Self-adaptive control method and system for primary frequency modulation of power grid

Abstract

The invention discloses a self-adaptive control method and system for primary frequency modulation of a power grid, and relates to the technical field of power system automation and new energy grid-connected control, and the method comprises the following steps: when the energy storage residual electric quantity is within the range of 15%-25%, the energy storage output power is greater than 30% of the rated power, and the absolute value of the frequency deviation of the power grid is greater than 0.08 Hz, controlling the frequency of the power grid; the fixed virtual inertia is replaced with a two-dimensional matrix changing in real time along with the residual electric quantity and the direct current bus voltage, virtual capacitance current in the direction for preventing the direct current bus voltage from dropping is synchronously generated, and the magnitude of the virtual capacitance current is in direct proportion to the change rate of the direct current bus voltage; in the cliff interval with the energy storage residual electric quantity of 15%-25%, the phenomenon of undervoltage locking at the moment of primary frequency modulation high-power discharge caused by fixed virtual inertia is thoroughly eliminated, the frequency secondary collapse probability is reduced to zero, and no hardware is added.

Description

An adaptive control method and system for primary frequency regulation of a power grid Technical Field

[0001] This invention relates to the field of power system automation and new energy grid connection control technology, specifically to an adaptive control method and system for primary frequency regulation of a power grid. Background Technology

[0002] Currently, primary frequency regulation of the power grid still relies mainly on the inherent speed regulation characteristics of synchronous generators, and their droop coefficient remains fixed once set. With the integration of a high proportion of renewable energy into the grid, system inertia decreases and disturbance frequency increases. Fixed droop strategies are prone to "over-adjustment-reverse adjustment" oscillations and have poor adaptability to continuous disturbances. To address this, some wind farms have attempted to use virtual inertia combined with droop control for primary frequency regulation: at the moment of frequency drop, the wind turbine uses rotor kinetic energy or reserved load shedding power to temporarily increase power, and then continues to support the system according to a fixed droop coefficient. However, when the wind turbine is in a deep load shedding state of around 20%, the reserved power is limited; if the energy storage system is simultaneously in the 15%~25% SOC "cliff range," its maximum discharge power is drastically compressed due to the DC bus voltage approaching its lower limit. At this point, if a fixed virtual inertia and droop coefficient are still used, the high-power discharge command at the beginning of frequency regulation will instantly lower the DC bus voltage, causing the energy storage converter to be undervoltage locked out, forcing the wind turbine to exit primary frequency regulation, resulting in a "secondary collapse" of the system frequency.

[0003] For example, Chinese patent CN111864807B proposes a primary frequency regulation method for wind turbines based on nonlinear droop control. By introducing a dynamic droop coefficient, it achieves adaptive adjustment of the droop coefficient, thereby effectively suppressing the power oscillation problem that may occur during frequency regulation in traditional droop control. However, the frequency regulation strategy of this patent is mainly based on a fixed droop characteristic curve and does not fully consider the dynamic adaptability of the wind turbine to the frequency response under unloaded operation. This results in a lag in active power regulation response and insufficient regulation accuracy when the frequency fluctuation is large, and it is prone to power oscillation, affecting the stability of the grid frequency. In addition, under the condition of low energy storage and unloaded wind turbine operation, the fixed inertia and droop coefficient cause the DC bus voltage to drop instantaneously, and the energy storage converter is disconnected due to undervoltage lockout, which in turn causes a secondary frequency drop.

