter updating often depends on experience coefficients or fuzzy rules, rather than real-time system state quantities. When the grid impedance, load characteristics or the number of parallel devices changes, the damping effect is difficult to maintain at the optimal level. Excessive damping will cause system response delay, and too small damping will limit the suppression effect. More complex situation is that the sensitivity of oscillation modes in different frequency bands to damping amplitude is not consistent, and the existing algorithms often lack frequency band differentiation mechanism, so they cannot achieve unified stability control in the full frequency domain.

[0007] Thirdly, the control strategy and event triggering mechanism are not closely coupled. Many so-called "event-driven" schemes only stay at the threshold crossing triggering level, the triggering logic is single, and the parameter configuration depends on manual setting. There is no priority judgment and action coordination between different event types, which leads to false triggering or response conflict when multiple disturbances occur. For example, when there are low-frequency power angle oscillation and high-frequency control oscillation in the system, the controller can only respond to one of the modes, and the other oscillation is delayed or even completely ignored. This uncoordinated response makes it difficult to guarantee the global stability of the control strategy.

[0008] In addition, the balance between algorithm complexity and real-time implementation has not been solved. Some multi-frequency adaptive control schemes proposed in the literature can theoretically cover a wider frequency domain, but they require higher computing resources and communication bandwidth support, which is a heavy burden for embedded control platforms. In the scenario of limited hardware power, developers have to reduce the algorithm order or sampling rate to ensure operability, thereby sacrificing control accuracy and dynamic performance. Especially in wind farm group control or multi-terminal DC systems, this compromise often leads to different responses of different subsystems, causing new oscillation sources.

[0009] Finally, the existing research lacks a mechanism for collaborative optimization from the system level. Most schemes still focus on single-machine or local control, and lack overall strategies considering the interaction and group behavior between stations. When multiple converters independently execute their own damping control, parameter competition or phase cancellation may occur, leading to local stability but global instability. Such problems are more prominent under the condition of large-scale new energy access, and in severe cases, they can even induce secondary oscillation. SUMMARY

[0010] In view of the above problems, the present application is proposed.

[0011] Therefore, the present application aims to the existing vibration suppression and virtual damping control method in theory research and engineering application although each has breakthrough, but under the complex working condition still is difficult to consider real-time, self-adaptability and multi-frequency band stability. Therefore, the present application aims at the problems such as that identification of wide-frequency oscillation is not timely, damping parameter adjustment is not flexible and event trigger response is not coordinated in the wide-frequency oscillation suppression control method of new energy station event-driven wide-frequency oscillation suppression control method.

[0012] To solve the above technical problems, the present application provides the following technical scheme: a new energy station event-driven wide-frequency oscillation suppression control method, comprising, Real-time monitoring is carried out on the station operation quantity through high-frequency sampling and signal fusion, multi-dimensional event sensing is carried out, when an abnormality occurs, a comprehensive index is calculated and a trigger event signal is judged, the oscillation frequency band is identified for the trigger event signal, and the signal is identified, the frequency band is decomposed, and the frequency band characteristics are output; a virtual damping parameter set is updated in real time, the system response residual and the oscillation energy change trend are calculated, the virtual damping parameter is dynamically adjusted, the optimal damping vector is formed; after the optimal damping vector is obtained, the virtual damping signal is injected according to the frequency band priority, the oscillation control is carried out, the weight is automatically distributed for constraint; the current oscillation control effect is judged through the self-evolution strategy, the current strategy weight is adjusted, and the knowledge graph mapping is updated, forming a knowledge chain; data encryption transmission and access authority hierarchical management are carried out through a three-level communication architecture, and visual display is carried out in cooperation, when communication abnormality or timeout is detected, the parameter is rolled back.

[0013] As a preferred scheme of the new energy station event-driven wide-frequency oscillation suppression control method, wherein: the trigger event signal judgment comprises setting a synchronous sampling node, real-time monitoring is carried out on the key operation quantity of the station through high-frequency sampling and signal fusion, and pretreatment is carried out. Abnormality is judged through single fusion score, when an abnormality occurs, a comprehensive index is calculated, the trigger event signal is judged through hysteresis logic, an event identifier is output, and written into a buffer.

[0014] As a preferred scheme of the new energy station event-driven wide-frequency oscillation suppression control method, wherein: the output frequency band characteristics comprise, the trigger event signal in the trigger interval is subjected to short-time Fourier transform, time-frequency spectrum energy distribution is obtained, the time-frequency spectrum energy distribution is clustered and divided; Main frequency components are obtained, each independent oscillation component is extracted, the oscillation frequency band is identified, the oscillation frequency band is decomposed through energy ratio and frequency domain clustering method, is identified and classified into first, second and third modes, and the corresponding frequency band characteristics are output.

[0015] As a preferred scheme of the new energy station event-driven wide-frequency oscillation suppression control method, the optimal damping vector is formed by updating the virtual damping parameter set in real time according to the oscillation frequency band category and the current operating state vector, taking the equivalent model of the power as the optimization target, and solving the minimum oscillation energy residual. The parameter sensitivity is recursively corrected through the event history data, a frequency band weighting coefficient is introduced to form a comprehensive update index, the optimal damping parameter at the current time is output, and is injected into the multi-loop coordination.

[0016] As a preferred scheme of the new energy station event-driven wide-frequency oscillation suppression control method, the oscillation control includes injecting the virtual damping signal into the power loop, the voltage loop and the current loop according to the frequency band priority for multi-loop coordinated damping after the optimal damping parameter is obtained. A hierarchical coordination structure is adopted, the power loop provides low-frequency support, the voltage loop provides medium-frequency stability, the current loop processes high-frequency interference, and the weights are distributed in layers and adjusted in real time by the frequency energy distribution to determine the coordinated injection signal. A frequency band weight function is introduced to automatically distribute the injection force according to the frequency, and a voltage command is comprehensively output, a dynamic amplitude limiting and phase coordination mechanism is set, and the amplitude limiting value of the injected voltage changes proportionally with the voltage level and the damping power. When there are two or more oscillation modes, the execution order is determined according to the energy proportion and the frequency weight comprehensive index, the frequency band control quantity with the maximum energy proportion and frequency weight comprehensive index is executed first, and the remaining loops enter the following or damping mode; the coordinated injection signal is input into the corresponding control loop of the converter controller after being limited in amplitude and phase, the injection effect is detected, and the weight and the injection force amplitude are adjusted.

[0017] The preferred technical scheme in the embodiment of the application has the following beneficial effects: natural frequency domain decoupling is realized through the bandwidth difference of the three loops to prevent overcompensation and reverse oscillation, the priority decision mechanism ensures the order of response when multiple events are superimposed, and the self-learning module forms dynamic complementation to improve the long-term stability of the system; the virtual damping injection is expanded from single-loop local adjustment to multi-loop coordinated execution, realizing integrated operation of frequency division damping- energy coordination-closed loop feedback, which is an important execution layer support for realizing wide-frequency stability control.

[0018] As a preferred scheme of the new energy station event-driven wide-frequency oscillation suppression control method, the self-evolution strategy includes automatically recording the time sequence data including the trigger time, the frequency band category, the damping parameter, the injection power and the system recovery process after each oscillation control event, defining a comprehensive performance index, comparing the comprehensive performance index with the historical results, and automatically strengthening the corresponding strategy weight if the current control effect is better than the historical results. By comparing the difference between the actual damping parameter and the historical optimal parameter, the parameter deviation vector is calculated, and the corresponding performance deviation is stored at the same time, if the performance deviation is greater than 0, the current parameter is better than the historical optimal parameter, then trigger strategy reinforcement, if the performance deviation is less than 0, execute weight backoff, the reinforcement and backoff process is updated by exponential weighting; Each event is abstracted as a structured record node, including an event feature vector and a corresponding optimal parameter, a new event is associated with a historical event through a similarity measurement function, if the similarity is lower than a preset threshold, the corresponding historical parameter template is reused and locally corrected to form a multi-dimensional knowledge graph; According to the statistical characteristics of the knowledge graph, the global strategy is retrained, the long-term average score of each type of strategy is calculated by the weighted accumulation of performance indicators, the strategy with the highest average score is listed as the priority template, and the strategy with the lowest score is eliminated.

