Self-adaptive layered suppression system and method for energy storage broadband oscillation of high-voltage cascade valve hall

Abstract

The invention discloses a high-voltage cascade valve hall energy storage broadband oscillation self-adaptive layered suppression system and method, and the system comprises a system layer controller, a valve hall layer controller and a unit layer controller which communicate through a dual-redundancy Ethernet. The system layer controller achieves the global oscillation monitoring and task distribution based on the electrical quantity of a grid-connected point. A valve hall layer controller monitors the state of the power unit, calculates the optimal control parameter and coordinates the action of the unit; and the unit layer controller receives the parameters, adaptively adjusts and executes control. The method comprises the steps that oscillation modes and risks are recognized through multi-domain detection; a Gaussian distribution function distributes a valve hall task weight; performing pole assignment to optimize PI parameters; performing feedforward compensation and PWM optimization execution suppression; the effect is evaluated regularly and adjusted dynamically, and reference parameters are recovered gradually after oscillation is eliminated. According to the invention, the layered architecture of the high-voltage cascade energy storage system is fully utilized, a three-level cooperative control strategy is designed, control parameters of each layer are adaptively adjusted according to oscillation characteristics, and rapid and accurate suppression of broadband oscillation is realized.

Description

Technical Field

[0001] This invention relates to the field of energy storage system technology, and in particular to a broadband oscillation adaptive hierarchical suppression system and method for high-voltage cascaded valve hall energy storage. Background Technology

[0002] With the increasing penetration of new energy sources and the large-scale application of power electronic equipment, the dynamic characteristics of power systems have undergone fundamental changes. High-voltage cascaded energy storage systems, as a representative of new energy storage technologies, play a crucial role in grid frequency regulation, peak shaving, and new energy consumption. However, because energy storage converters employ a large number of power electronic devices, the interaction between their rapid control response and grid impedance can easily lead to broadband oscillations.

[0003] High-voltage cascaded energy storage technology offers significant advantages in large-scale energy storage power plants, primarily in terms of efficiency, economy, and safety. Compared to low-voltage energy storage, high-voltage cascaded systems eliminate the need for step-up transformers and can be directly connected to high-voltage power grids (such as 10kV, 35kV, and 110kV voltage levels), greatly reducing system complexity and operating costs. High-voltage cascaded valve hall energy storage systems employ multiple power units connected in series. Each power unit includes an energy storage battery pack and a power converter, achieving high-voltage output through cascading. This topology inherently possesses hierarchical characteristics, providing the hardware foundation for hierarchical control.

[0004] Wideband oscillation has become a key factor restricting the safe and stable operation of high-voltage cascaded energy storage systems. The oscillation frequency range is extensive, from low-frequency oscillations of a few Hz to high-frequency oscillations of thousands of Hz, with different oscillation mechanisms and influencing factors at different frequency bands. Traditional oscillation suppression methods are insufficient to cover such a wide frequency range and cannot fully utilize the hierarchical architecture advantages of high-voltage cascaded systems.

[0005] Currently, the technical solutions for suppressing broadband oscillations in energy storage systems mainly have the following shortcomings: Simple control architecture: Existing technologies mostly adopt centralized control or simple partition control, which cannot fully utilize the hierarchical architecture characteristics of high-voltage cascaded systems. All control decisions are concentrated in one controller, resulting in slow response speed and the risk of single point of failure. The parameter adjustment lacks flexibility: Existing methods usually adopt a fixed parameter switching method, that is, two or more sets of control parameters are preset, and switching between different parameter groups occurs when oscillation occurs. This method cannot adapt to complex and ever-changing power grid conditions and has limited suppression effect. The system hardware characteristics need to be changed: Some methods suppress oscillations by changing the system impedance characteristics, which requires additional hardware or complex impedance reshaping algorithms, increasing system cost and control complexity; The suppression methods are limited and affect power output: Some methods change the system characteristics by blocking part of the converter. Although they can suppress oscillations, they will reduce the power output capability of the energy storage system and affect the normal operation of the system. Limited frequency coverage: Existing technologies are often designed for specific frequency bands, making it difficult to suppress low-frequency, mid-frequency, and high-frequency oscillations simultaneously, and lacking a unified solution for the entire frequency band.

[0006] Therefore, there is an urgent need for a collaborative suppression technology that is compatible with the hierarchical architecture of high-voltage cascaded systems, can adaptively adjust parameters, and can cover the entire frequency band of oscillations, so as to achieve fast and accurate suppression of wideband oscillations while ensuring the system's power output capability. Summary of the Invention

[0007] The purpose of this invention is to provide a broadband oscillation adaptive hierarchical suppression system and method for high-voltage cascaded valve hall energy storage. It makes full use of the hierarchical architecture of the high-voltage cascaded energy storage system, designs a three-level collaborative control strategy of system layer, valve hall layer and unit layer, and adaptively adjusts the control parameters of each layer according to the oscillation characteristics to achieve rapid and accurate suppression of broadband oscillation.

[0008] To achieve the above objectives, the present invention provides the following solution: A high-pressure cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression system includes: a system layer controller, a valve hall layer controller, and a unit layer controller. The controllers at each layer are connected via dual redundant Ethernet communication, and the communication protocol adopts the IEC61850 standard. The system-level controller is equipped with a high-speed data acquisition unit to collect three-phase voltage, three-phase current, active power, reactive power and frequency data at the grid connection point, and run real-time oscillation monitoring algorithm and task allocation algorithm to realize global oscillation monitoring, task allocation and coordination. The number of valve hall level controllers is consistent with the number of valve halls. Each valve hall level controller is electrically connected to multiple power units and is used to monitor the operating status of the power units in the valve hall, calculate the optimal control parameters according to the task instructions issued by the system level controller, and coordinate the control actions of each power unit in the valve hall. The unit-level controller corresponds one-to-one with the power unit. It is used to receive control parameters sent by the valve hall-level controller, perform adaptive parameter adjustment, execute current inner loop control, voltage outer loop control and phase-locked loop control operations, and provide feedback on the operating status of the corresponding power unit.

[0009] Furthermore, the high-speed data acquisition unit of the system layer controller adopts synchronous sampling technology to achieve time alignment of each phase data; the real-time oscillation monitoring algorithm adopts sliding window FFT analysis technology; and the task allocation algorithm adopts weighted optimization algorithm.

[0010] Furthermore, the power unit operating status monitored by the valve hall controller includes DC side voltage, AC side current, power unit temperature, switching device status, and battery SOC.

[0011] This invention also provides a method for adaptive hierarchical suppression of broadband oscillations in high-voltage cascaded valve hall energy storage, applied to the aforementioned high-voltage cascaded valve hall energy storage system, comprising the following steps: S1, Oscillation Detection and Identification: The system-level controller continuously monitors the electrical quantities at the grid connection point through a multi-domain detection method that combines time domain, frequency domain, and time-frequency domain. After detecting oscillation, it extracts multi-dimensional feature vectors, uses an SVM classifier to identify the oscillation mode, and uses an ARIMA model to predict the oscillation trend and assess the risk level. S2, Layered Task Allocation: The system layer controller calculates the task weight of each valve hall based on the oscillation frequency, amplitude and risk level using a Gaussian distribution function, generates a task allocation matrix and sends it to the valve hall layer controller. S3, Control Parameter Optimization: The valve hall controller calculates the PI control parameters of each power unit based on the task instructions issued by the system layer controller and the local power unit operating status, and corrects the parameters through a multi-unit coordination strategy. S4, Cooperative suppression execution: The unit-level controller receives the control parameters sent by the valve hall-level controller, and after a smooth switch, executes the PI control algorithm. It actively cancels the oscillation component through feedforward compensation control and optimizes the PWM modulation strategy. S5, Effect Evaluation and Iteration: The system-level controller evaluates the oscillation suppression effect every set time period, dynamically adjusts task allocation and control parameters based on the evaluation results, and gradually restores the baseline parameters according to an exponential law after the oscillation is eliminated.

