Frequency modulation method of energy storage power station under virtual synchronous machine control

Abstract

The present application relates to the technical field of power system operation and control, and discloses a frequency modulation method for energy storage power station under virtual synchronous machine control, wherein the energy storage power station is connected with an alternating current power grid through a virtual synchronous machine control converter, and is composed of a plurality of energy storage units connected in parallel through a common direct current bus; the method executes the following control process: S1, power grid frequency tracking and instruction generation: based on the virtual synchronous machine control converter, the alternating current power grid frequency is perceived, and a primary frequency modulation active power instruction is generated accordingly; the present application takes the direct current bus voltage deviation as a core control variable, and sets up a double-mode dynamic decision mechanism based on the deviation threshold, can quickly identify the disturbance intensity within the millisecond level time when the power grid frequency disturbance occurs, and immediately enters the fast power support mode, realizes the fastest and most sufficient response to the power shortage, thereby significantly improves the lowest point of the frequency, effectively suppresses the frequency drop rate, and shortens the system recovery time.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology, specifically to a frequency regulation method for energy storage power stations under virtual synchronous machine control. Background Technology

[0002] With the large-scale grid connection of high-proportion renewable energy sources such as wind power and photovoltaics, the inertia level and frequency regulation capability of the power system have been significantly weakened. To address this challenge, virtual synchronous machine (VSM) technology has been widely applied to power electronic interface devices such as energy storage converters. By simulating the rotor motion equations of synchronous generators, it provides virtual inertia and damping to the system, thereby supporting grid frequency stability. However, traditional VSM control methods have inherent limitations in dealing with rapid and large-amplitude power disturbances in the grid. A typical improvement scheme, such as the "A Model Predictive Virtual Synchronous Machine Control-Based Energy Storage Frequency Regulation Method" disclosed in authorization announcement number CN113890055B, focuses on establishing a frequency regulation method that includes frequency regulation... The cost function of rate increment and power increment is used, and rolling optimization is performed using model predictive control to improve frequency dynamic characteristics. This scheme represents a direction of current technological development. However, its control logic relies on online optimization calculation of the predictive model of AC side frequency changes, which poses challenges in computational complexity and real-time performance. More importantly, this method mainly focuses on optimizing AC side frequency indicators and fails to fully explore and utilize the synergistic potential between multiple energy storage units within the energy storage power station and the regulating role of the DC bus, a key physical quantity. This results in the system being unable to achieve optimal and fastest support for the grid frequency while maintaining internal DC voltage stability and inter-unit charge state balance under rapid power disturbances. Summary of the Invention

[0003] The purpose of this invention is to provide a frequency regulation method for energy storage power stations under virtual synchronous machine control, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a frequency regulation method for an energy storage power station under virtual synchronous machine control, wherein the energy storage power station is connected to the AC grid via a converter controlled by a virtual synchronous machine, and consists of multiple energy storage units connected in parallel via a common DC bus; the method executes the following control flow:

[0005] S1. Grid frequency tracking and command generation: Based on the virtual synchronous machine control converter, the AC grid frequency is sensed and a primary frequency regulation active power command is generated accordingly.

[0006] S2, DC side target voltage setting: Based on the primary frequency regulation active power command and the actual output power of the energy storage power station, set the target voltage reference value of the common DC bus;

[0007] S3. Dynamic decision-making on operation mode: Dynamically switch between the fast power support mode and the state of charge balance mode based on the deviation between the real-time voltage of the common DC bus and its target voltage reference value, as well as the preset threshold conditions;

[0008] S4. Coordinated power distribution in multiple modes: Generate corresponding power distribution commands for each energy storage unit according to the operation mode determined in step S3; among them, the fast power support mode takes maximizing the frequency modulation response speed as the priority goal, and the state of charge balance mode takes coordinating the state of charge of each energy storage unit as the priority goal;

[0009] S5. Execution of power commands and closed-loop control: Each energy storage unit adjusts its charge and discharge power through local control according to the allocated power command, so that the actual voltage of the common DC bus is stabilized near the target voltage reference value, and the total output power of the energy storage power station actively supports the grid frequency.

[0010] As a preferred technical solution of the present invention, the power-voltage droop relationship based on which the target voltage reference value is set in step S2 is:

[0011] ,

[0012] where, is the voltage reference value, is the rated voltage of the common DC bus, is the target total active power including the primary frequency modulation command, is the rated output active power of the energy storage power station, is the droop coefficient, and <0.