[0019] The preferred technical scheme in the embodiment of the application has the beneficial effects that: the experience of multiple events is explicit, which significantly shortens the parameter tuning time of a new scene; through performance-driven weight updating, long-term convergence of the algorithm is realized; automatic optimization can be achieved under different weather, grid strength and operation modes; the strategy library is dynamically updated and does not depend on manual setting, and has the ability of self-evolution; the leap from single event learning to cross-event knowledge evolution is realized, and the new energy station has the ability of long-term self-learning and self-correction.

[0020] As a preferred scheme of the new energy station event-driven wide frequency oscillation suppression control method, the three-level communication architecture includes a hierarchical security communication architecture, which is divided into a master control layer, a security gateway layer and an execution layer. All control commands and operation data need to be forwarded and verified through the security gateway. Operation data is transmitted in a transparent manner, and control commands are allowed to be executed after identity authentication and integrity verification. According to the use scene, users are divided into different levels, all access behaviors are recorded and have a traceability function, when detecting unauthorized or abnormal access, immediately issue a warning and automatically freeze the corresponding account. Continuously collect key operation quantities, and unify the display of multi-source data through a time synchronization mechanism. The visual interface adopts a hierarchical display method including a global situation view, an event view and a history playback view. When communication anomalies, data distortion, control conflicts or unauthorized operations are detected, the security mode is entered: the master control layer suspends the issuance of new commands, only local autonomous control is retained, and the virtual damping parameter is automatically rolled back to the scaled version of the last stable state to maintain the minimum stable margin: , wherein, is the damping parameter of the last normal operation, is a security scaling coefficient; an event report is automatically generated, including the oscillation type, the trigger reason, the control parameter change and the recovery process.

[0021] The beneficial effects of the preferred technical solution in the embodiments of the present invention are as follows: the intelligent control system not only has self-learning capabilities at the algorithm level, but also achieves autonomous protection and intelligent display at the security and interaction levels, ensuring that the algorithm layer and the human-machine interface are independent of each other, eliminating the risk of misoperation, improving the system's understandability and operation and maintenance efficiency, maintaining system autonomy and stability under abnormal working conditions, completing the closed loop from intelligent decision-making to safe execution, and realizing the coordinated coexistence of algorithms, communication and humans. It is a key supporting link to ensure the long-term reliable operation of the overall intelligent control system.

[0022] Another objective of this invention is to provide an event-driven wideband oscillation suppression and control system for new energy power stations.

[0023] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a new energy power station event-driven broadband oscillation suppression control system, comprising: a multi-dimensional event perception and anomaly triggering module, which monitors the power station's operating volume in real time through high-frequency sampling and signal fusion, performs multi-dimensional event perception, calculates comprehensive indicators and judges trigger event signals when an anomaly occurs; The spectrum decoupling and pattern recognition module identifies the oscillation frequency band of the trigger event signal, performs pattern recognition on the signal, decomposes the frequency band, and outputs the frequency band characteristics; The virtual damping parameter self-learning calculation module updates the virtual damping parameter set in real time. By calculating the system response residual and the trend of oscillation energy change, it dynamically adjusts the virtual damping parameters to form the optimal damping vector. The multi-loop coordinated injection and execution module obtains the optimal damping vector and then injects the virtual damping signal according to the frequency band priority to perform oscillation control and automatically allocates weights for constraints. The self-evolution strategy and knowledge mapping module uses the self-evolution strategy to determine the current oscillation control effect, adjust the current strategy weight, and update the knowledge graph mapping to form a knowledge chain. The security isolation and visualization collaboration module uses a three-level communication architecture to encrypt data transmission and manage access permissions hierarchically, and provides collaborative visualization. When a communication anomaly or timeout is detected, it enters a secure mode.

[0024] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the event-driven broadband oscillation suppression control method for new energy power stations.

[0025] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the event-driven broadband oscillation suppression control method for new energy power stations.

[0026] The beneficial effects of this invention: Compared with traditional vibration suppression methods for new energy power plants based on fixed damping parameters or single-loop control, this invention achieves significant improvements in control system architecture, algorithm mechanism, and system adaptive capability: ① It realizes a mechanism shift from "passive response" to "active prevention and control": Previous power plant control relied heavily on post-event response mechanisms, only passively adjusting after voltage or frequency anomalies, resulting in problems such as response lag and parameter mismatch. This invention, through event-driven and multi-source information fusion, can identify and trigger suppression actions in advance when oscillations just appear or even before they expand, achieving early intervention and rapid intervention. Under most simulation conditions, the total delay of event identification and response is controlled within 20-30 milliseconds, and the oscillation peak reduction rate exceeds 60%, significantly better than conventional delayed response control. ② It proposes a multi-frequency band hierarchical coordinated virtual damping control framework: Traditional virtual damping designs are mostly aimed at a single frequency band or a single element. This invention, for the first time, combines the virtual damping injection strategy with multiple control loops (power loop, voltage loop, current loop) to form a frequency-domain coordinated control architecture. This design not only ensures decoupling of control loops at different frequency bands, but also achieves coordinated energy absorption when low-frequency and high-frequency oscillations coexist. Simulation results show that within a wide frequency range of 1–10 Hz, the system damping ratio is improved by approximately 40%–50%, and the injected energy accounts for only 2% of the rated capacity, resulting in a significant improvement in control efficiency. ③ It possesses online self-learning and parameter self-optimization capabilities: Unlike traditional damping control that relies on manual tuning or fixed parameter tables, this invention introduces a lightweight self-learning mechanism that can real-time correct the parameter set based on oscillation characteristics and operating conditions. The adaptive algorithm uses performance metrics As a feedback variable, the system gradually approaches the optimal control solution while maintaining system stability. After multiple rounds of event verification, the average parameter tuning convergence time is less than 0.2 seconds, and the performance remains consistent after multiple disturbances, requiring no external intervention. This characteristic enables the control system to continuously evolve, automatically optimizing control performance according to changes in the power grid structure or the proportion of new energy sources. ④ An event-based knowledge-driven self-evolving control system is established: This invention not only focuses on the effectiveness of individual control actions but also archives the spectral characteristics, parameter selection, and control results of each event, forming a structured knowledge graph. When similar oscillations recur, the system can directly reuse the optimal control template without recalculation, significantly shortening the response time. This mechanism allows the system to gradually form statistical correlations between "events—strategies—results," achieving experience accumulation and autonomous optimization. Compared with memoryless control, this invention improves the response speed of similar events by approximately 35%, demonstrating significant learning and transfer capabilities. ⑤ Enhanced system security and engineering adaptability: Considering power grid communication delays and network security factors, this invention incorporates multi-level security isolation and parameter rollback mechanisms at the control execution layer. When communication is abnormal or latency exceeds limits, the control system will automatically enter autonomous mode, and parameters will be scaled up to safe values. It still maintains the necessary vibration suppression function. Furthermore, the system can seamlessly integrate with existing master stations or scheduling systems, requiring no modification to the original hardware structure for implementation, resulting in low engineering implementation costs and strong compatibility. ⑥ It achieves visualization and interpretability of the operation process: Unlike traditional "black box" control, this invention graphically displays key variables, oscillation spectra, parameter evolution, and control output processes. Maintenance personnel can view real-time changes in system oscillation energy, injected power trajectory, and strategy switching times, facilitating analysis of control effects and improving algorithm interpretability. This visualization design has practical value for the operation and maintenance, scheduling optimization, and subsequent algorithm verification of new energy power plants. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 The above is a flowchart of an event-driven broadband oscillation suppression control method for new energy power stations, provided as an embodiment of the present invention. Detailed Implementation

[0029] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0030] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a method for suppressing and controlling event-driven broadband oscillations in new energy power stations, comprising: S100: Real-time monitoring of station operation volume through high-frequency sampling and signal fusion; Multi-dimensional event perception; When an anomaly occurs, calculate comprehensive indicators and determine the trigger event signal. S200 identifies the oscillation frequency band of the trigger event signal, performs pattern recognition on the signal, decomposes the frequency band, and outputs the frequency band characteristics; S300: Real-time update of virtual damping parameter set; dynamically adjust virtual damping parameters by calculating the system response residual and oscillation energy change trend to form the optimal damping vector; S400: After obtaining the optimal damping vector, the virtual damping signal is injected according to the frequency band priority for oscillation control, and weights are automatically assigned for constraints. S500 uses a self-evolutionary strategy to determine the current oscillation control effect, adjusts the current strategy weights, and updates the knowledge graph mapping to form a knowledge chain. The S600 uses a three-level communication architecture to encrypt data transmission and manage access permissions hierarchically, and provides a visual display. When a communication anomaly or timeout is detected, it enters a safe mode and rolls back the parameters. It should be noted that existing technologies have varying degrees of shortcomings in terms of event triggering logic, adaptive damping parameters, cross-frequency band coordination, and engineering feasibility. Especially when the system's operating state changes rapidly or multiple types of oscillations coexist, traditional solutions often struggle to maintain sufficient robustness and coordination while ensuring real-time performance. Therefore, there is an urgent need for a novel control architecture that combines event recognition, dynamic parameter adjustment, and wide-frequency band collaborative suppression to achieve efficient and stable operation of new energy power stations under complex conditions.