[0004] Therefore, there is an urgent need for a primary frequency regulation adaptive control method that can adjust the droop coefficient and virtual inertia in real time according to the frequency deviation under extreme operating conditions such as wind turbine unloading and energy storage nearing undervoltage, completely eliminating instantaneous undervoltage lockout and preventing secondary frequency collapse, without adding any new hardware. Summary of the Invention

[0005] This invention proposes an adaptive control method and system for primary frequency regulation of the power grid. In the cliff range of 15% to 25% of the remaining energy storage capacity, it completely eliminates the instantaneous undervoltage blocking phenomenon of high-power discharge in primary frequency regulation caused by fixed virtual inertia, reduces the probability of secondary frequency collapse to zero, and does not add any hardware.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an adaptive control method for primary frequency regulation of a power grid, wherein under the conditions that the remaining energy storage capacity is in the range of 15% to 25%, the energy storage output power is greater than 30% of its rated power, and the absolute value of the power grid frequency deviation is greater than 0.08Hz, a fixed virtual inertia is replaced with a two-dimensional matrix that changes in real time with the remaining energy and the DC bus voltage, and a virtual capacitor current is generated synchronously in a direction that prevents the DC bus voltage from dropping, wherein the magnitude of the virtual capacitor current is proportional to the rate of change of the DC bus voltage.

[0007] In this technical solution, the energy storage output power command is adjusted by using the matrix and the virtual capacitor current together, so that the DC bus voltage rises to no less than 92% of the rated value within 200ms, thereby preventing the energy storage converter from being locked due to undervoltage.

[0008] Preferably, the upper left element of the two-dimensional matrix decreases as the remaining power decreases, and the lower right element increases as the remaining power increases. The upper right element and the lower left element are opposites of each other and are both proportional to the ratio of the DC bus voltage to the rated voltage, so as to form a power-frequency-voltage cross-damping.

[0009] Preferably, replacing the fixed virtual inertia with a two-dimensional matrix that changes in real time with the remaining power and DC bus voltage includes: dividing the disturbance into large disturbance, medium disturbance, and small disturbance according to the absolute value of the grid frequency deviation and its rate of change; replacing the fixed virtual inertia with a two-dimensional matrix only when the disturbance level reaches medium or large disturbance; and keeping the original fixed virtual inertia unchanged for small disturbance.

[0010] Preferably, the absolute value of the small disturbance deviation is no greater than 0.08 Hz and the absolute value of the rate of change is no greater than 0.3 Hz / s, the absolute value of the medium disturbance deviation is greater than 0.08 Hz and no greater than 0.3 Hz / s and the absolute value of the rate of change is no greater than 0.5 Hz / s, and the absolute value of the large disturbance deviation is greater than 0.3 Hz / s or the absolute value of the rate of change is greater than 0.5 Hz / s.

[0011] Preferably, corresponding fusion modes are used to obtain power adjustment amounts for large disturbances, medium disturbances, and small disturbances, including: a mode with both fuzzy judgment and virtual synchronizer characteristics is used for large disturbances, a mode with both prediction and adaptive correction characteristics is used for medium disturbances, and only an adaptive correction mode is used for small disturbances.

[0012] Preferably, the virtual capacitor current is equal to the product of 200μF and the DC bus voltage change rate, wherein the 200μF is set by software and no additional physical capacitor is required.

[0013] Preferably, if the DC bus voltage is not lower than 92% of the rated value within 200ms, or if the remaining charge falls outside the 15%~25% range, the matrix control will automatically exit and the fixed virtual inertia will be restored.

[0014] Preferably, while entering matrix control, the total response time of a single frequency modulation is kept to be no more than 40ms, the frequency recovery time is no more than 400ms, and the steady-state frequency deviation is no more than 0.05Hz.

[0015] The present invention also adopts the following technical solution: an adaptive control system for primary frequency regulation of a power grid, which realizes the above-mentioned adaptive control method for primary frequency regulation of a power grid, including: a sampling module acquiring DC bus voltage, output power, power grid frequency deviation and its rate of change at a rate of not less than 1kHz; a judgment module identifying whether the following conditions are met simultaneously: remaining power of 15%~25%, output power greater than 30% of rated power, and absolute value of frequency deviation greater than 0.08Hz; when the judgment module gives a positive result, the control module replaces the fixed virtual inertia with a two-dimensional matrix and generates a virtual capacitor current to jointly adjust the output power command.