[0013] As a preferred technical solution of the present invention, the value of the droop coefficient is dynamically adjusted by the online identified equivalent inertia time constant of the energy storage power station and the set value of the frequency modulation dead zone; specifically: when the identified equivalent inertia time constant decreases, increase the value of | |; when the set value of the frequency modulation dead zone shrinks, decrease the value of | |, so as to adaptively enhance the ability to suppress the high-frequency disturbance components of the power grid on the premise of maintaining the stability of the DC voltage.

[0014] As a preferred technical solution of the present invention, in the fast power support mode in step S4, the currently maximum charge and discharge power capacity based on which the power command is allocated to each energy storage unit is updated in real time through an online optimization model including dynamic slack variables; this model takes maximizing the total available power adjustment margin of all energy storage units as the optimization goal, and takes the state of charge safety interval, temperature limit and the instantaneous overload capacity of the converter of each unit as constraints, and solves and updates the capacity value every other short period.

[0015] As a preferred embodiment of the present invention, in step S4, under the state-of-charge equalization mode, the magnitude of the compensation power allocated to each energy storage unit is not only positively correlated with the degree to which the state of charge of that unit deviates from the average value, but also negatively correlated with the accumulated charge-discharge cycle fatigue of that unit; wherein, the fatigue degree F is evaluated according to the following formula:

[0016] ,

[0017] in, This represents the cumulative throughput of the energy storage unit. For its rated capacity, These are the weighting coefficients. The average charge / discharge rate is calculated based on the historical operating data of this unit.

[0018] As a preferred technical solution of the present invention, the local control of each energy storage unit in step S5 adopts an anti-saturation adaptive voltage outer loop structure based on an interference observer. This structure estimates and feeds forward the common DC bus voltage coupling interference caused by the power change of adjacent units in real time through the interference observer, and automatically adjusts the voltage loop controller parameters to prevent integral saturation when the local power command reaches the limit.

[0019] As a preferred embodiment of the present invention, in step S1, when generating the primary frequency modulation active power command, a feedforward compensation term based on frequency change rate prediction is introduced, specifically including the following steps:

[0020] S1a. Real-time acquisition of AC grid frequency sampling values ​​for the current and several consecutive previous control cycles;

[0021] S1b. Construct and solve a least-squares fitting curve describing the current dynamic process of the system, and use the slope of the curve as a prediction of the rate of change of the power grid frequency in the next control cycle.

[0022] S1c: Obtain the total available spinning reserve capacity of the energy storage power station at the current moment;

[0023] S1d. Multiply the frequency change rate predicted in step S1b by a dynamic gain coefficient that is proportional to the total available spinning reserve capacity described in step S1c to obtain the feedforward compensation term.

[0024] S1e. The feedforward compensation term is superimposed on the fundamental command generated based on the active-frequency droop control loop to jointly constitute the primary frequency modulation active power command.

[0025] Wherein, the dynamic gain coefficient Determined by the following formula:

[0026] ,

[0027] In the formula, This is an adjustable scaling factor. The total available spinning reserve capacity at the current moment is obtained in step S1c. This refers to the rated power of the energy storage power station.

[0028] As a preferred technical solution of the present invention, the actual output power of the energy storage power station in step S2 is obtained by a multi-rate data fusion method, specifically including: high-speed sampling and processing of the AC side electrical quantities of the virtual synchronous machine control converter to obtain a first power estimate, low-speed sampling and processing of the DC side electrical quantities of each energy storage unit to obtain a second power estimate, and then using a Kalman filter to fuse the first power estimate and the second power estimate to obtain the actual output power value used to set the target voltage reference value.

[0029] As a preferred embodiment of the present invention, the preset threshold condition in step S3 is adaptively adjusted according to the dynamic process of the power grid frequency: in the initial stage when the power grid frequency changes rapidly, the threshold is adjusted to expand the triggering range of the fast power support mode; in the recovery stage when the frequency tends to stabilize, the threshold is adjusted to expand the triggering range of the state of charge equalization mode.

[0030] As a preferred technical solution of the present invention, the execution of steps S3 and S4 adopts a hybrid driving mechanism that combines event triggering and time triggering: when a significant change in the absolute value of the common DC bus voltage deviation is detected to cross a preset threshold, mode decision and power allocation are immediately triggered; if there is no significant change, periodic triggering is performed with a preset base period.