[0031] Therefore, to address the aforementioned problems, through steps S100-S600, the event-driven intelligent control system for new energy power plants proposed in this invention constructs an integrated control framework of "event triggering—frequency domain coordination—virtual damping self-learning" to address issues such as diverse oscillation modes, lag in the response chain, and insufficient parameter adaptation under high-proportion new energy grid connection conditions. The system abandons the traditional approach based on a single predictive model or fixed parameter control, instead using events as the center, virtual damping injection as a means, and self-learning optimization as the evolution mechanism to achieve rapid autonomy and wide-frequency vibration suppression at the power plant level.

[0032] Example 2, refer to Figure 1 This is one embodiment of the present invention, which provides a method for suppressing and controlling event-driven broadband oscillations in new energy power stations, comprising: In this embodiment of the invention, S100 performs real-time monitoring of the station's operational volume through high-frequency sampling and signal fusion, and performs multi-dimensional event perception. When an anomaly occurs, a comprehensive index is calculated and a trigger event signal is determined, including the following steps S101-S102: S101. Set up synchronous sampling nodes to monitor key operational quantities of the station in real time through high-frequency sampling and signal fusion, and perform preprocessing. S101 is used to quickly sense and trigger anomalies of key electrical quantities during the grid-connected operation of the substation, serving as the entry point for subsequent frequency division identification and damping injection; In an embodiment of the present invention, key operational quantities of the station are monitored in real time and preprocessed, including the following steps A1-A2: A1. Arrange synchronous sampling nodes between the main control layer and the grid-connected converter to collect the station's operating quantities, including voltage (u), current (i), active / reactive power (P / Q), and phase angle difference (Δφ). A2. Lightweight preprocessing is performed on the edge side, including noise reduction, amplitude limiting, and missing measurement compensation.

[0033] In an optional implementation, the preprocessing in S101 can use asynchronous sampling nodes instead of synchronous sampling nodes to collect station operation data and align the data using software timestamps; lightweight preprocessing is performed at the edge side, using moving average filtering for noise reduction, fixed threshold limiting for amplitude limitation, and nearest neighbor interpolation for missing data compensation. However, this implementation will introduce time synchronization errors.

[0034] In another alternative implementation, the preprocessing in S101 can also be performed by centrally arranging sampling nodes at the main control layer (instead of dispersing them among the grid-connected converters) to collect the station's operating data and transmit the data using the existing communication network; the preprocessing is completed at the main control layer, using simple low-pass filtering for noise reduction, dynamic range limiting for amplitude limitation, and previous value holding for missing measurement compensation; however, this implementation will increase communication latency.

[0035] S102. Anomalies are identified through a single fusion score. When an anomaly occurs, a comprehensive index is calculated, and an event signal is triggered through hysteresis logic. An event identifier is output and written to a buffer. To avoid complex and computationally intensive detection algorithms, this invention constructs only a single fusion score and triggers it in conjunction with a hysteresis threshold; Criteria include: mutation rate (First-order difference in power or voltage), phase shift (PMU phase difference or equivalent frequency offset), target frequency band energy (Sliding window energy or bandpass filter output amplitude), fusion score (Convex combination): in, , , These represent the weights assigned to the criteria, and the hysteresis trigger criterion, where... : When a criterion is more stable (i.e., less volatile) in the most recent small window, its weight is slightly increased; if the volatility increases, it is appropriately decreased. The adjustment is performed smoothly within preset upper and lower limits with a fixed step size to avoid frequent reordering. A slight increase is made when there is a weak grid or multiple inverters are connected in parallel. Slightly increase when wind power fluctuates significantly. When signs of narrow-band oscillation appear, moderately increase , , The hysteresis threshold is set according to the station noise level and scheduling strategy. =0.7, = -0.1 or 0.2; When the fusion score exceeds the limit and is triggered, the module outputs an event identifier and brief accompanying information (including the trigger time, dominant criterion and preliminary frequency band indication), and writes the most recent sampling points into the buffer for direct use by the spectrum decoupling and pattern recognition module. This invention does not undertake complex diagnosis, and ensures that the anomaly is captured in a timely manner with minimal computational overhead, and the false trigger rate is controlled.

[0036] In this embodiment of the invention, step S200 involves identifying the oscillation frequency band of the trigger event signal, performing pattern recognition on the signal, decomposing the frequency band, and outputting the frequency band characteristics, including the following steps S201-S203: After completing multi-dimensional event perception and triggering, the next stage is spectrum decoupling and pattern recognition. The task of this invention is to perform frequency domain analysis and oscillation mode determination on the signal segments corresponding to the triggered events, thereby providing targeted frequency band information and control basis for subsequent damping injection. Signal decoupling is achieved using a "time-frequency joint analysis + adaptive filtering decomposition" approach. Unlike traditional fixed filtering or static Fourier transform, this invention introduces a short-time-window sliding strategy to continuously track instantaneous frequency characteristics while retaining real-time requirements. Specifically: In an embodiment of the present invention, S201, performing a short-time Fourier transform on the trigger event signal within the trigger interval to obtain the time-spectrum energy distribution includes the following steps B1-B2: B1. Perform a short-time Fourier transform on the trigger event signal within the trigger interval to obtain the time-spectrum energy distribution; specifically, analyze the voltage or power signal near the event trigger time. Perform a short-time Fourier transform to obtain its time-frequency energy distribution. : in, A time window function is used to limit the integration range; the energy density is calculated. =∣ |2, Identify the concentration of oscillation energy in different frequency bands to reduce computational burden; B2. The present invention adopts a sliding window update mechanism. The window length and update step size can be adaptively adjusted according to the event type. For example, a longer window is used when there is low-frequency oscillation, and the window length is shortened when there is high-frequency disturbance to improve the time resolution.

[0037] In an optional implementation, the time-frequency energy distribution obtained in S201 can be based on time-frequency energy analysis of wavelet transform. Continuous wavelet transform is performed on the trigger event signal within the trigger interval. A predefined wavelet basis function is selected, and wavelet coefficients are calculated through scaling and translation to obtain the energy distribution of the signal in different frequency bands and times. The time-frequency energy distribution is obtained by calculating the square of the modulus of the wavelet coefficients. A sliding window update mechanism is adopted, and the window length is adaptively adjusted according to the event type.

[0038] In another optional implementation, the time-frequency energy distribution obtained in S201 can also be based on the frequency band energy extraction of the filter bank, designing a set of fixed-bandwidth bandpass filters to cover the target frequency band, performing parallel filtering on the trigger event signal to obtain the signal components of each frequency band; calculating the short-time energy of the output signal of each filter to form the frequency band energy distribution; and using a sliding window to update the energy calculation, the window length of which can be adjusted manually or by simple rules according to the event type.

[0039] S202. Cluster the time-frequency energy distribution to obtain the main frequency components, extract each independent oscillation component, and identify the oscillation frequency band. To clarify the oscillation modes, the frequency domain energy sequence was clustered. The main analysis frequency band was set at 0.1–60 Hz, and the power spectral density was used for analysis. Peak search can yield a set of several main frequency points. To prevent misjudgment of noise spikes, a smoothing threshold method is used for peak detection. ,in, and These are the mean and standard deviation of the power spectral density, respectively. This is an empirical coefficient (generally taken as 2 to 3); The selected frequency points are mapped into low-frequency bands (<2Hz), mid-frequency bands (2–15Hz), and high-frequency bands (>15Hz) according to frequency ranges. This segmentation method not only facilitates the subsequent targeted damping injection by the controller, but also significantly improves the noise immunity of the identification.