[0012] Furthermore, in S1, the multi-domain detection method specifically includes: Time-domain detection: Using a detection window of a set length, the standard deviation of voltage and current within the window is calculated. If the standard deviation exceeds a set threshold, it is determined that oscillation may exist. Frequency domain detection: Perform FFT transformation on the acquired signal, analyze the spectral distribution, and determine that oscillation exists when the ratio of the amplitude of the non-fundamental frequency component to the amplitude of the fundamental frequency exceeds a set value; Time-frequency domain detection: Morlet wavelet is used for time-frequency analysis to identify the oscillation start time, frequency changes and amplitude evolution characteristics; In S1, the multidimensional feature vector includes: Time-domain characteristics: peak value, mean, standard deviation, skewness, kurtosis; Frequency domain characteristics: dominant frequency, spectral width, spectral centroid, spectral energy; Time-frequency characteristics: instantaneous frequency, modulation depth, and energy distribution.

[0013] Furthermore, in S2, the task weights are calculated using the following Gaussian distribution function:

[0014] in, These are the normalization coefficients; The detected oscillation frequency; For the first The center frequency of each valve chamber For the first The frequency distribution width of each valve chamber; When multiple oscillations are detected simultaneously, the system-level controller independently calculates a weight matrix for each oscillation frequency, and then performs a comprehensive analysis based on the weighted sum of the weights of each frequency.

[0015] Furthermore, in step 3, the PI control parameters are calculated using the pole placement method, as shown in the following formula:

[0016]

[0017] in, This is the proportionality coefficient. The integral coefficient is... For the damping ratio, Where is the natural frequency, L is the filter inductance, and R is the equivalent resistance.

[0018] Furthermore, the multi-unit coordination strategy in S3 is as follows: the parameter difference between adjacent power units does not exceed 10% to 20% of the average value. If the calculated parameters of a power unit exceed the consistency constraint, a weighted average method is used for correction, and the weight coefficient is determined according to the SOC state and available capacity of the power unit.

[0019] Furthermore, in S4, the feedforward compensation control calculates the feedforward compensation amount based on the frequency, amplitude, and phase information of the oscillation signal, with a compensation gain of 0.5 to 1.0; when a high-frequency oscillation greater than 500Hz is detected, the PWM carrier frequency is adaptively adjusted, and the dead-zone compensation is adaptively adjusted according to the modulation index.

[0020] Furthermore, in S5, the control parameters are adjusted based on an adaptive parameter adjustment mechanism, which includes online parameter optimization, parameter boundary and constraint handling, and multi-condition parameter adaptation, specifically: Online parameter optimization: A gradient descent-based algorithm is used, and a multi-objective optimization function is defined.

[0021] in, The sum of squares of the oscillation amplitude, Sum of squares of parameter adjustment, For power loss, , The weights are used as coefficients; the gradient of the objective function with respect to the parameters is calculated by numerical differentiation, and the parameters are updated using the gradient descent method with momentum. The learning rate is dynamically adjusted according to the severity of oscillations and the optimization process. Parameter boundary and constraint handling: Hard constraints adopt the truncation method, with the scaling factor adjusted from 0.3 to 3.0 times the baseline value and the integral factor adjusted from 0.2 to 5.0 times the baseline value; Soft constraints adopt the penalty function method, adding a penalty term to the objective function that increases with the number of iterations to ensure parameter consistency between units; Multi-condition parameter adaptation: Under charging condition, the damping ratio increases by 0.2 based on the reference value, and the natural frequency decreases to 0.8 times the reference value; under discharging condition, the damping ratio remains at the reference value, and the natural frequency increases to 1.2 times the reference value; under standby condition, the damping ratio increases by 0.3 based on the reference value, and the natural frequency decreases to 0.6 times the reference value.

[0022] According to specific embodiments provided by the present invention, the high-pressure cascaded valve hall energy storage broadband oscillation adaptive stratification suppression system and method disclosed in the present invention have the following technical effects: This invention adopts a three-level control architecture of system layer, valve hall layer, and unit layer, which makes full use of the natural hierarchical characteristics of high-voltage cascaded systems. Through the coordinated cooperation of each level, it achieves rapid detection and suppression of wideband oscillations, with fast response speed and no risk of single point of failure. Based on the adaptive allocation of oscillation frequency suppression tasks, the upper valve hall focuses on low-frequency oscillation suppression, the middle valve hall focuses on medium-frequency oscillation suppression, and the lower valve hall focuses on high-frequency oscillation suppression, so as to achieve targeted suppression of oscillations across the entire frequency band and improve the system stability margin. An adaptive parameter adjustment algorithm based on pole placement, combined with online gradient descent optimization, can dynamically adjust control parameters according to oscillation characteristics and changes in operating conditions, adapting to complex and ever-changing power grid conditions and achieving significant suppression effects. Without requiring additional hardware or changes to the system impedance characteristics, oscillation suppression is achieved through software algorithm optimization, avoiding equipment lockout and ensuring the power output capability of the energy storage system. It has a comprehensive protection mechanism and communication redundancy design, and can switch to local autonomous mode in the event of communication failure or extreme conditions to ensure the safe and stable operation of the system. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0024] Figure 1 This is a diagram of the three-layer control architecture of the high-pressure cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression system of the present invention; Figure 2 This is a flowchart of the oscillation detection and identification process of the present invention; Figure 3 This is a schematic diagram of frequency segmentation and task allocation in this invention; Figure 4 This is a timing diagram of the hierarchical collaborative suppression of the present invention; Figure 5 This is a flowchart of the adaptive adjustment process for PI parameters in this invention. Figure 6 This is a comparison chart of the oscillation suppression effect in Case 1 of Embodiment 2 of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The purpose of this invention is to provide a broadband oscillation adaptive hierarchical suppression system and method for high-voltage cascaded valve hall energy storage, aiming to: fully leverage the distributed control advantages of high-voltage cascaded systems to achieve rapid detection and suppression of oscillations; adapt to oscillation characteristics under different operating conditions through adaptive parameter adjustment to improve the suppression effect; avoid equipment lockout and impedance changes to ensure the power output capability of the energy storage system; and achieve targeted suppression of oscillations in different frequency bands to improve the stability margin of the system.

[0027] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] Example 1 like Figure 1 As shown, this invention provides a high-voltage cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression system. By constructing a three-level control architecture of system layer, valve hall layer, and unit layer, it achieves rapid detection, task allocation, and collaborative suppression of oscillations.