[0031] Compared with the prior art, the beneficial effects of the present invention are:

[0032] 1. This invention uses DC bus voltage deviation as the core control variable and establishes a dual-mode dynamic decision-making mechanism based on deviation threshold. It can quickly identify the intensity of the disturbance within milliseconds when the grid frequency disturbance occurs and immediately enter the fast power support mode. This mode prioritizes maximizing the power output speed and allocates power according to the real-time maximum charge and discharge capacity of each energy storage unit. This achieves the fastest and most sufficient response to the power deficit, thereby significantly improving the lowest frequency point, effectively suppressing the frequency drop rate, and shortening the system recovery time.

[0033] 2. This invention significantly mitigates voltage fluctuations on the common DC bus through a closed-loop DC voltage control architecture and anti-saturation local control based on an interference observer, ensuring the stability of the power transmission channel within the energy storage power station. Simultaneously, the introduced state-of-charge (SOC) balancing mode and its power allocation strategy incorporating battery fatigue compensation can intelligently adjust the charging and discharging states of each energy storage unit during the later stages or intervals of frequency regulation disturbances, effectively reducing SOC differences between units and slowing down battery aging rates. This provides a guarantee for the long-term, safe, and balanced operation of the energy storage power station.

[0034] 3. The present invention proposes several adaptive mechanisms, such as the downward coefficient dynamically adjusted according to the system's equivalent inertia and frequency regulation requirements, the operating mode switching threshold adaptively changing with the frequency dynamic process, and feedforward compensation based on least squares fitting and reserve capacity, which enable the control strategy to flexibly adapt to changes in the power grid's operating state and the needs of different frequency regulation scenarios. Combined with the hybrid driving mechanism of event triggering and time triggering, it ensures extremely fast response to sudden disturbances while reducing the computational overhead of the system in steady state, thereby improving the overall control efficiency and intelligence level.

[0035] 4. The actual output power of the power plant is obtained by adopting a multi-rate data fusion method, which combines the advantages of high-speed sampling on the AC side and low-speed sampling on the DC side. The optimal estimation is performed by Kalman filter, which effectively overcomes the influence of noise interference or faults at a single measurement point, and provides a reliable data foundation for subsequent accurate voltage reference value setting and power coordination allocation. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the overall process of the frequency regulation method for energy storage power stations under virtual synchronous machine control according to the present invention. Detailed Implementation

[0037] 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.

[0038] Example 1: Basic Example

[0039] This embodiment provides a frequency regulation method for an energy storage power station under virtual synchronous machine control. The method is applied to a 30MW / 60MWh lithium iron phosphate energy storage demonstration power station within the dispatching area of ​​a province in North China. The power station is connected to the 110kV AC grid through a 30MVA virtual synchronous machine control converter. The power station consists of 60 energy storage units with a rated power of 500kW and a rated capacity of 1MWh connected in parallel through a common DC bus with a rated voltage of 1500V.

[0040] This method executes the following control flow, such as Figure 1 As shown:

[0041] Step S1, Grid Frequency Tracking and Command Generation: Based on the frequency measurement unit inside the virtual synchronous machine control converter, the AC grid frequency is sensed in real time at a sampling frequency of 100Hz. Through the active-frequency droop control loop that simulates the rotor motion equation of the synchronous generator, the deviation between the grid frequency and the 50Hz rated frequency is converted into a primary frequency regulation active power command. The droop coefficient of this droop control loop is set to 4%.

[0042] Step S2, DC-side target voltage setting: Combining the primary frequency regulation active power command generated in step S1 with the actual output power of the energy storage power station, a target voltage reference value for the common DC bus is set. Specifically, the total active power currently actually output by the power station is obtained through a three-phase power transmitter installed on the AC output side of the virtual synchronous machine control converter. The primary frequency regulation active power command is added to the actual output power to obtain the target total active power. The voltage reference value is determined based on the power-voltage droop relationship.

[0043] Step S3, Dynamic Decision-Making of Operating Mode: Monitor the actual voltage of the common DC bus in real time at a sampling frequency of 10kHz, calculate the deviation between it and the target voltage reference value determined in step S2, set the first voltage deviation threshold to 15V and the second voltage deviation threshold to 5V, when the absolute value of the voltage deviation is greater than or equal to 15V, determine to enter the fast power support mode; when the absolute value of the voltage deviation is less than 15V but greater than or equal to 5V, determine to enter the state of charge equalization mode.

[0044] Step S4, Power Multi-mode Coordination and Allocation: Based on the operating mode determined in step S3, power allocation instructions are generated for each energy storage unit. In the fast power support mode, the priority is to maximize the frequency regulation response speed, and the allocation instructions are positively correlated with the current maximum charge and discharge power capability of each unit. In the state of charge balancing mode, the priority is to coordinate the state of charge of each energy storage unit, and compensation power is allocated to units whose state of charge deviates from the average value.