[0016] Preferably, the control module automatically restores the fixed virtual inertia after the voltage recovers or the remaining power drops out of the 15%~25% range, and the switching process is shock-free.

[0017] The beneficial effects of this invention are: 1) It completely eliminates undervoltage lockout in the power cliff region by using matrixed virtual inertia and virtual capacitor current in coordination; 2) The control still maintains millisecond-level response without sacrificing speed; 3) It reduces current surge and extends battery life; 4) All functions are implemented in software, requiring no additional hardware, and existing equipment can be directly upgraded. Attached Figure Description

[0018] Figure 1 is a flowchart of an adaptive control method for primary frequency regulation of a power grid according to the present invention.

[0019] Figure 2 is a comparison chart of RoCoF in Embodiment 3 of the present invention. Detailed Implementation

[0020] Example 1 This example provides an adaptive control method for primary frequency regulation of a power grid.

[0021] In actual operation, when the remaining battery charge is between 15% and 25%, the energy storage system will encounter a special operating condition. At this time, the internal resistance of the battery increases significantly, and the DC bus voltage's ability to withstand power surges decreases drastically.

[0022] In this situation, if the grid frequency suddenly drops and energy storage is needed to output a large amount of power instantly, the traditional fixed virtual inertia control method will cause the DC bus voltage to drop rapidly, which can easily trigger the undervoltage protection of the converter, causing the equipment to be locked out, and then causing a secondary frequency collapse.

[0023] In past engineering practices, this phenomenon was simply attributed to insufficient battery power and did not receive enough attention, nor was it systematically analyzed and resolved.

[0024] This embodiment describes in detail a control method specifically designed to address the aforementioned problems.

[0025] The core idea of ​​this method is to activate a special control mode when the remaining battery power is between 15% and 25%, the energy storage output power exceeds 30% of the rated power, and the absolute value of the grid frequency deviation exceeds 0.08Hz.

[0026] In this mode, the originally fixed virtual inertia is replaced by a matrix that can change in real time.

[0027] This matrix consists of four elements, each of which is calculated and adjusted in real time based on the current remaining battery power and DC bus voltage.

[0028] The top-left element of the matrix decreases as the remaining battery power decreases, while the bottom-right element increases as the remaining battery power increases. The top-right and bottom-left elements are dynamically adjusted based on the ratio of the DC bus voltage to the rated voltage.

[0029] This matrix-style virtual inertia can better adapt to the characteristics of the battery in the power cliff region, providing more flexible frequency support capabilities.

[0030] Meanwhile, a virtual capacitor current is also introduced into the control method. This current is not achieved by actually adding a hardware capacitor, but is calculated in real time by a software algorithm based on the rate of change of the DC bus voltage.

[0031] Specifically, the change in DC bus voltage is calculated every millisecond, multiplied by the equivalent capacitance value of 200μF, to obtain the magnitude of the virtual capacitor current, and its direction is set to prevent voltage drop.

[0032] This virtual capacitor current is directly superimposed on the AC side current command of the converter, which can provide additional support at the moment when the voltage begins to drop, effectively mitigating the rate of voltage drop on the DC bus.

[0033] When the DC bus voltage remains above 92% of the rated voltage for 200ms continuously, or when the remaining battery charge drops out of the range of 15% to 25%, the system will automatically exit this special control mode and revert to the traditional fixed virtual inertia control.

[0034] The entire control process is completed independently by the local controller of the energy storage converter, without the need to communicate with the host computer or add any additional hardware.

[0035] This control method not only solves the undervoltage lockout problem in the power cliff area, but also maintains the original response speed and stability of the system, ensuring that the energy storage system can operate reliably under various operating conditions.

[0036] Example 2, based on Example 1, further describes the application of the above control method to a complete energy storage system.

[0037] The energy storage system consists of four energy storage branches, each of which includes a battery compartment with a rated capacity of 500kWh and a converter with a rated power of 500kW. All branches are connected in parallel to the same 380V AC bus.