[0029] The high-voltage cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression system includes: a system-level controller, a valve hall-level controller, and a unit-level controller. It fully utilizes the inherent hierarchical characteristics of the high-voltage cascaded system, achieving effective suppression of broadband oscillations through the coordinated operation of each level. The specific system architecture design is as follows: 1.1 System Layer (Monitoring Layer) Design The system layer, located at the station control layer of the energy storage power station, is the "brain" of the entire oscillation suppression system. This layer, through the system layer controller, performs the following main functions: Data Acquisition and Processing: The system layer is equipped with a high-speed data acquisition unit with a sampling frequency of no less than 10kHz, capable of capturing oscillation components up to 5kHz. The acquired data includes key electrical quantities such as three-phase voltage, three-phase current, active power, reactive power, and frequency at the grid connection point. Synchronous sampling technology is used to ensure time alignment of data for each phase, with an error not exceeding 1 microsecond.

[0030] Global Oscillation Monitoring: The system-level controller runs a real-time oscillation monitoring algorithm, employing a sliding window FFT analysis technique. The window length is set to 20 fundamental frequency periods (400ms), and the window sliding step size is 1 fundamental frequency period (20ms), ensuring the real-time performance and accuracy of oscillation detection. Simultaneously, the system layer maintains an oscillation history database, recording all oscillation events over the past 24 hours for oscillation pattern analysis and prediction.

[0031] Task allocation and coordination: The system-level controller calculates the suppression task weights for each valve chamber based on oscillation characteristics, generating a task allocation matrix. Task allocation considers multiple factors: the matching degree between the oscillation frequency and the natural frequency of the valve chamber, the current load rate of each valve chamber, the available capacity of each valve chamber, and historical suppression effects. The system layer employs a weighted optimization algorithm to ensure the rationality and efficiency of task allocation.

[0032] Communication Interface: The system-level controller communicates with each valve hall-level controller via dual-redundant Ethernet, with a communication cycle of 5ms. A time-triggered mechanism ensures deterministic data transmission. The communication protocol adopts the IEC 61850 standard to ensure interoperability.

[0033] 1.2 Valve Hall Layer (Coordination Layer) Design The valve hall layer houses local controllers within each valve hall, serving as a bridge between the system layer and the unit layer. Each valve hall controller manages all power units within its respective valve hall; typically, a valve hall contains 10 to 40 power units.

[0034] Area Status Monitoring: The valve hall layer monitors the operating status of all power units within its area in real time, including DC-side voltage, AC-side current, power unit temperature, switching device status, and battery SOC. The monitoring cycle is 1ms, ensuring accurate understanding of the unit status.

[0035] Parameter optimization calculation: The valve hall level controller calculates the optimal control parameters for each power unit based on the task instructions issued by the system level controller and local status information. The calculation employs an online optimization algorithm, and the objective function comprehensively considers multiple indicators such as oscillation suppression effect, power output capability, and inter-unit balance. The optimization calculation is completed within 10ms to ensure rapid response.

[0036] Inter-unit coordination: The valve hall controller coordinates the control actions of each power unit within its region, ensuring phase synchronization and amplitude coordination among units. The coordination mechanism includes: a unified control cycle, synchronized PWM trigger signals, and a coordinated power distribution strategy. For a valve hall containing 20 power units, the coordination algorithm ensures that the phase error of the output voltage of each unit does not exceed 0.5 degrees, and the amplitude error does not exceed 2%.

[0037] Fault Handling: When a power unit in this valve hall fails, the valve hall controller immediately controls the relevant switching devices to isolate the faulty unit and recalculates the control parameters of the remaining units to ensure that the oscillation suppression function is not interrupted. The fault detection and isolation time does not exceed 2ms.

[0038] 1.3 Unit Layer (Execution Layer) Design The unit layer, located in the local controller of each energy storage power unit, is the final executor of the oscillation suppression strategy. Each power unit contains hardware such as a battery pack, DC / AC converter, and LC filter.

[0039] Local control execution: The unit level executes specific control algorithms based on the control parameters issued by the valve hall level controller. These algorithms include inner current loop control, outer voltage loop control, and phase-locked loop control. The control cycle is 100 microseconds to ensure rapid dynamic response.

[0040] Parameter adaptive adjustment: The unit-level controller has a certain degree of autonomous adjustment capability, making fine adjustments within the parameter range given by the valve hall level to adapt to local real-time operating condition changes. For example, it can fine-tune the power output based on the battery SOC status of the unit, and adjust the switching frequency based on temperature conditions.

[0041] Status feedback: Every 1ms, the unit-level controller feeds back the operating status and control effect of its unit to the valve hall level, providing data support for the valve hall level's coordinated decision-making. Status data includes actual output voltage, current, power, modulation depth, switching frequency, etc.

[0042] Safety Protection: The unit layer implements basic safety protection functions, including overcurrent protection, overvoltage protection, and overtemperature protection. The protection action time does not exceed 10 microseconds to ensure the safety of power devices.

[0043] Example 2 This invention also provides a method for adaptive hierarchical suppression of broadband oscillations in high-voltage cascaded valve hall energy storage, applied to the aforementioned high-voltage cascaded valve hall energy storage system, comprising the following steps: S1, Oscillation Detection and Identification: The system-level controller continuously monitors the electrical quantities at the grid connection point through a multi-domain detection method that combines time domain, frequency domain, and time-frequency domain. After detecting oscillation, it extracts multi-dimensional feature vectors, uses an SVM classifier to identify the oscillation mode, and uses an ARIMA model to predict the oscillation trend and assess the risk level. S2, Layered Task Allocation: The system layer controller calculates the task weight of each valve hall based on the oscillation frequency, amplitude and risk level using a Gaussian distribution function, generates a task allocation matrix and sends it to the valve hall layer controller. S3, Control Parameter Optimization: The valve hall controller calculates the PI control parameters of each power unit based on the task instructions issued by the system layer controller and the local power unit operating status, and corrects the parameters through a multi-unit coordination strategy. S4, Cooperative suppression execution: The unit-level controller receives the control parameters sent by the valve hall-level controller, and after a smooth switch, executes the PI control algorithm. It actively cancels the oscillation component through feedforward compensation control and optimizes the PWM modulation strategy. S5, Effect Evaluation and Iteration: The system-level controller evaluates the oscillation suppression effect every set time period, dynamically adjusts task allocation and control parameters based on the evaluation results, and gradually restores the baseline parameters according to an exponential law after the oscillation is eliminated.

[0044] In this embodiment, the specific details of each key step in the high-voltage cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression method are as follows: 1. Oscillation detection and identification: Oscillation detection and identification are prerequisites for oscillation suppression. This invention employs a method combining multi-domain detection and intelligent identification, such as... Figure 2 As shown.

[0045] 1.1 Multi-domain oscillation detection method This invention employs a multi-domain oscillation detection method that combines time domain, frequency domain, and time-frequency domain to improve the accuracy and reliability of detection.

[0046] Time-domain detection: This method identifies abnormal fluctuations by monitoring the time-domain waveforms of voltage and current. A detection window length of 100ms is set, and the standard deviations of voltage and current within the window are calculated. When the standard deviation exceeds a set threshold, oscillations are considered to be present. The advantage of time-domain detection is its fast response time, capable of detecting anomalies within 10 to 20ms after an oscillation occurs.

[0047] Frequency domain detection: The acquired signal is subjected to FFT transformation to analyze the spectral distribution. The FFT calculation uses 2048 points, with a frequency resolution of approximately 4.88 Hz. All frequency components within the range of 5 Hz to 2000 Hz are scanned to identify frequency points with abnormal amplitudes. The oscillation detection criterion is as follows:

[0048] in, The fundamental frequency is 50Hz. For any non-fundamental frequency, The threshold value is typically 3% to 5%. A similar criterion applies to current. The advantage of frequency domain detection is its ability to accurately identify oscillation frequencies, providing a basis for subsequent hierarchical task allocation.