[0045] Step S5, Power Command Execution and Closed-Loop Control: Each energy storage unit receives power distribution commands through its local converter controller. The local controller adopts a dual-loop control structure, with the outer loop being the power loop and the inner loop being the current loop. By adjusting the charging and discharging power of each unit, the actual voltage of the common DC bus is stabilized near the target voltage reference value, ultimately achieving rapid and accurate support of the total output power of the energy storage power station for changes in grid frequency.

[0046] Example 2: Example with adaptive droop coefficient adjustment

[0047] Based on Example 1, this embodiment further defines the dynamic adjustment method of the droop coefficient K in step S2, taking a 20MW / 40MWh energy storage power station connected to a power grid in a certain region of East China as an example.

[0048] First, the equivalent inertial time constant of the energy storage power station is identified online using the recursive least squares method. The identification module takes the grid frequency change rate and the power station power response data as input and updates the identification results every 5 seconds. At the same time, it obtains the real-time frequency dead zone setting value issued by the grid dispatch, which can be dynamically adjusted within the range of ±0.05Hz to ±0.15Hz.

[0049] Sag coefficient The initial value is set to -0.01V / kW, and the dynamic adjustment rule is as follows: when the identified equivalent inertial time constant decreases from 10 seconds to 8 seconds, | The value of | increases from 0.01 to 0.012. When the power grid frequency regulation dead zone setting value decreases from ±0.1Hz to ±0.05Hz, | The value of | is reduced from 0.01 to 0.008. This adjustment strategy aims to enhance the suppression of high-frequency disturbances by using a more sensitive voltage-power response when the system inertia is weakened; when the frequency regulation requirements are more stringent, the response sensitivity is appropriately reduced to avoid frequent operation at the dead zone boundary, thereby adaptively enhancing the ability to suppress high-frequency disturbance components of the power grid while maintaining DC voltage stability.

[0050] Example 3: Example of Charge State Equilibrium Based on Fatigue Assessment

[0051] Based on Example 1, this embodiment further refines the state-of-charge balancing mode in step S4 and applies it to a 5MW / 10MWh cascaded energy storage system in an industrial park in South China.

[0052] In the state-of-charge (POC) balancing mode, the central controller first calculates the average POC value of all online-operating energy storage units, and then... Each energy storage unit calculates its compensation power, which is determined by two parts: the first part is positively correlated with the degree to which the unit's state of charge deviates from the average value, with a proportionality coefficient set at 0.5 MW / %SOC; the second part is negatively correlated with the unit's current accumulated charge-discharge cycle fatigue.

[0053] Fatigue The evaluation formula is as follows:

[0054] ,

[0055] in, This represents the cumulative throughput of the unit since it was put into operation, read through its local battery management system (BMS), in kWh. Its rated capacity, in kWh; This is the weighting coefficient, which is empirically set to 0.1 based on the battery type (lithium iron phosphate in this example); The average charge / discharge rate, expressed in C, is calculated based on the unit's operating data over the past 24 hours.

[0056] This formula has the same mathematical form and physical meaning as the fatigue assessment formula mentioned above, where the subscripts... Used only for identifying the first One energy storage unit.

[0057] For example, if a unit's cumulative throughput is twice its rated capacity, and the historical average capacity is 0.2C, then its fatigue level... The controller will prioritize allocating more compensation power to units with lower fatigue levels to achieve lifespan balancing. Ultimately, the compensation power command for that unit will be... Determined by the following formula:

[0058] ,

[0059] in, The average state of charge (expressed as a percentage) of all currently online energy storage units;

[0060] For the first Current state of charge (in percentage) of each energy storage unit;

[0061] This is the state-of-charge deviation adjustment coefficient;

[0062] This is the fatigue adjustment coefficient;

[0063] For the first The fatigue assessment value of each energy storage unit is calculated according to the aforementioned formula.

[0064] Example 4: Feedforward Compensation Example Based on Least Squares Prediction and Dynamic Gain

[0065] This example elaborates in detail the specific implementation process of introducing the feedforward compensation term in step S1 of Example 1, taking a 50MW / 100MWh energy storage power station supporting a high-proportion new energy base in a certain northwestern region as an example.

[0066] Step S1a: The frequency sampling module of the controller acquires the AC grid frequency sampling values of the current and the previous 9 control cycles at a frequency of 1kHz, forming a time series containing 10 data points.