[0038] The system operates in a high-proportion renewable energy scenario, with a photovoltaic penetration rate of 65%, a load peak-valley difference of 50%, and more than 120 frequency disturbances per day.

[0039] In this environment, energy storage systems need to perform frequent charging and discharging operations to maintain the stability of the grid frequency.

[0040] To verify the effectiveness of this control method, the initial remaining battery charge of the four energy storage branches was artificially set to 18%, 20%, 22%, and 24%, respectively, placing them all within the charge cliff zone. When the system detected a drop in grid frequency to 49.70Hz due to a sudden change in photovoltaic output, each of the four converters independently determined that the conditions for entering a special control mode were met. At this point, the control algorithm within each converter automatically replaced the fixed virtual inertia with a real-time changing matrix and generated a corresponding virtual capacitor current. Because the remaining battery charge of each branch is slightly different, the elements of the matrix will also differ, allowing the output power command of each branch to be finely adjusted according to its own battery state.

[0041] Throughout the entire control process, the host computer's energy management system is only responsible for distributing the total frequency modulation power command equally to the four energy storage branches, with each branch undertaking the same power support task.

[0042] The voltage anti-blocking control of individual branches is entirely achieved by local matrix and virtual capacitor current, without the need to upload the battery's remaining power information to the host computer, nor does the host computer need to perform complex coordination calculations.

[0043] This distributed control method not only reduces the communication burden but also improves the system's response speed and reliability. After 200ms of control adjustment, the DC bus voltage of all branches has recovered to no less than 92% of the rated voltage, and the system frequency has also recovered to above 49.95Hz.

[0044] At this point, each converter exits the special control mode in turn, reverts to traditional fixed virtual inertia control, and continues to perform normal frequency regulation tasks.

[0045] Through long-term operation and testing, the system did not experience any undervoltage lockout events during high-power discharge in the power cliff zone, the number of frequency non-compliance points remained at zero, the maximum steady-state error of the power output curves of the four energy storage branches was only 1.6%, and the difference in battery cycle life was also controlled within 4%.

[0046] These indicators are all superior to traditional control methods, which fully demonstrates that this control method can maintain the high-precision power distribution and long-life operation characteristics of the system while solving the undervoltage lockout problem in the power cliff area.

[0047] The entire control process requires no additional hardware or changes to the existing communication architecture, offering excellent compatibility and scalability, and is suitable for energy storage systems of various sizes.

[0048] Example 3: Adaptive Control Method for Primary Frequency Regulation of Power Grid In actual operation, it needs to deal with frequency disturbances of varying intensities. In order to maintain the stability and response speed of the energy storage system under various disturbance conditions, this example provides an adaptive control method for primary frequency regulation of power grid. As shown in Figure 1, the disturbance is divided into three intervals—large, medium, and small—based on the absolute value of the power grid frequency deviation and its rate of change, and the control strategy is adjusted according to the disturbance level.

[0049] During normal grid operation, the control unit of the energy storage system samples the grid frequency at 1 kHz and calculates the frequency deviation Δf and its rate of change df / dt. The frequency deviation Δf is the difference between the current frequency and the rated frequency of 50 Hz, and the rate of change df / dt is the derivative of the frequency deviation with respect to time, reflecting the speed of frequency change. Based on the magnitude of these two parameters, the control unit classifies disturbances into three levels: Small disturbance: |Δf| ≤ 0.08 Hz and |df / dt| ≤ 0.3 Hz / s; Medium disturbance: 0.08 Hz < |Δf| ≤ 0.3 Hz and |df / dt| ≤ 0.5 Hz / s; Large disturbance: |Δf| > 0.3 Hz or |df / dt| > 0.5 Hz / s. These classification criteria are based on grid operation experience and the physical characteristics of the energy storage equipment. Small disturbances correspond to daily load fluctuations or small changes in new energy output, where the grid frequency deviates little from the rated value and changes slowly. Medium-sized disturbances are usually caused by large load switching or sudden changes in renewable energy output, with moderate frequency deviations and rates of change. Large disturbances, on the other hand, may be caused by serious faults such as line tripping, large generating units disconnecting from the grid, or extreme weather leading to sudden changes in renewable energy output, with large frequency deviations and drastic changes.