[0049] Time-frequency domain detection: Wavelet transform is used for time-frequency analysis, which can simultaneously obtain the time and frequency information of oscillations. The wavelet transform uses Morlet wavelets as the mother wavelet, and can identify features such as the onset time, frequency changes, and amplitude evolution of oscillations. Time-frequency domain detection is particularly suitable for detecting non-stationary and transient oscillations.

[0050] 1.2 Intelligent recognition of oscillation modes Upon detecting oscillations, the system needs to identify the oscillation development patterns to provide a basis for selecting suppression strategies. This invention proposes an oscillation pattern recognition method based on machine learning.

[0051] Feature extraction: Extract multi-dimensional feature vectors from the oscillating signal, including time-domain features (peak value, mean, standard deviation, skewness, kurtosis), frequency-domain features (dominant frequency, spectral width, spectral centroid, spectral energy), and time-frequency features (instantaneous frequency, modulation depth, energy distribution). The feature vectors have a dimension of 15 and form a feature matrix.

[0052] Pattern Classification: A Support Vector Machine (SVM) classifier is used to classify oscillation patterns. Classification categories include: divergent oscillations (oscillation amplitude continuously increases, system is unstable), convergent oscillations (oscillation amplitude gradually decreases, system is self-stabilizing), constant-amplitude oscillations (oscillation amplitude remains constant, system is critically stable), intermittent oscillations (oscillations appear and disappear periodically), and chaotic oscillations (oscillation frequency and amplitude change irregularly). The SVM classifier is trained offline using historical oscillation data, with training samples containing over 1000 labeled oscillation events. Online classification time is less than 5ms.

[0053] Trend Prediction: For divergent oscillations, the system needs to predict the development trend of the oscillations and assess the risk level. A time series prediction model is used to predict the oscillation amplitude for the next 100 ms based on the oscillation data of the past 50 ms. The prediction model adopts the ARIMA (Autoregressive Moving Average) model. Based on the prediction results, the oscillation risk level is calculated as follows: low risk (predicted amplitude less than 120% of the threshold), medium risk (predicted amplitude between 120% and 150% of the threshold), and high risk (predicted amplitude greater than 150% of the threshold). Different risk levels correspond to different suppression response speeds and intensities.

[0054] 2. Hierarchical task allocation: The hierarchical task allocation adopts a hierarchical control strategy, which achieves precise suppression of wideband oscillations through three-layer collaboration.

[0055] 2.1 Detailed Description of System-Level Control Strategies The core of the system-level control strategy is the task allocation algorithm, which rationally allocates suppression tasks to each valve chamber based on oscillation characteristics and system state.

[0056] Frequency segmentation strategy: Divide the wideband into multiple sub-bands, each sub-band corresponding to a specific control loop and valve chamber, such as... Figure 3 As shown. The ultra-low frequency band (5Hz to 20Hz) is mainly caused by outer loop power control, with a weighting of 70% for the upper valve hall, 20% for the middle valve hall, and 10% for the lower valve hall. The low frequency band (20Hz to 100Hz) is mainly caused by phase-locked loop (PLL) control, with a weighting of 60% for the upper valve hall, 30% for the middle valve hall, and 10% for the lower valve hall. The mid-low frequency band (100Hz to 300Hz) is mainly caused by voltage outer loop control, with a weighting of 30% for the upper valve hall, 50% for the middle valve hall, and 20% for the lower valve hall. The mid-frequency band (300Hz to 600Hz) is mainly caused by current inner loop control, with a weighting of 20% for the upper valve hall, 60% for the middle valve hall, and 20% for the lower valve hall. The high-frequency band (600Hz to 1000Hz) is mainly caused by digital control delay and sampling, with the weighting for this band being 10% for the upper valve hall, 30% for the middle valve hall, and 60% for the lower valve hall. The ultra-high-frequency band (1000Hz to 2000Hz) is mainly caused by PWM carrier and LC filter resonance, with the weighting for this band being 5% for the upper valve hall, 15% for the middle valve hall, and 80% for the lower valve hall.

[0057] Weighting calculation formula: for the detected oscillation frequency , No. The task weights for each valve chamber are calculated using a Gaussian distribution function:

[0058] in, The normalization coefficients ensure that the sum of all weights is 1; For the first The center frequency of each valve hall is 50Hz for the upper valve hall, 300Hz for the middle valve hall, and 1000Hz for the lower valve hall. The frequency distribution width is set to 100Hz for the upper layer, 200Hz for the middle layer, and 500Hz for the lower layer.

[0059] Multi-frequency oscillation handling: When the system detects oscillations at multiple frequencies simultaneously, the system layer independently calculates a weight matrix for each oscillation frequency, and then performs comprehensive optimization. The comprehensive weight is the weighted sum of the weights of each frequency, and the weight coefficients are related to the oscillation amplitude and risk level.

[0060] Dynamic adjustment mechanism: The system layer evaluates the oscillation suppression effect every 100ms and dynamically adjusts the task allocation based on the evaluation results. The evaluation index comprehensively considers the oscillation suppression rate, response time, and power loss, with weight coefficients typically set to 0.6, 0.3, and 0.1. If the comprehensive evaluation index is lower than the expected value of 0.7, the system layer recalculates the weight allocation, with an adjustment range not exceeding 20%, to avoid frequent and large adjustments that could lead to system instability.

[0061] 2.2 Detailed Explanation of Valve Hall Layer Control Strategy The valve chamber layer is the core layer for parameter optimization. Based on the task allocation and local state information of the system layer, it calculates the optimal control parameters for each power unit, such as... Figure 5 As shown.

[0062] PI parameter adaptive optimization: The current loop control of energy storage converters typically uses a PI controller, whose transfer function is:

[0063] The performance of the controller is determined by the proportional coefficient. and integral coefficient Decision. This invention proposes a PI parameter optimization method based on pole placement. For the current loop of the energy storage converter, the filter inductor is considered. and equivalent resistance Construct a closed-loop system. This is a second-order system. To obtain good dynamic performance and damping characteristics, the target pole is designed as follows:

[0064] in, For the damping ratio, The natural frequency. Using the pole placement method, the relationship between the PI parameters and the target pole can be obtained:

[0065]

[0066] Adaptive adjustment strategy: When oscillation is detected, the target damping ratio and natural frequency are adjusted according to the oscillation characteristics. The adjustment amount of the damping ratio is proportional to the task weight of this valve chamber and the degree to which the oscillation amplitude exceeds the threshold, with a typical adjustment range of 0.1 to 0.3. The adjustment amount of the natural frequency is proportional to the degree to which the oscillation frequency deviates from the reference frequency, with a typical adjustment coefficient of 0.2 to 0.5. After adjustment, the PI parameters are recalculated according to the pole placement formula.

[0067] Parameter constraints and protection: To ensure system stability, parameter adjustments must meet stability constraints (both proportional coefficient and integral coefficient are greater than zero), adjustment range constraints (new parameters are within the range of 0.5 to 2.0 times the baseline parameters), and phase margin constraints (not less than 45 degrees).

[0068] Multi-unit coordination strategy: The valve hall contains multiple power units, and the control parameters of each unit need to be coordinated to avoid inconsistencies between units that could lead to increased circulating current or oscillations. Parameter consistency between units requires that the parameter differences between adjacent units do not exceed 10% to 20% of the average value. If the calculated parameters of a unit exceed the consistency constraint, a weighted average method is used for correction. The weighting coefficient is determined based on the unit's SOC status and available capacity; units with larger capacities have higher weights.