[0067] Step S1b: Taking the sequence number of the time series as the independent variable and the frequency value as the dependent variable, a first-order linear model is constructed, and the least squares method is used for fitting to solve the slope of the fitting straight line. This slope is the predicted value of the grid frequency change rate for the next control cycle, with the unit of Hz / s.

[0068] Step S1c: Through the power station energy management system EMS, the sum of the available charge-discharge powers of all online energy storage units at the current moment is acquired in real time as the total available spinning reserve capacity .

[0069] Step S1d: Multiply the predicted frequency change rate by a dynamic gain coefficient , to obtain the feedforward compensation term. The dynamic gain coefficient is determined by the following formula:

[0070] ,

[0071] where is an adjustable proportional coefficient, set to 2.0 according to the system's requirement for response speed; is the rated power of the energy storage power station, 50MW. For example, if the current total reserve capacity is 20MW and the predicted frequency change rate is -0.5Hz / s, then , the feedforward compensation term = -0.5Hz / s × 0.8 = -0.4Hz / s. This value means that the predicted frequency will decrease, and the energy storage power station needs to additionally generate active power.

[0072] Step S1e: Add the calculated feedforward compensation term to the fundamental wave command generated by the traditional active-frequency droop control link to jointly form the final primary frequency regulation active power command and send it to the subsequent link.

[0073] Example 5: Power Measurement Example of Multi-Rate Data Fusion

[0074] This embodiment details the specific method for obtaining the actual output power in step S2 of embodiment 1. The system configuration is as follows: The AC side of the virtual synchronous machine control converter uses LEM's CT 0.5 series high-precision current sensor and VT 1000 series voltage sensor for measurement; the DC side of each energy storage unit uses the domestic Huali Technology HL-DCM series DC power metering module.

[0075] Step S21 (High-speed sampling and processing): Synchronously sample the three-phase currents and voltages (A, B, and C) on the AC side of the virtual synchronous machine control converter at a sampling frequency of 20kHz. Obtain the three-phase instantaneous power using an instantaneous power calculation algorithm. Then, pass the sampled power through a first-order low-pass digital filter with a cutoff frequency of 100Hz to filter out switching frequency and high-frequency noise, obtaining the first power estimate P. est1 The update frequency is 10kHz.

[0076] Step S22 (Low-speed sampling and processing): Sample the voltage and current on the DC side of each energy storage unit at a sampling frequency of 1kHz. Calculate the DC side power of each unit and aggregate the data via the CAN bus to the central controller for summation, obtaining the second power estimate P. est2 The update frequency is 1kHz.

[0077] Step S23 (Kalman Filter Fusion): Establish a discrete Kalman filter, where the system state variable is the true value P of the actual output power of the energy storage power station. real , will P est1 and P est2 As two asynchronous observation inputs with different noise statistical characteristics, the filter performs optimal data fusion based on the update rate of the two and the preset observation noise covariance matrix, and finally outputs the optimal estimate of the actual output power used to set the target voltage reference value. This method effectively improves the noise immunity and reliability of power measurement and avoids control inaccuracy caused by single measurement point failure or disturbance.

[0078] Example 6: Frequency-Dynamic Threshold Adaptive Adjustment Example

[0079] This embodiment, based on embodiment 1, provides a detailed explanation of the adaptive adjustment of the preset threshold conditions in step S3.

[0080] The system continuously monitors the rate of change of the power grid frequency and sets a rapid change judgment threshold df / dt. threshold It is 0.3 Hz / s.

[0081] Initial stage judgment and threshold adjustment: When |df / dt| > 0.3 Hz / s is detected, it is determined that the grid frequency has entered the initial stage of rapid change. At this time, the controller automatically temporarily relaxes the first voltage deviation threshold from 15V to 25V and tightens the second voltage deviation threshold from 5V to 3V. This adjustment strategy aims to expand the trigger range of the fast power support mode, so that the energy storage power station can release or absorb power earlier and more, prioritize the fast power support speed to the grid, and suppress abrupt frequency changes.

[0082] Recovery Phase Judgment and Threshold Adjustment: When the grid frequency remains stable within the ±0.02Hz dead zone of 49.98Hz to 50.02Hz for 5 consecutive seconds, the recovery equilibrium phase is determined. At this time, the controller restores the first voltage deviation threshold to 15V and relaxes the second voltage deviation threshold from 5V to 8V. This adjustment aims to expand the trigger range of the state of charge equalization distribution mode, promote energy balance adjustment among energy storage units after the initial resolution of the frequency crisis, and prepare for the next disturbance.