[0050] The control unit determines the current disturbance level in real time and adopts different control strategies according to the level: when a small disturbance is detected, the system maintains the normal control mode, using a fixed virtual inertia H0 (e.g., 0.08s) and normal PID parameters. At this time, the energy storage system only needs to provide limited power support to maintain the frequency within the allowable range. During small disturbances, the matrix virtual inertia and virtual capacitor current are not activated to avoid unnecessary control actions and reduce equipment wear and energy loss.

[0051] When a moderate disturbance is detected, the system initiates adaptive control mode. First, it checks whether the remaining energy storage capacity is within the 15%–25% energy cliff zone, and simultaneously checks whether the output power exceeds 30% of the rated power. If both conditions are met, the matrix virtual inertia and virtual capacitor current are further activated to prevent undervoltage lockout triggered by a DC bus voltage drop due to high-power discharge in the energy cliff zone. The four elements of the matrix virtual inertia are updated in real-time based on the current remaining energy and DC bus voltage, forming cross-damping to improve the system's adaptability to moderate disturbances. The virtual capacitor current is calculated based on the DC bus voltage change rate, with its direction aimed at preventing voltage drops, and its magnitude is 200μF multiplied by the voltage change rate. During moderate disturbances, the matrix and virtual capacitor current work together to ensure that the DC bus voltage recovers to at least 92% of the rated voltage within 200ms, thereby avoiding lockout while maintaining frequency recovery speed and stability.

[0052] Upon detecting a large disturbance, the system immediately enters a high-enhancement control mode. Regardless of whether the remaining power is in the cliff zone, matrix virtual inertia and virtual capacitor current are forcibly activated, and the matrix gain and virtual capacitor current coefficient are further increased. During large disturbances, the scaling factor K of the matrix elements is set to 1.4, and the virtual capacitor current coefficient is increased to 280μF to provide greater damping and voltage support. Simultaneously, the rate of change of energy storage output power is limited to prevent excessive current surges. In large disturbance mode, the system can provide frequency support of no less than 30% of rated power within 40ms, with a frequency recovery time not exceeding 400ms and a steady-state frequency deviation not exceeding 0.05Hz. Under high-enhancement control, even in the face of severe frequency drops, the system can maintain the DC bus voltage above the undervoltage protection threshold, ensuring that equipment does not lock out, while simultaneously helping the grid to quickly restore frequency.

[0053] In actual implementation, the control unit uses a DSP chip to implement the above logic. The DSP interrupts once every millisecond to complete frequency sampling, deviation and rate of change calculation, disturbance level judgment, matrix element update, virtual capacitor current calculation, and power command output. The matrix element update formula is fixed in the DSP program in C language to ensure fast and accurate calculation. The virtual capacitor current is directly superimposed on the AC side current command of the converter, requiring no additional hardware. The entire control logic is completed locally, without relying on communication with a host computer, adapting to extreme situations such as communication interruption.

[0054] Furthermore, to improve control reliability, the system incorporates a disturbance level switching delay and hysteresis mechanism. When the disturbance level changes, the control mode is not switched immediately. Instead, it continuously monitors for 20ms, and only switches after confirming the new level has stabilized, avoiding frequent jumps that could cause system oscillations. Simultaneously, hysteresis intervals are set between different levels. For example, when reducing from a medium disturbance to a small disturbance, |Δf| is required to fall below 0.06Hz before exiting the medium disturbance mode, preventing repeated switching near boundary values.