[0040] After obtaining the main frequency components, the phase relationship, energy decay rate, and signal coupling characteristics of each component are further extracted, indicating that the target signal is the sum of each independent oscillation component: in For component amplitude, The attenuation coefficient is... Let n be the phase, n be the total number of frequency points, and i be the variable index. Belongs to the main frequency point set middle.

[0041] According to each component and The system can determine the type of oscillation: if the frequency is low and the decay is slow, it is mostly power angle or power-type oscillation; if the frequency is medium and the energy is concentrated in the converter control bandwidth, it is mostly control-coupled oscillation; if the frequency is high and the decay is extremely fast, it is mostly transient oscillation caused by sampling or communication noise.

[0042] To prevent misclassification, an energy proportion criterion is introduced to assist decision-making, that is, when the energy of a certain frequency band... The proportion of total energy exceeds the set threshold At that time, this frequency band was identified as the dominant oscillation region: threshold It can be set according to operating experience, and is generally taken as 0.25 to 0.35.

[0043] S203. By using the energy ratio and frequency domain clustering method, the oscillation frequency band is decomposed, and the signal is divided into the first mode, the third mode, and the frequency band characteristics are output. After spectrum decoupling is completed, the module output includes: oscillation frequency and corresponding energy percentage; frequency band category (low, medium, high frequency) and mode label; recommended damping injection path indication information; specifically, the signal is divided into: the first mode is low frequency (power angle type), the second mode is medium frequency (control interaction type), and the third mode is high frequency (local circulation type). These results are passed down in the form of structured data and synchronously written into the policy library for subsequent self-learning optimization. Compared with traditional fixed filtering or single spectrum methods, the time-frequency joint decoupling method of this invention can significantly improve the recognition accuracy while maintaining real-time performance. Through the comprehensive use of sliding window update, smooth threshold detection and energy ratio criterion, the system still has stable recognition capability under low signal-to-noise ratio conditions. This invention can be implemented on an embedded controller.

[0044] In an embodiment of the present invention, the virtual damping parameter set is updated in real time in S300. By calculating the system response residual and the trend of oscillation energy change, the virtual damping parameters are dynamically adjusted to form the optimal damping vector, including the following steps S301-S302: After spectrum decoupling and pattern recognition are completed, the virtual damping parameter self-learning calculation stage is entered. Based on the real-time oscillation mode and operating status, the optimal combination of virtual damping parameters is calculated so as to achieve rapid and targeted energy injection in the subsequent execution layer. Unlike the traditional approach of setting parameters through offline tuning or manual experience, this invention automatically updates the damping parameters to achieve adaptive and evolvable control characteristics.

[0045] S301. Based on the oscillation frequency band category and the current operating state vector, update the virtual damping parameter set in real time, and use the equivalent model of damping power as the optimization objective to minimize the oscillation energy residual. The system first determines the oscillation frequency band category (low frequency, medium frequency, or high frequency) and the current operating state vector based on the spectrum identification results: in, , , , , These represent active power, reactive power, voltage amplitude, phase difference, and equivalent impedance at the grid connection point, respectively; t is the time variable; the current operating state vector serves as the input to the self-learning model for online calculation of the virtual resistance. Virtual conductivity and virtual inertia Three types of parameters; to balance stability and dynamic response speed, the system uses damping power The equivalent model is the optimization objective: The optimization objective reflects the contribution relationship between the three injection mechanisms—virtual resistance, conductance, and inertia—to the system's energy absorption and power regulation. The objective of parameter solving is to minimize the oscillatory energy residual under constraints. in For the identified oscillatory active component, the objective function J represents the uncompensated part of the oscillation power; the smaller the value, the better the damping matching. and The oscillation frequency band is the start and end time; i(t) is the grid-connected current, and u(t) is the bus voltage. Angular frequency, i.e., the rate of change of frequency.

[0046] In an embodiment of the present invention, S302, dynamically adjusting the virtual damping parameters to form the optimal damping vector, includes the following steps C1-C5: C1. The parameter sensitivity is recursively corrected by using historical event data and a frequency band weighting coefficient is introduced to form a comprehensive update index. The optimal damping parameter at the current moment is output and injected into the multi-loop coordination. C2. A lightweight self-learning strategy is adopted, which recursively corrects the parameter sensitivity using historical event data, and the currently estimated parameters are... The update rate can be expressed as: ,in The learning rate coefficient can be adaptively adjusted according to the event type. To prevent over-correction from causing control oscillations, the module introduces constraints: , and These are the lower and upper limits of the parameters. In this way, parameter updates are kept within a safe range, avoiding sudden changes.

[0047] C3. Because the sensitivity of oscillations in different frequency bands to parameters varies significantly, frequency band weighting coefficients are introduced when solving the system. This forms a comprehensive update index, yielding the optimal damping parameter vector for the current moment. : in These represent the parameter sets corresponding to low-frequency, mid-frequency, and high-frequency oscillations, respectively.

[0048] C4. The frequency band weights can be adjusted in real time according to the energy ratio or event level, so that the algorithm responds faster to the dominant frequency band and remains stable to the secondary frequency band.

[0049] C5. After the calculation is completed, the module outputs the optimal damping parameter vector at the current moment. The system then sends the results to the multi-loop coordination injection module for execution. Simultaneously, it feeds back the event status, oscillation decay effect, and final J value to the self-learning storage unit for continuous updates to the strategy library. Once multiple event results converge, the system automatically generates parameter mapping rules, enabling online self-optimization without manual intervention.

[0050] In an optional implementation, the data encryption transmission in S302 can be a fixed parameter fixed key symmetric encryption transmission. During system initialization, a fixed symmetric encryption key is preset and stored in the secure storage area of ​​the main control layer and the grid-connected converter. After the optimal damping parameter vector is output, the parameter vector is encrypted using the fixed key to generate a ciphertext data packet. The ciphertext data packet is then sent to the multi-loop coordinated injection module through the station's internal network. The receiving end uses the same fixed key to decrypt the ciphertext, recover the parameter vector, and perform damping injection.

[0051] In another optional implementation, the data encryption transmission in S302 can also be performed by calculating a simple hash value of the parameter vector after outputting the optimal damping parameter vector for single linear identification, using it as a check code, packaging the parameter vector and hash value into a data frame, and transmitting it to the multi-loop coordination injection module through the station's internal network. The receiving end recalculates the hash value of the parameter vector and compares it with the received hash value. If they match, the parameter is accepted; otherwise, it is discarded and a retransmission is requested.

[0052] Compared with control methods based on fixed parameters or single linear identification, the update law of this invention only requires first-order gradient calculation, making it suitable for embedded processing; the parameters can be dynamically adjusted according to changes in power grid strength and operating status; interval constraints and weighted smoothing are introduced to avoid control overshoot; the learning process gradually converges in the accumulation of events, continuously improving the control effect; and the virtual damping parameter is transformed from static setting to dynamic self-learning, which is the core bridge for the entire system to achieve intelligent and self-optimizing control.

[0053] In this embodiment of the invention, after obtaining the optimal damping vector in S400, the virtual damping signal is injected according to the frequency band priority for oscillation control, and weights are automatically assigned for constraints, including the following steps S401-S404: Obtaining the optimal parameter vector of virtual damping Then, the system enters the execution phase. The function of this module is to allocate different types of virtual damping injection signals to the current loop, voltage loop, and power loop according to their frequency band characteristics and control priorities, achieving multi-loop coordinated regulation. Its goal is to ensure rapid vibration suppression while avoiding mutual interference or phase cancellation between multiple control loops.