[0083] Example 7: Hybrid Trigger-Driven Mechanism Example

[0084] This embodiment illustrates the execution driving mechanism of steps S3 and S4, and controls the reference time period T of the system. base Set to 20ms.

[0085] At the beginning of each reference cycle, the system checks whether the absolute value of the common DC bus voltage deviation has changed significantly since the previous cycle, crossing the first threshold (15V) or the second threshold (5V). The judgment is based on the change in the sign of the deviation value or the crossing from one side of the threshold to the other.

[0086] Event triggering: If such a significant change event is detected, the periodic loop is immediately broken, triggering an immediate mode re-decision and power reallocation process. This ensures that the control response is almost instantaneous in the event of a sudden change in operating conditions.

[0087] Time-triggered: If no significant change event is detected, a regular mode decision and power allocation process is triggered after the current baseline cycle ends. In the fast power support mode, the shortest power adjustment cycle is 5ms, which is much shorter than the base cycle, ensuring the speed of this mode. In the state-of-charge equalization mode, a cycle of 20ms is sufficient to complete fine adjustment.

[0088] This hybrid driving mechanism, which combines event-triggered and time-triggered mechanisms, ensures that the system can respond quickly to sudden events while avoiding high-frequency operations when there are no changes, thus reducing system overhead.

[0089] Comparison with Example 1: Traditional virtual synchronous machine control (without the multi-mode coordination and voltage deviation decision-making of this invention)

[0090] Using the same energy storage power station model in North China as in Example 1, the mode decision-making link based on DC voltage deviation and the multi-mode power coordination and allocation link are eliminated in the comparative example. The virtual synchronous machine control converter generates power commands only based on the grid frequency droop characteristics, and adopts a simple proportional allocation strategy based on the energy storage unit capacity to issue the total power command to each unit. Each unit independently executes local power control, without a collaborative closed loop based on the common DC bus voltage. This comparative example is used to simulate the traditional VSG control strategy commonly found in the prior art.

[0091] Compare with Example 2: Fixed threshold and fixed period trigger control

[0092] Based on Example 1, the first and second voltage deviation thresholds in step S3 are fixed at 15V and 5V respectively, and are not adjusted dynamically with frequency. At the same time, the execution mechanism of steps S3 and S4 is changed to a pure fixed-cycle trigger with a fixed cycle of 20ms, and the event trigger mechanism is canceled. This comparative example is used to verify the effectiveness of the threshold adaptive and hybrid triggering mechanism in this invention.

[0093] Experimental verification and data analysis

[0094] To verify the effectiveness of the method proposed in this invention, a regional power grid simulation model including thermal power units, wind power, loads and energy storage power stations was constructed based on the MATLAB / Simulink and PLECS joint simulation platform. The methods represented by Example 1, Comparative Example 1 and Comparative Example 2 were tested respectively. The test scenario simulated a 0.5Hz frequency drop event caused by a large-capacity wind turbine disconnecting from the grid in a certain area of ​​North China.

[0095] Table 1 Comparison of key performance indicators for frequency event response under different control methods

[0096] Key performance indicators Compare with Example 1 (Traditional VSG) Compare with Example 2 (fixed threshold / period) Example 1 (Method of the Invention) Improvement rate (vs. Example 1) Lowest frequency point (Hz) 49.47 49.56 49.63 +0.16 Hz Maximum frequency drop rate (Hz / s) -0.92 -0.68 -0.45 Reduced by 51.1% Time (s) required for the frequency to recover to 49.9 Hz 10.2 7.8 5.6 Shortened by 45.1% The frequency modulation process always has a peak active power output (MW). 24.3 26.1 28.7 An increase of 18.1%. Maximum dispersion of SOC after the event (%) 12.3 7.5 4.2 Reduced by 65.9% Maximum fluctuation rate of DC bus voltage (%) ±3.2％ ±1.8％ ±1.1％ Reduced by 65.6% Average response time to control commands (ms) 25.0 25.0 8.5 (Event Triggered) Reduced by 66.0%

[0097] As can be seen from the table above:

[0098] I. Comprehensive Improvement in Frequency Support Performance

[0099] 1. Significant improvement in the lowest frequency point: Example 1 raises the lowest frequency point to 49.63Hz, which is 0.16Hz higher than the traditional VSG method (49.47Hz). This improvement is directly due to the DC voltage deviation-dual-mode power distribution architecture established in this invention. When the frequency starts to drop, the DC bus voltage deviates due to the rapid power output. The system immediately enters the rapid power support mode and distributes power according to the real-time maximum adjustable capability of each energy storage unit, realizing rapid and maximum replenishment of the power deficit.