[0069] 2.3 Detailed Explanation of Unit-Level Control Strategy The unit layer is the execution layer of the control strategy, which executes the specific control algorithm according to the parameters issued by the valve hall layer.

[0070] Current loop control implementation: The current loop control uses a digital PI controller, implemented in a discrete manner. The sampling period is 100 microseconds. The discrete PI controller is implemented incrementally, and the control quantity in each sampling period is equal to the control quantity in the previous period plus the increment of the proportional term and the increment of the integral term.

[0071] Feedforward Compensation Control: To actively suppress oscillations, a feedforward compensation stage is introduced into the controller. The feedforward compensation amount is calculated based on the detected oscillation signal, including oscillation frequency, amplitude, and phase information. The compensation gain is determined through offline experiments, with typical values ​​ranging from 0.5 to 1.0. The compensation phase needs to account for system delay to ensure that the compensation signal is out of phase with the oscillation signal.

[0072] Modulation strategy optimization: Under high-frequency oscillation conditions, PWM modulation may interact with the oscillation frequency, exacerbating the oscillation. Therefore, the modulation strategy needs to be optimized. When high-frequency oscillation (greater than 500Hz) is detected, the PWM carrier frequency is adaptively adjusted to avoid overlap with harmonics of the oscillation frequency. Dead time introduces nonlinearity; dead time compensation is adaptively adjusted according to the modulation index, with a larger compensation amount at low modulation indices.

[0073] 3. Adaptive parameter adjustment mechanism 3.1 Online Parameter Optimization Algorithm This invention employs an online parameter optimization algorithm based on gradient descent to adjust control parameters in real time to minimize oscillation amplitude.

[0074] Objective function design: Define a multi-objective optimization function that comprehensively considers oscillation suppression, control smoothness, and power loss.

[0075] in, The sum of squares of the oscillation amplitude, Sum of squares of parameter adjustment, For power loss, , The weighting coefficients are usually taken as 0.1 or 0.05.

[0076] Gradient Calculation and Parameter Update: The gradient of the objective function with respect to the parameters is calculated using numerical differentiation, with a small perturbation of 0.1% of the parameter value. Parameters are updated using gradient descent with momentum, typically with a momentum coefficient of 0.9. The learning rate is dynamically adjusted based on the severity of oscillations and the optimization progress; the more severe the oscillations, the larger the learning rate, and the smaller the learning rate gradually becomes with the number of iterations.

[0077] 3.2 Parameter Boundary and Constraint Handling Hard constraint handling: Parameter adjustments must meet hard constraint conditions, using a truncation method. If the calculated new parameters exceed the allowable range, they are truncated to boundary values. Parameter boundaries are determined based on system stability analysis, with the proportional coefficient ranging from 0.3 to 3.0 times the baseline value and the integral coefficient ranging from 0.2 to 5.0 times the baseline value.

[0078] Soft constraint handling: For soft constraints such as inter-unit consistency, a penalty function method is used. A penalty term is added to the objective function, and the penalty coefficient increases with the number of iterations, so that the parameters gradually meet the consistency requirements.

[0079] 3.3 Multi-condition parameter adaptive The dynamic characteristics of a system differ under different operating conditions, and parameter optimization strategies need to be adjusted accordingly.

[0080] Charging condition parameter adjustment: During charging, the current flows from the grid to the energy storage, and the system exhibits capacitive characteristics. At this time, the damping ratio is increased by 0.2 from the reference value; the natural frequency is reduced to 0.8 times the reference value. This setting enhances system stability and avoids oscillations during charging.

[0081] Discharge condition parameter adjustment: Under discharge conditions, the current flows from the energy storage to the grid, and the system exhibits inductive characteristics. At this time, the damping ratio remains at the reference value; the natural frequency is increased to 1.2 times the reference value. This setting improves the system's dynamic response speed and ensures rapid power output.

[0082] Standby operating parameters adjustment: In standby mode, the system does not exchange power but needs to maintain grid connection. At this time, the damping ratio is at its maximum, increased by 0.3 from the reference value; the natural frequency decreases to 0.6 times the reference value. This setting can suppress oscillations to the greatest extent while reducing system losses.

[0083] 4. Synergistic Inhibition Process Specifically, in this embodiment of the invention, the collaborative suppression process of the high-voltage cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression method is as follows: Figure 4 As shown, it includes: Step 1: Continuous monitoring (0 to 10 ms) The system-level oscillation monitoring system continuously acquires voltage and current signals at the grid connection point at a sampling rate of 10kHz. An FFT analysis is performed every 20ms, scanning all frequency components within the range of 5Hz to 2000Hz. The detection algorithm is as follows: voltage and current are acquired in each sampling period and stored in a sliding window buffer; a 2048-point FFT is performed on the window data in each fundamental frequency period to calculate the amplitude of each frequency component; if the ratio of the amplitude of a certain frequency component to the fundamental frequency amplitude is greater than 3%, the oscillation frequency and amplitude are recorded, triggering an oscillation event.

[0084] Step 2: Oscillation Confirmation and Classification (10 to 30 ms) After oscillation detection is triggered, the confirmation phase begins, employing a multi-cycle verification mechanism. Five oscillation cycles are continuously monitored, and the amplitude of each cycle is recorded. The rate of change of amplitude is calculated; if the rate of change is positive and large, it is determined to be a high-risk divergent oscillation; if the rate of change is negative, it is determined to be a convergent oscillation requiring no intervention; otherwise, it is determined to be a medium-risk constant-amplitude oscillation. For both divergent and constant-amplitude oscillations, feature vectors are extracted and input into an SVM classifier for further classification.

[0085] Step 3: Task allocation decision (30 to 50 ms) The system layer calculates the task weights for each valve chamber based on oscillation characteristics. For each valve chamber: a base weight (based on a Gaussian distribution function) is calculated; valve chamber state corrections are considered (weights are reduced if the load rate is too high or a faulty unit exists); and weights are normalized so that the sum of all weights is 1. Task instructions, including target frequency, target suppression rate, weights, and time limits, are generated and distributed to each valve chamber via the communication network.

[0086] Step 4: Valve Hall Parameter Calculation (50 to 100 ms) After receiving the task command, each valve chamber independently calculates its local optimal parameters. It reads the task weights and local status (load rate, SOC distribution, temperature, etc.); calculates the target damping ratio and natural frequency, limiting them to a reasonable range; and calculates the PI parameters according to the pole placement method. For each unit within its valve chamber: it calculates the weights based on the unit's SOC and capacity, performs a weighted average correction on the PI parameters, checks parameter constraints, and adjusts the parameters to the constraint boundaries if any are violated. Finally, it generates control commands and sends them to each unit.

[0087] Step 5: Unit-level execution and feedback (100ms to 2s) Each power unit performs parameter updates and control. Upon receiving new parameters, it smoothly switches parameters within 10ms to avoid abrupt changes. PI control is executed to calculate errors and control values, and feedforward compensation is added to actively cancel oscillation components before outputting to the PWM modulator. Status feedback is sent to the valve hall level every 1ms.