[0054] S401. After obtaining the optimal damping parameters, the virtual damping signal is injected into the power loop, voltage loop and current loop according to the frequency band priority to perform multi-loop coordinated vibration suppression. S402. A hierarchical coordination structure is adopted, with the power loop providing low-frequency support, the voltage loop providing mid-frequency stability, and the current loop handling high-frequency interference. The weights are allocated hierarchically and adjusted in real time by the frequency band energy distribution to determine the coordinated injection signal. A hierarchical coordination structure is adopted, and the responsibilities of each loop are as follows: the current loop is responsible for the current balance and suppression of high-frequency (>15Hz) disturbances and has the fastest response; The voltage loop should be designed to handle control-coupled oscillations in the mid-frequency (2–15Hz) range; The power loop handles low-frequency (<2Hz) power angle oscillations and power swings.

[0055] The three components are structurally a combination of serial and parallel components, with their weights adjusted in real time by the frequency band energy distribution. The coordinated injection signal is uniformly represented as: in, The optimal virtual resistance, For optimal virtual conductance, The optimal virtual inertia is determined. This coordinated injection signal is superimposed as a correction factor onto the adjustment signal of the corresponding control loop, with different superposition weights for different frequency bands, to achieve stratified energy absorption.

[0056] S403 introduces a frequency band weighting function to automatically allocate the injection force according to the frequency, integrates the output voltage command, sets a dynamic limiting and phase coordination mechanism, and the limiting value of the injection voltage changes proportionally with the voltage level and damping power. To ensure decoupling between the control loops, the system introduces a frequency band weighting function. Its function is to automatically distribute the injection force according to the frequency, and the comprehensive output voltage command is: in, include , , , , , These are the respective weights included in the frequency band weighting function. , , The outputs of the power loop, voltage loop, and current loop are respectively, and their sum forms the final injection control command. The weighting function satisfies the normalization condition. , Low-pass type is acceptable. Take the bandpass type, By adopting a high-pass filter, frequency band separation can be naturally achieved; this ensures that the control action at the same frequency point comes from only one dominant loop, avoiding competition between control loops.

[0057] To prevent excessive injection or control oscillations under large disturbances, a dynamic limiting and phase coordination mechanism is set up, and the limit value of the injected voltage varies proportionally with the system voltage level and damping power: in The rated voltage is α, which is a proportional coefficient ranging from 0.1 to 0.3. For real-time damping power, This is the rated power of the station. Simultaneously, the phase of the injected signal is constrained to ensure that the virtual damping component and the oscillation component remain approximately out of phase (180°±15°) to maximize energy absorption.

[0058] S404. When there are two or more oscillation modes, the execution order is determined based on the comprehensive index of energy ratio and frequency weight, the injection effect is detected, and the weight and injection force amplitude are adjusted. When multiple oscillation modes coexist (e.g., low-frequency power angle fluctuations superimposed with mid-frequency controlled oscillations), the module uses a comprehensive index based on energy proportion and frequency weight. Determine the execution order: in, For the energy of this frequency band, The dynamic weights assigned to the system The total oscillation energy. When The control input with the largest frequency band is executed first, while the remaining loops enter either follow or suppress mode. This strategy enables the system to achieve clear primary and secondary control logic in complex scenarios with multiple superimposed oscillation sources, avoiding overcompensation or reverse cancellation.

[0059] After amplitude limiting and phase processing, the coordinated injection signals are input to each control loop of the converter controller. The actual injected voltage, current, and power response signals are sampled in real time by the monitoring unit and fed back to the self-learning module for performance evaluation and parameter correction. If insufficient injection effect or excessive phase drift is detected, the module automatically adjusts the weights of each loop. Alternatively, the injection force can be reduced to prevent over-response. The entire execution process is completed within a millisecond-level closed loop, ensuring that control actions are synchronized with event changes.

[0060] In this embodiment of the invention, step S500 involves determining the current oscillation control effect through a self-evolutionary strategy, adjusting the current strategy weights, and updating the knowledge graph mapping to form a knowledge chain. This includes the following steps S501-S504: S501. After each control event, the system automatically records relevant timing data, including trigger time, frequency band type, damping parameters, injected power, and system recovery process. To quantify control effectiveness, a comprehensive performance index is defined. : in: , These represent the initial amplitude of the oscillation and the residual amplitude after decay, respectively. This is the time required for the system to recover to a steady state. Phase shift before and after the event; , , These are weighting coefficients used to balance the effects of decay rate, time, and phase parameters; The reference time required for the system to recover to steady state. The normalization factor is π (3.14159...).

[0061] The larger the value, the more ideal the control effect. After the event ends, the system compares this value with historical results and uses it as a basis for strategy adjustment. If the current control effect is better than the historical results, the corresponding strategy weight will be automatically strengthened. S502, the system compares the actual damping parameters in this instance. Compared with historical best parameters The difference is used to calculate the parameter deviation vector: At the same time, the corresponding performance deviation Stored together. If the system detects... If this indicates that the current parameters are better than the historical best, then policy reinforcement is triggered; if If so, weight rollback is executed. The reinforcement and rollback process is completed through exponential weighted updates: Where λ∈(0,1) is the forgetting factor. This represents the historical average performance metric. This weight update mechanism allows for smooth adjustment of policy preferences across multiple events, enabling the system to gradually accumulate experience.

[0062] S503. To enable experience reuse, this module abstracts each event into a structured record node, including an event feature vector. With corresponding optimal parameters The system uses a similarity measurement function. Linking new events with historical events: in, As feature weights, higher weights are typically assigned to frequency and impedance features. The dominant frequency represents the main frequency component of the oscillation. The dominant frequency represents the main frequency component of the oscillation. For equivalent impedance, For network strength, For the feature vector of historical events, and Let represent the i-th feature value of a new event and a historical event, respectively, and z be the total number of events.

[0063] If the similarity is below a preset threshold, the system will reuse the corresponding historical parameter template and make local corrections to avoid repeated learning and improve response speed. As events accumulate, a multi-dimensional knowledge graph is formed to map the intrinsic relationship between "scene features - control strategies - expected effects".

[0064] S504: Based on the statistical characteristics of the knowledge graph, the module periodically retrains the global strategy. Through weighted accumulation of performance metrics, it calculates the long-term average score for each strategy type. : in, The number of events for the current strategy. Let j be the performance metric for the j-th event, where j is the variable index; By prioritizing strategies with higher average scores and appropriately eliminating poorly performing old strategies, the evolutionary characteristic of "survival of the fittest" is maintained in the strategy library. This can be regarded as an implementation of lightweight reinforcement learning, which retains the ability to optimize in the long term while avoiding the computational burden of high-complexity algorithms.

[0065] Finally, the updated policy mapping table and weight distribution matrix are output and passed back to the self-learning parameter module. When a new round of events is triggered, the policy parameters with the best recent performance will be given priority to shorten the calculation time and improve stability. In addition, the knowledge graph can be exported as a structured file periodically for analysis and review by the upper-level scheduling system or external control center.

[0066] In an embodiment of the present invention, S600 uses a three-level communication architecture for encrypted data transmission and hierarchical access control, along with a visual display. When a communication anomaly or timeout is detected, a secure mode is entered, and parameters are rolled back, including the following steps S601-S605: As the final layer of protection and interactive interface for system operation, S600 aims to ensure the safe, stable, monitorable and controllable operation of the entire intelligent control system in complex power grid environments; it constructs a complete security and collaboration system from five aspects: communication security, access control, operation monitoring, visualization display and anomaly protection.

[0067] In an embodiment of the present invention, S601, data encryption transmission is performed through a three-level communication architecture, including the following steps D1-D2: D1. A layered secure communication architecture is adopted, which is divided into three parts: the main control layer, the security gateway layer, and the execution layer. All control commands and running data must be forwarded and verified through the security gateway to prevent unauthorized external access or command injection. D2. Uplink operating data (measurement, status) is transmitted transparently, and downlink control commands can only be executed after identity authentication and integrity verification. To further prevent communication interference, redundant link switching and encrypted transmission mechanisms are supported to ensure safe operation even when communication fluctuates or some links fail.

[0068] In an optional implementation, the encrypted data transmission in S601 can simplify the communication architecture to two layers: a control layer and an execution layer, eliminating the independent security gateway layer. However, an encryption module is built into the control layer. Before all data transmission, dynamic encryption is performed according to the role permissions of the data sender and receiver. A pre-shared key is used to encrypt control commands and running data. When the execution layer device receives data, it verifies the legality of the encryption key. Only when the key matches will the command be executed. A heartbeat mechanism is used to monitor the status of the communication link. When a communication abnormality is detected, it automatically switches to a backup communication channel, which also uses encrypted transmission.