[0100] 2. Effective suppression of frequency drop rate: The maximum frequency change rate was reduced from -0.92Hz / s to -0.45Hz / s, a decrease of 51.1%. This significant improvement is due to the frequency change rate prediction feedforward compensation based on least squares fitting. This feedforward link can predict the frequency change trend in advance and inject additional power support commands before the frequency starts to accelerate down, smoothing the frequency drop trajectory and reducing the transient impact of the system.

[0101] 3. Increased frequency recovery speed: The time to recover the frequency to 49.9Hz was reduced from 10.2 seconds to 5.6 seconds, a reduction of 45.1%. This indicates that the control method of the present invention not only responds quickly in the early stage of disturbance, but also maintains the continuous support capability of the energy storage power station through timely intervention of the state of charge balancing mode during the system recovery stage, avoiding premature exit from frequency regulation due to some units approaching the limit of SOC.

[0102] II. Optimization of the operating status of the energy storage power station itself

[0103] 1. Significant improvement in SOC balancing effect: After the frequency regulation event, the maximum SOC dispersion among the energy storage units decreased from 12.3% to 4.2%, which directly proves the effectiveness of the state-of-charge balancing allocation strategy that integrates fatigue assessment. This strategy is automatically activated in the later stage of frequency disturbances. By accurately compensating the SOC deviation units and taking into account the historical operating fatigue of the batteries, it realizes the intelligent redistribution of energy within the power station, thus reserving a balanced response capability for subsequent continuous disturbances.

[0104] 2. Enhanced DC bus voltage stability: The maximum fluctuation rate of DC bus voltage was reduced from ±3.2% to ±1.1%. This improvement verifies the effectiveness of the anti-saturation adaptive voltage outer loop control based on the interference observer. This control structure can estimate and compensate for the coupling interference caused by the power change of adjacent units in real time, and automatically adjust the control parameters when the power command reaches the limit, ensuring the high stability of the DC side voltage and providing a guarantee for the safe operation of the converter.

[0105] III. Optimization of Real-Time Performance of the Control System

[0106] The average response delay of control commands was reduced from 25ms under fixed-cycle triggering to 8.5ms under event triggering. This 66% delay reduction demonstrates the superiority of the hybrid triggering mechanism. When the DC voltage deviation is detected to cross the threshold, the system immediately triggers the control decision, avoiding the response lag caused by fixed-cycle waiting. This hybrid mechanism of "event-driven + time-based" ensures the real-time performance of the system while taking into account the operating efficiency.

[0107] IV. Comparative Analysis with Other Improvement Methods

[0108] The performance of Comparative Example 2 (fixed threshold / period) is between that of the conventional VSG and the present invention. This further illustrates the importance of the threshold adaptive adjustment and hybrid triggering mechanism in the present invention. Fixed threshold cannot optimize the mode switching timing according to the frequency dynamic process, and fixed period triggering cannot capture abrupt events between two samplings. These limitations are verified in the data of Comparative Example 2.

[0109] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A frequency regulation method for an energy storage power station under virtual synchronous machine control, characterized in that, The energy storage power station is connected to the AC grid via a virtual synchronous machine-controlled converter and consists of multiple energy storage units connected in parallel via a common DC bus; the method executes the following control flow: S1. Grid frequency tracking and command generation: Based on the virtual synchronous machine control converter, the AC grid frequency is sensed and a primary frequency regulation active power command is generated accordingly. S2, DC side target voltage setting: Based on the primary frequency regulation active power command and the actual output power of the energy storage power station, set the target voltage reference value of the common DC bus; S3. Dynamic decision-making for operating mode: Based on the deviation between the real-time voltage of the common DC bus and its target voltage reference value, as well as the preset threshold conditions, the system dynamically switches between fast power support mode and state of charge balance mode. S4. Power multi-mode coordinated allocation: Based on the operating mode determined in step S3, generate corresponding power allocation instructions for each energy storage unit; wherein, the fast power support mode prioritizes maximizing the frequency regulation response speed, and the state of charge balancing mode prioritizes coordinating the state of charge of each energy storage unit. S5. Power Command Execution and Closed-Loop Control: Each energy storage unit adjusts its charging and discharging power through local control according to the power command it is allocated, so that the actual voltage of the common DC bus is stabilized near the target voltage reference value, and the total output power of the energy storage power station actively supports the grid frequency.