[0055] Through the aforementioned zoned control, the energy storage system can maintain optimal control performance under various disturbance intensities. It responds gently to small disturbances, reducing wear; provides precise support to medium disturbances, preventing blockage; and provides strong support to large disturbances, ensuring grid stability. This method of classifying disturbance levels based on frequency deviation and rate of change and adaptively adjusting the control strategy significantly improves the reliability, response speed, and lifespan of the energy storage system in primary frequency regulation, while maintaining its advantages of simple implementation and convenient upgrades, making it suitable for various energy storage application scenarios.

[0056] To prevent the power grid from instantly collapsing to low frequencies when a large power supply is suddenly lost, this invention incorporates "significantly reducing the rate of frequency change" into the control objective. Specifically, this is achieved without relying on external large capacitors or additional generators. Instead, the virtual inertia within the VSG is directly increased from the conventional 0.08 s to 0.22 s. Simultaneously, the damping coefficient is no longer fixed but fluctuates in real-time with the frequency deviation: the larger the deviation, the stronger the damping; the smaller the deviation, the automatically relaxed damping, thus avoiding over-adjustment and saving energy.

[0057] In the actual scenario, the dispatcher conducted two identical impact tests on the plant side, and the RoCoF comparison chart is shown in Figure 2.

[0058] Each instantaneous removal of a 1 MW load is equivalent to a hard pull on the power grid. The first time, using traditional fixed inertia, the peak frequency change rate reached 1.2 Hz / s, almost triggering the relay protection setting. The second time, by writing the parameters of this invention into the DSP, the same impact resulted in a peak frequency change rate of only 0.45 Hz / s, a reduction of 62%, remaining within the most stringent allowable range of GB / T33593-2017 throughout.

[0059] Throughout the process, the energy storage DC bus did not experience any additional fluctuations, and the module current did not show any new spikes. This proves that "increasing inertia and changing damping" is purely a software operation that has no negative impact on the equipment's lifespan. Instead, it provides the power grid with valuable "breathing time," creating a window for subsequent primary and secondary frequency regulation of the generating units. This truly achieves enhanced inertia support that "does more with less money."

[0060] Example 4: The biggest fear in primary frequency regulation of the power grid is that line tripping or large generator load shedding can cause a sudden collapse in frequency.

[0061] This embodiment simulates a "large disturbance" scenario: the frequency drop exceeds 0.3Hz, the rate of change is higher than 0.5Hz / s, and the dispatch command requires the 500kW energy storage power to be directly increased from standby to 520kW within 40ms, and the output to be sustained for no less than 10s. If a fixed virtual inertia is still used at this time, the DC bus will be instantly pulled down to below 0.81pu, the converter will immediately be undervoltage lockout, and a secondary frequency collapse will be almost inevitable.

[0062] Specifically, once the DSP detects |Δf|>0.3Hz and df / dt>0.5Hz / s within a 50ms cycle, it immediately sets the "large disturbance flag", and the entire subsequent cycle follows the "fuzzy logic-dominated" branch.

[0063] Fuzzy logic takes frequency deviation, rate of change, and SOC as inputs and outputs three algorithm weights: under large disturbances, the VSG weight is pushed to 0.65, the MPC weight is reduced to 0.2, and the PID weight is only 0.15, which ensures both millisecond-level power impact and sufficient steady-state accuracy.

[0064] At the same time, the scaling factor K of the matrix virtual inertia element was set to 1.4, and the virtual capacitor current factor was increased from 200μF to 280μF, instantly "supporting" the DC bus voltage drop.

[0065] The power command is issued after current sharing by the MGCC. The actual current rise rate of each converter is limited to 0.9 pu / ms, which satisfies the grid RoCoF < 0.5 Hz / s requirement and prevents the battery voltage from collapsing.

[0066] Throughout the process, the local controller continuously monitors the bus voltage. If the voltage does not rise back to 0.92 pu within 200 ms, the matrix mode is maintained. Once the voltage returns to the safe zone and the frequency rises back to above 49.9 Hz, the K value automatically drops back to 1.0, and the system smoothly exits the large disturbance mode.