[0069] In another alternative implementation, the encrypted data transmission in S601 can also retain a three-tier communication architecture, but the security gateway layer focuses on time synchronization and dynamic key generation; the master control layer and execution layer devices keep time synchronized with the security gateway, and each communication session uses a timestamp-based dynamic encryption key; uplink data is decrypted and re-encrypted at the security gateway before being forwarded, and downlink commands are also encrypted and verified using dynamic keys; if communication times out or key verification fails, the system falls back to a predefined static encryption mode to ensure basic communication security.

[0070] S602. Based on the usage scenario, users are divided into different levels such as operation monitoring personnel, maintenance personnel, and scheduling management personnel. The permissions of each level of user are strictly limited to the corresponding scope. For example, the monitoring layer can view real-time data but has no control permissions, and the maintenance layer can perform parameter calibration but cannot directly issue control commands. All access activities are recorded and traceable. When unauthorized or abnormal access is detected, the system will immediately issue an alert and automatically freeze the relevant account.

[0071] S603: The monitoring terminal continuously collects key operational data and displays multi-source data in a unified manner through a time synchronization mechanism. The visualization interface adopts a layered display method. The global situation view is used to display the overall operating status and oscillation spectrum distribution of the site; The event view displays the entire process of triggering, identifying, controlling, and recovering; The history playback view supports event reconstruction and strategy review.

[0072] S604. When communication anomalies, data distortion, control conflicts, or unauthorized operations are detected, the system will automatically enter safe mode. In safe mode, the main control layer suspends the issuance of new instructions, retains only local autonomous control, and the virtual damping parameters automatically revert to the scaled version of the most recent stable state to maintain minimum stability margin. ,in, These are the damping parameters from the last normal operation. The scaling factor is set to 0.8 for safety. This method can maintain the basic vibration suppression capability of the system in the state of disconnection or failure, and ensure that the local area of ​​the station still has self-stabilizing characteristics and will not cause the operation to run out of control due to the interruption of external communication.

[0073] S605, security isolation not only ensures the security of communication links, but also provides real-time decision support to operation and maintenance personnel through a visualization platform.

[0074] The system automatically generates event reports, including oscillation type, triggering cause, changes in control parameters, and recovery process, and displays them in the form of charts and summaries.

[0075] Maintenance personnel can directly confirm policies, compare parameters, and trace faults on this interface without having to access the underlying control module, thereby reducing the risk of human error.

[0076] In addition, the system supports interoperability with the upper-level dispatch center, enabling cross-site coordination control and information linkage at a higher level.

[0077] In summary, this invention establishes an event-driven broadband oscillation suppression method capable of cross-frequency band identification, self-learning optimization, and rapid response in complex operating environments. By organically integrating the four stages of detection, judgment, calculation, and control, this invention can complete oscillation identification and damping injection within milliseconds, providing a new stable control approach for new energy power plants that combines intelligence, speed, and universality.

[0078] Example 3: An embodiment of the present invention provides a method for suppressing and controlling event-driven broadband oscillations in new energy power stations. To verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0079] Experiment 1: Event-Driven Broadband Oscillation Active Suppression Control Based on Wind-Solar-Storage Hybrid Power Station: This experiment verifies the engineering feasibility and control effect of the "Event-Driven Broadband Oscillation Suppression Control Method for New Energy Power Stations with Virtual Damping Injection Function" proposed in this invention. The test object is a typical wind-solar-storage hybrid new energy power station, whose topology, equipment parameters, and operating conditions are as follows: 1. System Configuration and Operating Conditions: The experimental subject is a 500MW-level renewable energy power station in a region of Northwest China, consisting of three parts: a wind farm with an installed capacity of 300MW, using 150 2MW doubly-fed induction generator (DFIG) units; a photovoltaic (PV) farm with an installed capacity of 150MW, using string inverters (1.5MW capacity per unit); and an energy storage system with a capacity of 50MW / 100MWh, employing lithium battery energy storage units connected to the same 220kV busbar. The power station transmits power to the 500kV main grid through a ±320kV bipolar flexible DC system (VSC-HVDC). The station is equipped with 8 PMUs and 4 IED devices, with a sampling frequency of 10kHz and a timing accuracy of ±5µs. The short-circuit ratio (SCR) at the grid connection point is 3.5, and the system exhibits significant weak grid characteristics. The system operates under typical midday high-power conditions: wind power output of 280MW, PV power of 130MW, and the energy storage system in a 15MW charging state.

[0080] 2. Disturbance and Event Detection: At the 35th second of the simulation, a short-term voltage drop occurred on the 500kV bus of the external power grid (from 1.0 pu to 0.88 pu, lasting 160ms). The event sensing module calculates the voltage change rate and power deviation rate in real time. Adaptive threshold The event was identified as a disturbance. A trigger signal was transmitted to the main control layer within 18ms. Within 30ms, the spectrum decoupling module calculated the spectral energy distribution based on the short-time Fourier transform (STFT) and detected a significant coexistence of a mid-frequency component (6.7Hz) and a low-frequency oscillation component (1.5Hz). This characteristic was labeled as a "compound oscillation event," simultaneously activating both the mid-frequency and low-frequency control branches.

[0081] 3. Virtual damping parameter self-learning and optimization calculation: The control system calls the similar event template (number #A312) in the knowledge base. The template parameters are: Based on real-time measurements and error vector: The system calculates the objective function: The optimal parameters are obtained through gradient descent and memory update: This parameter is calculated within the average convergence step of the past 10 events (approximately 180ms) and then executed after confirmation by the main control layer.

[0082] 4. Multi-loop coordinated injection execution: Current loop (high frequency band >10Hz) injection of virtual admittance Achieving rapid damping of high-frequency oscillations; virtual damping is injected into the voltage loop (mid-frequency band 2–10Hz). Phase difference between control loops; virtual inertia injected into the power loop (low frequency band <2Hz). This enhances inertia support and smooths power output. The injected signal is standardized as follows: Weight allocation through multi-ring coordination: The total injected power peaked at 7.8MW, accounting for 1.6% of the rated power. Within 0.48s, the voltage recovered to 0.985pu, the oscillation amplitude decreased by 70%, and the main frequency energy decreased by 22dB.

[0083] 5. Self-evolution and performance recording: After the event ends, the system calculates performance metrics: This represents a 31% improvement over the traditional fixed virtual damping strategy. The event was automatically numbered #A341 and compared with template #A312, resulting in a performance improvement of ΔQ = +0.09. The system updates the event node weights accordingly. A new knowledge graph mapping entry is formed: "Oscillation type: Composite (1.5Hz + 6.7Hz); Operating condition: High power operation; Optimal damping parameters: [0.93, 0.16, 0.08]; Expected damping time: 0.5s" 6. Safety and Visual Verification: During control execution, the system detected that the communication link latency increased from 18ms to 24ms, which did not exceed the threshold. The safety isolation module maintained the normal channel without triggering a rollback. Maintenance personnel can view the following curves and charts in real time through the visualization interface: bus voltage time-domain waveform and spectrum; injected power and energy distribution of each control loop; real-time evolution trend of the performance index Qperf; and automatically generated event diagnostic reports (including trigger time, frequency band classification, parameter curves, and recovery time).

[0084] 7. Results and Analysis: Control results show that the system voltage recovery time was shortened from 1.35s to 0.49s; the peak power fluctuation decreased from ±8% to ±2.1%; the frequency deviation decreased from 0.32Hz to 0.09Hz; and the overall damping ratio increased by approximately 45%. Experimental verification demonstrates that the method of this invention can achieve millisecond-level event response, online parameter optimization, and multi-loop coordinated control under weak grid conditions, while exhibiting high stability and repeatability. Compared with the traditional "fixed damping + central dispatch control," this method shows significant advantages in response speed, energy utilization, and disturbance rejection performance, making it suitable for widespread application in large-scale wind-solar-storage integrated power plants.