2. The method according to claim 1, characterized in that, In step S2, the power-voltage droop relationship on which the target voltage reference value is set is as follows: ， in, This is the voltage reference value. The rated voltage of the common DC bus. The target total active power includes the primary frequency regulation command. The rated output active power of the energy storage power station The droop coefficient is... and <0.

3. The method according to claim 2, characterized in that, The sag coefficient is dynamically adjusted according to the equivalent inertia time constant of the energy storage power station identified online and the set value of the frequency modulation dead zone. Specifically, when the identified equivalent inertia time constant decreases, the value of | | is increased; when the set value of the frequency modulation dead zone shrinks, the value of | | is decreased, so as to adaptively enhance the ability to suppress the high-frequency disturbance components of the power grid on the premise of maintaining the stability of the DC voltage.

4. The method according to claim 1, characterized in that, In step S4, under the fast power support mode, the current maximum chargeable and dischargeable power capability on which the power command is allocated to each energy storage unit is updated in real time through an online optimization model containing dynamic relaxation variables. The model aims to maximize the total available power adjustment margin of all energy storage units, and is constrained by the safe charge state range, temperature limit and instantaneous overload capability of each unit. The capability value is solved and updated every short cycle.

5. The method according to claim 1, characterized in that, In step S4, under the state-of-charge (POC) balancing mode, the magnitude of the compensation power allocated to each energy storage unit is not only positively correlated with the degree to which the unit's POC deviates from the average value, but also negatively correlated with the unit's accumulated charge-discharge cycle fatigue. The fatigue degree F is evaluated using the following formula: ， in, This represents the cumulative throughput of the energy storage unit. For its rated capacity, These are the weighting coefficients. The average charge / discharge rate is calculated based on the historical operating data of this unit.

6. The method according to claim 1, characterized in that, In step S5, the local control of each energy storage unit adopts an anti-saturation adaptive voltage outer loop structure based on an interference observer. This structure estimates and feeds forward the common DC bus voltage coupling interference caused by power abrupt changes in adjacent units in real time through the interference observer, and automatically adjusts the voltage loop controller parameters to prevent integral saturation when the local power command reaches the limit.

7. The method according to claim 1, characterized in that, In step S1, when generating the primary frequency modulation active power command, a feedforward compensation term based on frequency change rate prediction is introduced, specifically including the following steps: S1a. Real-time acquisition of AC grid frequency sampling values ​​for the current and several consecutive previous control cycles; S1b. Construct and solve a least-squares fitting curve describing the current dynamic process of the system, and use the slope of the curve as a prediction of the rate of change of the power grid frequency in the next control cycle. S1c: Obtain the total available spinning reserve capacity of the energy storage power station at the current moment; S1d. Multiply the frequency change rate predicted in step S1b by a dynamic gain coefficient that is proportional to the total available spinning reserve capacity described in step S1c to obtain the feedforward compensation term. S1e. The feedforward compensation term is superimposed on the fundamental command generated based on the active-frequency droop control loop to jointly constitute the primary frequency modulation active power command. Wherein, the dynamic gain coefficient Determined by the following formula: ， In the formula, This is an adjustable scaling factor. The total available spinning reserve capacity at the current moment is obtained in step S1c. This refers to the rated power of the energy storage power station.

8. The method according to claim 1, characterized in that, In step S2, the actual output power of the energy storage power station is obtained through a multi-rate data fusion method, which specifically includes: high-speed sampling and processing of the AC side electrical quantities of the virtual synchronous machine control converter to obtain a first power estimate; low-speed sampling and processing of the DC side electrical quantities of each energy storage unit to obtain a second power estimate; and then using a Kalman filter to fuse the first power estimate and the second power estimate to obtain the actual output power value used to set the target voltage reference value.

9. The method according to claim 1, characterized in that, The preset threshold condition in step S3 is adaptively adjusted according to the dynamic process of the power grid frequency: in the initial stage when the power grid frequency changes rapidly, the threshold is adjusted to expand the triggering range of the fast power support mode; in the recovery stage when the frequency tends to stabilize, the threshold is adjusted to expand the triggering range of the state of charge equalization mode.

10. The method according to claim 1, characterized in that, The execution of steps S3 and S4 adopts a hybrid driving mechanism that combines event triggering and time triggering: when a significant change in the absolute value of the common DC bus voltage deviation is detected that crosses a preset threshold, mode decision and power allocation are immediately triggered; if there is no significant change, periodic triggering is performed with a preset base period.

Images