[0088] Step 6: Effect Evaluation and Iteration (evaluate every 100ms) The system layer continuously monitors the oscillation suppression effect. It calculates the current oscillation amplitude and suppression rate every 100ms. If the suppression rate is less than 50% and the timeout has not occurred, the suppression effect is poor, requiring strategy adjustments, appropriately increasing task weights, recalculating parameters, and issuing new ones. If the oscillation amplitude drops below the threshold, suppression is successful, and the system enters the parameter recovery phase. If the timeout has occurred and the suppression is unsuccessful, emergency measures are initiated to reduce power output.

[0089] Step 7: Gradual parameter recovery (2s to 12s) After the oscillation disappears, the system does not immediately restore the original parameters, but rather recovers gradually. The recovery time constant is set to 5 seconds. The parameters gradually recover from the suppressed parameters to the reference parameters according to an exponential law. During the recovery process, continuous monitoring is conducted to check for the recurrence of oscillations. If an oscillation is detected, the recovery process is stopped and suppression is restarted. After the recovery is complete, the system enters normal monitoring mode.

[0090] Furthermore, the high-voltage cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression method of the present invention also includes a multi-band parallel processing scheme: when the system detects oscillations at multiple frequencies simultaneously, a parallel processing strategy is adopted to independently calculate the weight of each valve hall for each oscillation frequency, generate independent tasks, and after each valve hall receives multiple tasks, calculates the comprehensive weight. The weight coefficient is determined according to the oscillation risk level, and a multi-objective optimization method is used to calculate the comprehensive objective parameters to ensure that there is a suppression effect on all frequencies. The optimal PI parameter is solved through numerical optimization.

[0091] The high-voltage cascaded valve hall energy storage broadband oscillation adaptive stratification suppression method of the present invention further includes: 5. Working condition adaptability design scheme: 5.1 Detailed charging strategy: During charging, the energy storage system acts as a load, absorbing power from the grid. Under these conditions, the dynamic characteristics of the system are significantly different from those under discharging conditions.

[0092] Charging power reserve: To ensure control margin during oscillation suppression, the charging power should not be operated at full load, and the maximum charging power is set to 95% of the rated power. When oscillation is detected, the charging power should be appropriately reduced, and the reduction amount is related to the task weight and the severity of the oscillation.

[0093] Reactive power assistance: During charging, active power is mainly controlled, but reactive power can be used to improve damping. The amplitude of the injected reactive power is proportional to the amplitude of the oscillation, and the phase selection makes the reactive power and the oscillation power out of phase.

[0094] Charging current distribution optimization: The charging current of each valve hall is redistributed according to the oscillation suppression task. Valve halls with high task weights have their charging current appropriately reduced, while valve halls with low task weights have their charging current increased accordingly to compensate, ensuring that the total charging power remains basically unchanged.

[0095] 5.2 Detailed strategies for discharge conditions: During discharge, the energy storage system transmits power to the grid, and it is necessary to ensure the stability of the power output.

[0096] Power redistribution strategy: Maintaining a constant total output power, oscillation suppression is achieved through inter-unit redistribution. Valve halls with high task weights reduce power output, with a typical power adjustment factor of 0.1 to 0.2. Valve halls with low task weights increase power output to compensate, ensuring overall power balance.

[0097] SOC balancing considerations: Power redistribution requires consideration of the SOC status of each valve hall. If the SOC of a valve hall is low, its power increase is limited; if the SOC of a valve hall is high, its power output is increased preferentially.

[0098] Dynamic power factor adjustment: Improves system damping characteristics by adjusting the power factor. The adjusted power factor is reduced by a certain amount from the reference value, and the amount of reduction is related to the task weight and the degree to which the oscillation frequency deviates from the reference frequency.

[0099] 5.3 Detailed strategy for standby operation: In standby mode, the system does not exchange active power, but it can use reactive power regulation capabilities to participate in oscillation suppression.

[0100] Active reactive power injection: Upon detecting oscillation, the required reactive power is calculated. The reactive power amplitude is proportional to the oscillation current amplitude and the grid voltage. Reactive power is allocated to each valve hall according to its weight. When executing reactive power control, the d-axis current command is adjusted to keep the q-axis current zero.

[0101] DC voltage fine-tuning: The dynamic characteristics of the system are altered by fine-tuning the DC-side voltage. The adjusted DC voltage increases or decreases slightly from the reference value, with a typical adjustment factor of 2% to 5%. Voltage adjustment is achieved through short-term micro-charging or micro-discharging, with the micro-charging / discharging power being approximately 1% of the rated power.

[0102] The adaptive stratified suppression method for broadband oscillations in high-voltage cascaded valve hall energy storage described in this invention further includes: 6. Protection and Restriction Schemes: 6.1 Multi-layered protection mechanism System-level protection: Monitors the overall status and activates system-level protection when a serious anomaly occurs. Monitoring indicators include: grid connection point voltage within 90% to 110% of the rated value, grid connection point current not exceeding 120% of the rated value, frequency deviation not exceeding 0.5Hz, and oscillation duration not exceeding 10 seconds. If any indicator exceeds the limit, corresponding measures are taken: if the voltage exceeds the limit, immediately reduce power and increase reactive power support; if the current exceeds the limit, gradually reduce power; if the frequency exceeds the limit, switch to frequency support mode; if the oscillation timeout occurs, significantly reduce power and record the alarm.

[0103] Valve Hall Protection: Monitors the status of the area to protect the equipment in this valve hall. It monitors indicators such as DC bus voltage, valve hall temperature, and number of power unit failures for each valve hall. If the DC voltage exceeds the limit, it adjusts the charging / discharging power balance voltage; if the voltage exceeds the limit for more than 1 second, it locks the valve hall and switches to bypass. If the temperature exceeds the limit, it reduces the power of the valve hall and initiates forced cooling. If there are too many faulty units, it redistributes power to healthy units; if the remaining capacity is insufficient, it reduces the rated capacity of the valve hall.

[0104] Unit-level protection: Each power unit implements rapid protection. Hardware protection response time is less than 10 microseconds, including overcurrent protection, overvoltage protection, and short-circuit protection. Software protection response time is less than 1 millisecond, including overtemperature protection, overload protection, and communication failure protection. When protection is activated, hardware protection immediately shuts down the IGBT and switches to fault bypass; software protection reduces power in an orderly manner, and if the fault persists, the unit is shut down.

[0105] 6.2 Handling Communication Failures Communication redundancy design: Dual-network redundant communication is used between the system layer and the valve hall layer, automatically switching to the backup network in case of primary network failure. Communication monitoring checks the communication status every 5ms. If primary network communication fails, it switches to the backup network and records the failure time; if the backup network also fails, it initiates local autonomous mode.

[0106] Local Autonomous Mode: When communication fails completely, each layer operates independently. The system layer continues to monitor oscillations but cannot issue task commands, recording oscillation data for subsequent analysis. The valve hall layer runs a preset local oscillation suppression strategy, employing fixed parameter switching and power limiting. If local oscillation is detected, it switches to a preset parameter group and reduces power. The unit layer performs basic current and voltage control, maintaining communication with the valve hall layer.

[0107] Synchronization after communication restoration: After communication is restored, the system layer reacquires the status of each valve chamber and checks for any parameter inconsistencies. If inconsistencies are found, oscillation suppression is paused, all parameters are updated uniformly, and collaborative control is restarted; otherwise, normal collaborative mode is restored.

[0108] The following detailed examples illustrate the practical application effects of the method of the present invention.