[0067] The on-site oscilloscope recording showed that the lowest frequency was 49.52Hz, and the energy storage power reached 520kW after 38ms. The lowest bus voltage was 793V. Undervoltage lockout was not triggered throughout the process, and the grid frequency returned to 49.95Hz within 0.4s, meeting the most stringent grid operation guidelines.

[0068] In Example 5, the disturbance is the most common situation encountered in daily scheduling: the load suddenly increases by hundreds of megawatts, the frequency drops by about 0.15Hz, and the rate of change is 0.3Hz / s. It requires rapid power support, but also must avoid over-tuning and equipment fatigue.

[0069] In this embodiment, the photovoltaic output is cut off by 30% within 1 second on the RT-LAB platform, causing the frequency to drop to 49.85Hz, which is completely in line with the "medium disturbance" range.

[0070] After detecting 0.08Hz < |Δf| ≤ 0.3Hz and df / dt ≤ 0.5Hz / s in a 50ms clock cycle, the system enters the "MPC-driven, adaptive PID-assisted" path.

[0071] The MPC model uses energy storage power, bus voltage, and frequency deviation as state variables. It performs a 3-step prediction in the time domain and a 1-step control in the time domain, solving a quadratic programming problem online to obtain the optimal power trajectory for the next 150ms. Meanwhile, the adaptive PID exists in the outer loop to eliminate steady-state error.

[0072] In terms of weight allocation, MPC accounts for 0.6, PID accounts for 0.3, and VSG fast response accounts for only 0.1, which ensures optimal trajectory while retaining sufficient damping. The matrix virtual inertia K value is kept at 1.0, and the virtual capacitor current is maintained at 200μF without further amplification to prevent "over-braking" under moderate disturbances.

[0073] The MGCC distributes the general command according to the SOC level, and the four energy storage units follow the MPC trajectory respectively. The power rise rate is limited to 0.5 pu / ms, and the frequency recovery process is smooth without overshoot.

[0074] Records show that the frequency returned to 50Hz within 2.5s, the bus voltage was as low as 820V (1.025pu), the steady-state frequency deviation was 0.02Hz, the energy storage output had no oscillation, the battery current surge was reduced by 30%, the equipment temperature rise was reduced by 5℃, and the module life was significantly extended.

[0075] Example 6: Small disturbances are the most common and are also the "invisible killer" of equipment wear.

[0076] This embodiment simulates slight load fluctuations: the frequency oscillates slowly between 49.97Hz and 50.03Hz, with a maximum deviation of 0.03Hz and a rate of change of 0.1Hz / s, which falls entirely within the "small disturbance" range.

[0077] After determining that |Δf|≤0.08Hz and |df / dt|≤0.3Hz / s at a 50ms clock cycle, the system only enables the "adaptive PID-dominated" branch, and sets the weights of VSG and MPC to 0 to avoid large actions.

[0078] The adaptive PID uses the square of the frequency deviation as the loss function and updates the proportional, integral, and derivative parameters via online gradient descent, with the update step size limited to within 1% to prevent oscillations caused by the self-learning process itself. The matrix virtual inertia and virtual capacitor current are completely uninvolved, and the DC bus voltage is naturally adjusted by the original voltage loop, which is both energy-saving and quiet.

[0079] Due to the minimal disturbance, the MGCC current sharing coefficient remains almost constant. The four energy storage units are evenly distributed with a fine-tuning power of approximately ±5kW, resulting in an output current change of less than 0.02pu and a module temperature fluctuation of less than 1℃.

[0080] 24-hour continuous operation records show that the frequency qualification rate is 100%, the number of energy storage cycles is reduced by 15% compared with traditional fixed parameter PID, the steady-state frequency deviation is reduced to 0.01Hz, the SOC difference between battery packs is reduced to 2%, and the entire system achieves "silent frequency modulation" under small disturbances, which makes a positive contribution to equipment life and power quality.