[0085] Experiment 2: Verification of Dynamic Coordination and Vibration Suppression with Energy Storage: This experiment uses a new energy power station with energy storage as the object to verify the dynamic coordination performance of the event-driven virtual damping control method proposed in this invention under complex disturbances. This embodiment specifically considers the rapid power support characteristics of the energy storage unit and the nonlinear coupling effect of wind power and photovoltaic output.

[0086] 1. System Composition and Operating Conditions: This embodiment constructs a simulation model based on actual power plant data. Total installed capacity: 400MW (250MW wind power + 120MW photovoltaic + 30MW / 60MWh energy storage); Grid connection: 220kV bus connected to a ±320kV flexible DC system, transmitting power to the 500kV public grid; Sampling frequency: 20kHz, timing accuracy ±2µs; Grid short-circuit ratio (SCR): 3.0, typical weak grid; Operating conditions: 230MW wind power output, 100MW photovoltaic output, and energy storage operating in 10MW discharge mode. The simulation platform is built using PSCAD / EMTDC, with a time step of 25µs. The model includes 14 control nodes and 2 independent converter control units.

[0087] 2. Disturbance Event and Spectrum Identification: At 20 seconds into the simulation, a single-phase ground fault occurs on transmission line L1 (48km long), lasting 120ms. The bus voltage momentarily drops to 0.86 pu, recovering after the fault is cleared. The multi-dimensional event sensing module calculates the voltage change rate and power surge: The test result exceeded the dynamic threshold. =0.1, triggering the control strategy. Subsequently, the spectrum decoupling module uses a 64ms sliding window STFT to perform time-frequency transformation on the sampled signal to obtain the dominant frequency component: This indicates that the system exhibits both low-frequency power angle oscillations and mid-frequency control-interactive oscillations.

[0088] 3. Virtual damping parameter self-learning calculation: The system calculates based on event feature vectors. Match the historical event template (number #B204) in the knowledge graph to extract initial parameters. Gradient updates are performed using a real-time error function. After 8 iterations (approximately 160ms), the system converges to the optimal parameters: ; 4. Coordinated Injection and Energy Storage Participation in Control: The control module of this invention superimposes the virtual damped injection signal onto the power, voltage, and current three-loop control quantities. Weights are assigned based on the real-time frequency component energy ratio: =0.35:0.5:0.15; Meanwhile, the energy storage unit adjusts according to the main frequency. =2.1Hz dynamic power output adjustment, the control objective is: in =35 MW / Hz. Within 0.6s after control injection, the voltage recovered to 0.98pu, the oscillation energy decreased by 67%, and the peak power output of the energy storage was 12MW, accounting for 3% of the system capacity.

[0089] 5. Performance Evaluation and Strategy Evolution: Comprehensive System Computational Performance Indicators: Take weight =0.5, =0.3, =0.2, calculation result =0.91. Compared to the original fixed parameter control scheme. Compared to the previous version, this represents an improvement of approximately 26.4%. Subsequently, the self-evolution module updates the strategy weights based on performance changes. The system generates a new event node #B205 and updates the mapping in the knowledge graph: "Event Type: Dual-Frequency Composite". " 6. Security Protection and Rollback Verification: At the 25th second of the simulation, the communication link latency briefly increased to 32ms (threshold is 30ms). The security isolation module automatically triggered "local autonomous mode," scaling the damping parameters proportionally. The system completed parameter switching within 0.4 seconds, with no significant fluctuations in control. After communication was restored, the parameters automatically returned to normal operating conditions. The comparison results are shown in Table 1. Table 1 Comparison Results

[0090] 8. Results: Experiment 2 fully verified the effectiveness of the present invention under conditions of multiple disturbances, multi-frequency band coupling, and energy storage participation. The control system can respond to events in milliseconds, realizing adaptive virtual damping injection, multi-loop coordinated execution, and dynamic energy storage support, significantly improving the system stability and disturbance rejection capability of new energy power plants. Meanwhile, the safety rollback mechanism and knowledge graph update mechanism ensure long-term operational reliability and sustainable optimization capabilities.

[0091] Example 4 is an embodiment of the present invention. The above is a schematic scheme of an event-driven broadband oscillation suppression control method for new energy power stations. It should be noted that the technical solution of an event-driven broadband oscillation suppression control system for new energy power stations belongs to the same concept as the technical solution of the event-driven broadband oscillation suppression control method for new energy power stations described above. Details not described in detail in the technical solution of the event-driven broadband oscillation suppression control system for new energy power stations in this embodiment can be found in the description of the technical solution of the event-driven broadband oscillation suppression control method for new energy power stations described above.

[0092] This embodiment provides an event-driven broadband oscillation suppression and control system for new energy power stations, including: a multi-dimensional event perception and abnormal triggering module, a spectrum decoupling and pattern recognition module, a virtual damping parameter self-learning calculation module, a multi-loop coordinated injection and execution module, a self-evolutionary strategy and knowledge mapping module, and a security isolation and visualization collaboration module. The multi-dimensional event perception and anomaly triggering module first monitors multi-dimensional quantities such as voltage, current, power, and frequency in real time through high-speed sampling and edge computing nodes. A joint criterion is constructed using mutation rate, phase drift, and energy distribution characteristics to automatically trigger events when out-of-limit features or potential oscillation signs are detected. Compared to traditional timed sampling, this module replaces sampling frequency with event density, significantly reducing computational burden and enabling millisecond-level response to potential oscillations.

[0093] The spectrum decoupling and pattern recognition module automatically enters the frequency domain analysis phase after an event is triggered. Through joint time-frequency analysis and an improved adaptive filter, it performs frequency division and decoupling on the oscillation signal, enabling the identification of three typical modes: low-frequency (power angle type), mid-frequency (power control type), and high-frequency (current loop coupling type). Unlike traditional fixed threshold judgments, this module can dynamically adjust the recognition boundary based on grid impedance, power fluctuation amplitude, and control loop bandwidth, thereby achieving adaptive frequency division recognition.

[0094] The virtual damping parameter self-learning calculation module is the core of the system's computation. By modeling the current operating state of the power station (such as equivalent impedance matrix, power change rate, and oscillation energy attenuation characteristics) and combining historical event samples, it uses reinforcement learning and gradient constraint algorithms to solve for the optimal set of virtual damping parameters in real time. Unlike traditional empirical setting methods, this module can automatically balance the proportions of virtual resistance, virtual conductance, and virtual inertia based on the system's frequency characteristics, achieving dynamic matching and cross-frequency band adaptive adjustment.

[0095] The multi-loop coordinated injection and execution module, after parameter calculation, selects the injection path based on the oscillation frequency band and risk level: low-frequency disturbances are injected with virtual inertia through the power loop, mid-frequency disturbances are superimposed with virtual damping through the voltage loop, and high-frequency disturbances are embedded with virtual conductance in the current loop, achieving layered energy absorption. To avoid interference between multiple loops, this module introduces a "priority discrimination and amplitude limiting modulation" mechanism, enabling multiple control paths to run in parallel without conflict, ensuring smooth damping energy injection and phase stability.

[0096] The self-evolution strategy and knowledge mapping module automatically performs quantitative analysis of the control effect after each event, extracts the event fingerprint (including frequency, amplitude, phase, decay time constant, etc.), and performs similarity matching with the historical event database. Through incremental learning and strategy transfer mechanisms, a knowledge base based on a ternary mapping of "event type - control strategy - effect feedback" is gradually formed, enabling the system to have continuous learning and self-evolution capabilities, and optimize performance without relying on external manual tuning.

[0097] The security isolation and visualization collaboration module is responsible for controlling access permissions, data isolation, and status display of the control links to ensure safe deployment and operation. The system achieves local autonomy and information interference prevention from the upper-level scheduling system through a communication isolation gateway. Simultaneously, it provides a multi-dimensional visualization interface at the monitoring layer, displaying real-time oscillation spectrum, injected power, damping contribution, and system stability margin. This module ensures the verifiability and controllability of the algorithm in actual field operations.

[0098] Through the coordinated operation of the above six modules, this invention constructs an intelligent control system that combines event response, frequency band hierarchical management, and self-learning optimization, achieving a technological leap from "feedback suppression" to "feedforward intervention." The system can not only quickly identify and inject damping at the initial stage of oscillation, but also achieve self-evolution of the control strategy after multiple runs, thereby significantly improving the dynamic stability, disturbance rejection capability, and operational self-consistency of new energy power plants.