[0109] Case 1: Suppression of Mid-Frequency Oscillations System configuration: Energy storage power station capacity 100MW / 200MWh, grid connection voltage level 35kV, number of valve halls 3 (upper, middle and lower layers), number of power units in each valve hall 20, and capacity of a single power unit 1.67MW.

[0110] Operating conditions: The working mode is discharge, the output power is 50MW (50% of the rated power), the power factor is 1.0, and the grid short-circuit ratio is 3.0.

[0111] Oscillation event: At t=0, the system detected a 180Hz oscillation due to the sudden change in the grid impedance from 0.5Ω to 0.8Ω.

[0112] Detailed inhibition process: Oscillation occurs at time T0 (0ms): Sudden changes in grid impedance cause system oscillation, with a small initial amplitude.

[0113] Oscillation detection at time T1 (10ms): System-level FFT analysis detected a 180Hz oscillation component, with a current oscillation amplitude of 45.2A (5.4% of the fundamental current of 833A) and a voltage oscillation amplitude of 890V (4.45% of the fundamental voltage of 20kV). This triggered an oscillation alarm and initiated the suppression program.

[0114] Oscillation confirmation at time T2 (20ms): Monitoring over 5 consecutive periods showed oscillation amplitudes of 45.2A, 47.8A, 50.6A, 53.8A, and 57.2A. The calculated growth rate was 540A / s. This was determined to be a high-risk divergent oscillation.

[0115] Task allocation at time T3 (30ms): The system layer calculates the weights of each valve hall. 180Hz belongs to the mid-frequency band, and based on the Gaussian distribution function, the weights are calculated as follows: Valve Hall 1 weight 0.24 (upper-level center frequency 50Hz), Valve Hall 2 weight 0.59 (middle-level center frequency 300Hz, mainly responsible), and Valve Hall 3 weight 0.17 (lower-level center frequency 1000Hz). Task instructions are then issued to each valve hall.

[0116] Parameter calculation at time T4 (40ms): Valve Hall 2 is the main responsible unit for parameter calculation. The baseline parameters are K_p=0.5, Ki=50, ζ=0.707, ω_n=2π×200rad / s. Based on the task weight of 0.59 and the oscillation amplitude ratio of 2.29, the new damping ratio is calculated to be 0.936, and the new natural frequency is 1283.5rad / s. According to the pole placement formula (L=2mH, R=0.1Ω), K_p=4.70 and Ki=3297 are calculated. After parameter constraint checks, K_p is adjusted to 1.0 and Ki=250 (both reaching the upper limit).

[0117] At time T5 (50ms), parameters are issued and executed: the 20 power units in valve hall 2 receive new parameters and smoothly switch parameters within 10ms. Feedforward compensation is added while executing PI control; a system delay of 150μs corresponds to a phase compensation of 9.7 degrees.

[0118] Preliminary effect assessment at time T6 (100ms): Oscillation amplitude decreased to 48.5A, suppression rate 15.2%. The effect is not significant, continue execution.

[0119] Mid-term assessment at time T7 (500ms): Oscillation amplitude decreased to 32.7A, suppression rate 42.8%. A suppression rate less than 50% necessitates strategy adjustment. The system-level decision-making weight of valve hall 2 was increased to 0.71, and the parameters of valve hall 2 were recalculated to further increase the damping ratio to 1.05 (overdamped).

[0120] Significant improvement was observed at time T8 (1000ms): the oscillation amplitude decreased to 18.3A, with a suppression rate of 68.0%. The oscillation amplitude has been reduced to below the threshold of 25A, indicating successful suppression.

[0121] Oscillation eliminated at time T9 (2000ms): Oscillation amplitude reduced to 6.8A, suppression rate 88.1%. Oscillation basically eliminated, parameter recovery procedure initiated.

[0122] Parameter recovery at time T10 (2000ms to 12000ms): A recovery time constant of 5 seconds is set, and the parameters gradually recover from the suppressed parameters to the reference parameters according to an exponential law. The oscillation suppression effect is as follows: Figure 6 As shown, the horizontal axis represents time (0 to 12 seconds), and the vertical axis represents the amplitude of the oscillating current (0 to 60A). Key time points: t=2s, recovery factor 0.33.K p =0.835、 K i =184; at t=5s, the recovery factor is 0.63. K p =0.685、 K i =124; at t=10s, the recovery factor is 0.86. K p =0.57、 K i =78. Finally, the system fully recovered to the baseline parameters and returned to normal operation. Figure 6 The key time points and suppression rates are marked in the middle, and the oscillation threshold of 25A is marked with a dashed line.

[0123] Summary of suppression effect: detection response time 20ms, parameter calculation time 20ms, significant suppression time 1000ms, final suppression rate 88.1%, power output maintained at 50MW without impact throughout.

[0124] Case 2: Concurrent Processing of Multi-Frequency Oscillations Oscillation event: The system simultaneously detected oscillations at two frequencies, 85Hz and 520Hz.

[0125] During the T1 time detection phase: FFT analysis results show that the 85Hz oscillation amplitude of 35.6A (4.3%) is a medium risk of constant amplitude oscillation, and the 520Hz oscillation amplitude of 28.4A (3.4%) is a medium risk of slow divergence.

[0126] Task allocation at time T2: For the 85Hz oscillation, based on the frequency segmentation strategy and Gaussian distribution function, the weights are calculated as follows: Valve Hall 1: 0.68 (the upper layer mainly handles low frequencies), Valve Hall 2: 0.25, Valve Hall 3: 0.07. For the 520Hz oscillation: Valve Hall 1: 0.08, Valve Hall 2: 0.32, Valve Hall 3: 0.60 (the lower layer mainly handles high frequencies).

[0127] When calculating the overall weight, the weight coefficient of each oscillation is calculated first. The weight coefficient for 85Hz is 3.04 (based on amplitude ratio and risk level), and the weight coefficient for 520Hz is 1.93. After aggregation, the total weights are: Valve Hall 1 0.45, Valve Hall 2 0.28, and Valve Hall 3 0.27.

[0128] Parameter optimization at time T3: The optimization objective of valve hall 1 is primarily to suppress 85Hz, with 85Hz accounting for 90% of the weight and 520Hz accounting for 10% in the multi-objective optimization function. The optimization objective of valve hall 3 is primarily to suppress 520Hz, with 85Hz accounting for 10% of the weight and 520Hz accounting for 90% in the multi-objective optimization function. Valve hall 2 considers both frequencies, with each frequency accounting for 50%. The optimal PI parameters for each valve hall are solved through numerical optimization.

[0129] Execution and Results at Time T4: At t=1s, the 85Hz oscillation amplitude decreased to 25.3A (suppression rate 29%), and the 520Hz oscillation amplitude decreased to 19.7A (suppression rate 31%). At t=2s, the 85Hz oscillation amplitude decreased to 16.8A (suppression rate 53%), and the 520Hz oscillation amplitude decreased to 12.1A (suppression rate 57%). Both frequency oscillations were effectively suppressed, verifying the effectiveness of the multi-band parallel processing strategy.

[0130] Case Summary: This case demonstrates the system's ability to handle multi-frequency oscillations. By independently calculating task weights for each oscillation frequency and then comprehensively optimizing the control parameters of each valve chamber, simultaneous suppression of oscillations across different frequency bands is achieved. Each valve chamber undertakes different suppression tasks based on its own frequency characteristics, fully leveraging the advantages of a hierarchical architecture.