Claims

1. An adaptive control method for primary frequency regulation of a power grid, characterized in that, Under the conditions that the remaining energy storage capacity is in the range of 15% to 25%, the energy storage output power is greater than 30% of its rated power, and the absolute value of the grid frequency deviation is greater than 0.08Hz, the fixed virtual inertia is replaced with a two-dimensional matrix that changes in real time with the remaining energy capacity and the DC bus voltage, and a virtual capacitor current is generated in the direction of preventing the DC bus voltage from dropping. The magnitude of the virtual capacitor current is proportional to the rate of change of the DC bus voltage.

2. The adaptive control method for primary frequency regulation of a power grid according to claim 1, characterized in that, The top-left element of the two-dimensional matrix decreases as the remaining charge decreases, while the bottom-right element increases as the remaining charge increases. The top-right and bottom-left elements are opposites of each other and are both proportional to the ratio of the DC bus voltage to the rated voltage.

3. The adaptive control method for primary frequency regulation of a power grid according to claim 1, characterized in that, Replacing the fixed virtual inertia with a two-dimensional matrix that changes in real time with the remaining power and DC bus voltage includes: classifying the disturbance into large, medium, and small disturbances according to the absolute value and rate of change of the grid frequency deviation; replacing the fixed virtual inertia with a two-dimensional matrix only when the disturbance level reaches medium or large disturbance; and keeping the original fixed virtual inertia unchanged for small disturbances.

4. The adaptive control method for primary frequency regulation of a power grid according to claim 3, characterized in that, The absolute value of the small disturbance deviation is no greater than 0.08 Hz and the absolute value of the rate of change is no greater than 0.3 Hz / s; the absolute value of the medium disturbance deviation is greater than 0.08 Hz and no greater than 0.3 Hz / s and the absolute value of the rate of change is no greater than 0.5 Hz / s; and the absolute value of the large disturbance deviation is greater than 0.3 Hz / s or the absolute value of the rate of change is greater than 0.5 Hz / s.

5. The adaptive control method for primary frequency regulation of a power grid according to claim 3, characterized in that, For large, medium, and small disturbances, corresponding fusion modes are used to obtain the power adjustment amount, including: large disturbances use a mode that combines fuzzy judgment and virtual synchronous machine characteristics, medium disturbances use a mode that combines prediction and adaptive correction characteristics, and small disturbances use only the adaptive correction mode.

6. The adaptive control method for primary frequency regulation of a power grid according to claim 1, characterized in that, The magnitude of the virtual capacitor current is equal to the product of 200μF and the rate of change of the DC bus voltage.

7. The adaptive control method for primary frequency regulation of a power grid according to claim 1, characterized in that, If the DC bus voltage is not lower than 92% of the rated value within 200ms, or if the remaining charge falls outside the 15%~25% range, the matrix control will automatically exit and the fixed virtual inertia will be restored.

8. The adaptive control method for primary frequency regulation of a power grid according to claim 1 or 7, characterized in that, While entering matrix control, the total response time of a single frequency modulation should not exceed 40ms, the frequency recovery time should not exceed 400ms, and the steady-state frequency deviation should not exceed 0.05Hz.

9. An adaptive control system for primary frequency regulation of a power grid, implementing the adaptive control method for primary frequency regulation of a power grid as described in any one of claims 1-8, characterized in that, include: The sampling module acquires DC bus voltage, output power, grid frequency deviation and its rate of change at a rate of not less than 1kHz; the judgment module identifies whether the following conditions are met simultaneously: remaining power of 15%~25%, output power greater than 30% of rated power, and absolute value of frequency deviation greater than 0.08Hz; when the judgment module gives a positive result, the control module replaces the fixed virtual inertia with a two-dimensional matrix and generates virtual capacitor current to jointly adjust the output power command.

10. An adaptive control system for primary frequency regulation of a power grid according to claim 9, characterized in that, The control module automatically restores the fixed virtual inertia after the voltage recovers or the remaining power drops out of the 15%~25% range.

Images