[0099] This embodiment also provides an electronic device applicable to a new energy power station event-driven broadband oscillation suppression control method, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the new energy power station event-driven broadband oscillation suppression control method proposed in the above embodiment.

[0100] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a new energy power station event-driven broadband oscillation suppression control method as proposed in the above embodiment.

[0101] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for implementing event-driven broadband oscillation suppression control of new energy power stations proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0102] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0103] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for suppressing and controlling event-driven broadband oscillations in new energy power plants, characterized in that: include, Real-time monitoring of station operation volume is achieved through high-frequency sampling and signal fusion, and multi-dimensional event perception is performed. When an anomaly occurs, comprehensive indicators are calculated and trigger event signals are determined. The system identifies the oscillation frequency band of the trigger event signal, performs pattern recognition on the signal, decomposes the frequency band, and outputs the frequency band characteristics. The virtual damping parameter set is updated in real time, and the virtual damping parameters are dynamically adjusted by calculating the system response residual and the trend of oscillation energy change to form the optimal damping vector; After obtaining the optimal damping vector, the virtual damping signal is injected according to the frequency band priority for oscillation control, and weights are automatically assigned for constraints. The current oscillation control effect is judged by the self-evolution strategy, the current strategy weight is adjusted, and the knowledge graph mapping is updated to form a knowledge chain; Data encryption and access control are managed through a three-tier communication architecture, with integrated visualization. When a communication anomaly or timeout is detected, the system enters a secure mode and rolls back the parameters.

2. The event-driven broadband oscillation suppression and control method for new energy power stations as described in claim 1, characterized in that: The determination of the trigger event signal includes setting up a synchronous sampling node, using high-frequency sampling and signal fusion to monitor key operational quantities of the station in real time, and performing preprocessing. Anomalies are identified by a single fusion score. When an anomaly occurs, a comprehensive index is calculated, and an event signal is triggered by hysteresis logic. The event identifier is then output and written to the buffer.

3. The event-driven broadband oscillation suppression and control method for new energy power stations as described in claim 2, characterized in that: The output frequency band characteristics include performing a short-time Fourier transform on the trigger event signal within the trigger interval to obtain the time-frequency energy distribution, and then performing clustering on the time-frequency energy distribution; The main frequency components are obtained, each independent oscillation component is extracted, the oscillation frequency band is identified, and the oscillation frequency band is decomposed by energy ratio and frequency domain clustering method. The oscillation frequency band is identified and classified into the first, second and third modes, and the corresponding frequency band characteristics are output.

4. The event-driven broadband oscillation suppression and control method for new energy power stations as described in claim 3, characterized in that: The process of forming the optimal damping vector includes updating the set of virtual damping parameters in real time according to the oscillation frequency band category and the current operating state vector, and minimizing the oscillation energy residual by taking the equivalent model of damping power as the optimization objective. The parameter sensitivity is recursively corrected using historical event data, and a frequency band weighting coefficient is introduced to form a comprehensive update index. The optimal damping parameter at the current moment is then output and injected into the multi-loop coordination.

5. The event-driven broadband oscillation suppression and control method for new energy power stations as described in claim 4, characterized in that: The oscillation control includes obtaining the optimal damping parameters and then injecting the virtual damping signal into the power loop, voltage loop and current loop according to the frequency band priority to perform multi-loop coordinated oscillation suppression. A hierarchical coordination structure is adopted, with the power loop providing low-frequency support, the voltage loop stabilizing the mid-frequency, and the current loop handling high-frequency interference. The weights are allocated hierarchically and adjusted in real time by the frequency band energy distribution to determine the coordinated injection signal. A frequency band weighting function is introduced to automatically allocate the injection force according to the frequency, and the output voltage command is integrated. A dynamic limiting and phase coordination mechanism is set, and the limiting value of the injection voltage changes proportionally with the voltage level and damping power. When there are two or more oscillation modes, the execution order is determined by the comprehensive index of energy ratio and frequency weight. The frequency band control quantity with the largest comprehensive index of energy ratio and frequency weight is executed first, and the remaining loops enter the follow or suppression mode. After the coordinated injection signal is limited and phase processed, it is input to the corresponding control loop of the converter controller to detect the injection effect and adjust the weight and injection force amplitude.

6. The event-driven broadband oscillation suppression and control method for new energy power stations as described in claim 5, characterized in that: The self-evolutionary strategy includes automatically recording time-series data, including trigger time, frequency band category, damping parameters, injected power, and system recovery process, after each oscillation control event ends, defining a comprehensive performance index, comparing the comprehensive performance index with historical results, and automatically strengthening the corresponding strategy weight if the current control effect is better than the historical results. By comparing the actual damping parameters with the historical best parameters, the parameter deviation vector is calculated. At the same time, the corresponding performance deviation is stored synchronously. If the performance deviation is greater than 0, the current parameters are better than the historical best, and the strategy reinforcement is triggered. If the performance deviation is less than 0, the weight rollback is executed. The reinforcement and rollback processes are updated through exponential weighting. Each event is abstracted into a structured record node, including the event feature vector and the corresponding optimal parameters. New events are associated with historical events through a similarity measurement function. If the similarity is lower than a preset threshold, the corresponding historical parameter template is reused and local corrections are made to form a multidimensional knowledge graph. Global strategies are retrained based on the statistical characteristics of the knowledge graph. The long-term average score of each strategy is calculated by weighted accumulation of performance indicators. The strategy with the highest average score is listed as the priority template, and the strategy with the lowest score is eliminated.

7. The event-driven broadband oscillation suppression and control method for new energy power stations as described in claim 6, characterized in that: The three-tier communication architecture includes a layered secure communication architecture, divided into three parts: the main control layer, the security gateway layer, and the execution layer. All control commands and operational data must be forwarded and verified through the security gateway. Operational data is transmitted transparently, while control commands are allowed to be executed after identity authentication and integrity verification. Users are divided into different levels according to the usage scenario. All access behaviors are recorded and have traceability functions. When unauthorized or abnormal access is detected, an alert is immediately issued and the corresponding account is automatically frozen. Key operational volumes are continuously collected, and multi-source data is displayed uniformly through a time synchronization mechanism. The visualization interface adopts a layered display method, including a global situation view, an event view, and a historical playback view. When communication anomalies, data distortion, control conflicts, or unauthorized operations are detected, a safe mode is entered: the master control layer suspends the issuance of new instructions, retains only local autonomous control, and the virtual damping parameters automatically revert to a scaled version of the most recent stable state to maintain minimum stability margin. ,in, These are the damping parameters from the last normal operation. Set a safe scaling factor; automatically generate event reports, including oscillation type, trigger cause, control parameter changes, and recovery process.

8. A new energy power station event-driven broadband oscillation suppression control system, employing the new energy power station event-driven broadband oscillation suppression control method as described in any one of claims 1 to 7, characterized in that, include: The multi-dimensional event perception and anomaly triggering module monitors the station's operational volume in real time through high-frequency sampling and signal fusion, performs multi-dimensional event perception, calculates comprehensive indicators and determines the trigger event signal when an anomaly occurs; The spectrum decoupling and pattern recognition module identifies the oscillation frequency band of the trigger event signal, performs pattern recognition on the signal, decomposes the frequency band, and outputs the frequency band characteristics; The virtual damping parameter self-learning calculation module updates the virtual damping parameter set in real time. By calculating the system response residual and the trend of oscillation energy change, it dynamically adjusts the virtual damping parameters to form the optimal damping vector. The multi-loop coordinated injection and execution module obtains the optimal damping vector and then injects the virtual damping signal according to the frequency band priority to perform oscillation control and automatically allocates weights for constraints. The self-evolution strategy and knowledge mapping module uses the self-evolution strategy to determine the current oscillation control effect, adjust the current strategy weight, and update the knowledge graph mapping to form a knowledge chain. The security isolation and visualization collaboration module uses a three-level communication architecture to encrypt data transmission and manage access permissions hierarchically, and provides collaborative visualization. When a communication anomaly or timeout is detected, it enters a secure mode.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the event-driven broadband oscillation suppression control method for new energy power stations as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the event-driven broadband oscillation suppression control method for new energy power stations as described in any one of claims 1 to 7.

Images