[0131] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A high-pressure cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression system, characterized in that, include: The system-level controller, valve hall-level controller, and unit-level controller are connected to each other via dual-redundant Ethernet communication, and the communication protocol adopts the IEC61850 standard. The system-level controller is equipped with a high-speed data acquisition unit to collect three-phase voltage, three-phase current, active power, reactive power and frequency data at the grid connection point, and run real-time oscillation monitoring algorithm and task allocation algorithm to realize global oscillation monitoring, task allocation and coordination. The number of valve hall level controllers is consistent with the number of valve halls. Each valve hall level controller is electrically connected to multiple power units and is used to monitor the operating status of the power units in the valve hall, calculate the optimal control parameters according to the task instructions issued by the system level controller, and coordinate the control actions of each power unit in the valve hall. The unit-level controller corresponds one-to-one with the power unit. It is used to receive control parameters sent by the valve hall-level controller, perform adaptive parameter adjustment, execute current inner loop control, voltage outer loop control and phase-locked loop control operations, and provide feedback on the operating status of the corresponding power unit.

2. The high-pressure cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression system according to claim 1, characterized in that, The high-speed data acquisition unit of the system layer controller adopts synchronous sampling technology to achieve time alignment of each phase data; The real-time oscillation monitoring algorithm uses sliding window FFT analysis technology; the task allocation algorithm uses a weighted optimization algorithm.

3. The high-pressure cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression system according to claim 1, characterized in that, The power unit operating status monitored by the valve hall controller includes DC side voltage, AC side current, power unit temperature, switching device status, and battery SOC.

4. A method for adaptive hierarchical suppression of broadband oscillations in high-voltage cascaded valve hall energy storage, applied to the high-voltage cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression system according to any one of claims 1-3, characterized in that, Includes the following steps: S1, Oscillation Detection and Identification: The system-level controller continuously monitors the electrical quantities at the grid connection point through a multi-domain detection method that combines time domain, frequency domain, and time-frequency domain. After detecting oscillation, it extracts multi-dimensional feature vectors, uses an SVM classifier to identify the oscillation mode, and uses an ARIMA model to predict the oscillation trend and assess the risk level. S2, Layered Task Allocation: The system layer controller calculates the task weight of each valve hall based on the oscillation frequency, amplitude and risk level using a Gaussian distribution function, generates a task allocation matrix and sends it to the valve hall layer controller. S3, Control Parameter Optimization: The valve hall controller calculates the PI control parameters of each power unit based on the task instructions issued by the system layer controller and the local power unit operating status, and corrects the parameters through a multi-unit coordination strategy. S4, Cooperative suppression execution: The unit-level controller receives the control parameters sent by the valve hall-level controller, and after a smooth switch, executes the PI control algorithm. It actively cancels the oscillation component through feedforward compensation control and optimizes the PWM modulation strategy. S5, Effect Evaluation and Iteration: The system-level controller evaluates the oscillation suppression effect every set time period, dynamically adjusts task allocation and control parameters based on the evaluation results, and gradually restores the baseline parameters according to an exponential law after the oscillation is eliminated.

5. The adaptive hierarchical suppression method for broadband oscillations in high-pressure cascaded valve hall energy storage according to claim 4, characterized in that, In S1, the multi-domain detection method specifically includes: Time-domain detection: Using a detection window of a set length, the standard deviation of voltage and current within the window is calculated. If the standard deviation exceeds a set threshold, it is determined that oscillation may exist. Frequency domain detection: Perform FFT transformation on the acquired signal, analyze the spectral distribution, and determine that oscillation exists when the ratio of the amplitude of the non-fundamental frequency component to the amplitude of the fundamental frequency exceeds a set value; Time-frequency domain detection: Morlet wavelet is used for time-frequency analysis to identify the oscillation start time, frequency changes and amplitude evolution characteristics; In S1, the multidimensional feature vector includes: Time-domain characteristics: peak value, mean, standard deviation, skewness, kurtosis; Frequency domain characteristics: dominant frequency, spectral width, spectral centroid, spectral energy; Time-frequency characteristics: instantaneous frequency, modulation depth, and energy distribution.

6. The adaptive hierarchical suppression method for broadband oscillations in high-pressure cascaded valve hall energy storage according to claim 4, characterized in that, In S2, the task weights are calculated using the following Gaussian distribution function: in, These are the normalization coefficients; The detected oscillation frequency; For the first The center frequency of each valve chamber For the first The frequency distribution width of each valve chamber; When multiple oscillations are detected simultaneously, the system-level controller independently calculates a weight matrix for each oscillation frequency, and then performs a comprehensive analysis based on the weighted sum of the weights of each frequency.

7. The adaptive hierarchical suppression method for broadband oscillations in high-pressure cascaded valve hall energy storage according to claim 4, characterized in that, In step 3, the PI control parameters are calculated using the pole placement method, as shown in the following formula: in, This is the proportionality coefficient. The integral coefficient is... For the damping ratio, Where is the natural frequency, L is the filter inductance, and R is the equivalent resistance.

8. The high-pressure cascaded valve hall energy storage broadband oscillation adaptive hierarchical suppression method according to claim 4, characterized in that, The multi-unit coordination strategy in S3 is as follows: the parameter difference between adjacent power units does not exceed 10% to 20% of the average value. If the calculated parameters of a power unit exceed the consistency constraint, a weighted average method is used for correction. The weight coefficient is determined according to the SOC status and available capacity of the power unit.

9. The adaptive hierarchical suppression method for broadband oscillations in high-pressure cascaded valve hall energy storage according to claim 4, characterized in that, In S4, the feedforward compensation control calculates the feedforward compensation amount based on the frequency, amplitude, and phase information of the oscillation signal, with a compensation gain of 0.5 to 1.0; when a high-frequency oscillation greater than 500Hz is detected, the PWM carrier frequency is adaptively adjusted, and the dead-zone compensation is adaptively adjusted according to the modulation index.

10. The adaptive hierarchical suppression method for broadband oscillations in high-pressure cascaded valve hall energy storage according to claim 4, characterized in that, In step S5, the control parameters are adjusted based on an adaptive parameter adjustment mechanism. This adaptive parameter adjustment mechanism includes online parameter optimization, parameter boundary and constraint handling, and multi-condition parameter adaptation. Specifically: Online parameter optimization: A gradient descent-based algorithm is used, and a multi-objective optimization function is defined. in, The sum of squares of the oscillation amplitudes, Sum of squares of parameter adjustment, For power loss, , The weights are used as coefficients; the gradient of the objective function with respect to the parameters is calculated by numerical differentiation, and the parameters are updated using the gradient descent method with momentum. The learning rate is dynamically adjusted according to the severity of oscillations and the optimization process. Parameter boundary and constraint handling: Hard constraints adopt the truncation method, with the scaling factor adjusted from 0.3 to 3.0 times the baseline value and the integral factor adjusted from 0.2 to 5.0 times the baseline value; Soft constraints adopt the penalty function method, adding a penalty term to the objective function that increases with the number of iterations to ensure parameter consistency between units; Multi-condition parameter adaptation: Under charging condition, the damping ratio increases by 0.2 based on the reference value, and the natural frequency decreases to 0.8 times the reference value; under discharging condition, the damping ratio remains at the reference value, and the natural frequency increases to 1.2 times the reference value; under standby condition, the damping ratio increases by 0.3 based on the reference value, and the natural frequency decreases to 0.6 times the reference value.

Images