A flywheel energy storage system control method based on data center photovoltaic power supply

Abstract

The application discloses a flywheel energy storage system control method based on data center photovoltaic power supply, belongs to the technical field of power systems and new energy grid connection control, and comprises the following steps: acquiring output prediction data of photovoltaic power generation, real-time load power, city power grid voltage fluctuation parameters and real-time rotating speed state of the flywheel energy storage; taking the real-time rotating speed state of the flywheel energy storage as a dynamic constraint variable to generate a power gap prediction value; generating a flywheel city power collaborative control strategy according to the city power voltage fluctuation parameters; detecting the temperature change rate of the flywheel motor in real time to generate a control instruction; synchronously sending the control instruction to a photovoltaic inverter and the flywheel motor, adjusting photovoltaic inverter parameters and flywheel rotating speed response data, and dynamically updating a control parameter set. The application adopts a double regulation mechanism of flywheel rotating speed dynamic constraint and voltage fluctuation rate weight distribution, combines a temperature gradient harmonic collaborative suppression strategy, and can realize millisecond-level power compensation response, reduce grid dependence and prolong the service life of the energy storage system.

Description

Technical Field

[0001] The present invention relates to the field of power system and new energy grid-connected control technology, and in particular to a flywheel energy storage system control method based on photovoltaic power supply for a data center. Background Art

[0002] As a high-energy-consuming scenario, data centers rely on traditional mains electricity supply, which leads to problems such as high energy costs and large carbon emissions.

[0003] Existing technologies combine photovoltaic power generation with battery energy storage, but this technology suffers from significant response delays when dealing with intermittent photovoltaic power generation and sudden load changes, and batteries suffer from short cycle life and temperature sensitivity. Static weighting of utility power and energy storage cannot adapt to grid voltage fluctuations and high-frequency harmonic interference, leading to reduced power quality and even equipment damage. Furthermore, conventional prediction models fail to fully account for the physical constraints of flywheel energy storage, making it prone to overload and disconnection of the energy storage system when the speed approaches its limit. Summary of the Invention

[0004] To solve the above problems, the present invention provides a flywheel energy storage system control method based on photovoltaic power supply for data centers. It adopts a dual adjustment mechanism of dynamic constraint of flywheel speed and weight distribution of voltage fluctuation rate, combined with a temperature gradient harmonic collaborative suppression strategy, which can achieve millisecond-level power compensation response, reduce grid dependence and extend the life of the energy storage system.

[0005] The above objectives can be achieved through the following solutions:

[0006] A control method for a flywheel energy storage system based on photovoltaic power supply for a data center comprises obtaining output forecast data of photovoltaic power generation, real-time load power of the data center, voltage fluctuation parameters of the mains grid, and real-time speed status of the flywheel energy storage; using the real-time speed status of the flywheel energy storage as a dynamic constraint variable, correcting the time series matching relationship between photovoltaic output forecast data and load power, and generating a power gap prediction value with speed constraint; extracting a voltage fluctuation rate based on the mains voltage fluctuation parameter, generating a flywheel weight based on the magnitude of the voltage fluctuation rate, and generating a flywheel-mains coordinated control strategy in combination with the power gap prediction value with speed constraint; detecting the temperature change rate of the flywheel motor in real time, and gradient-constraining the power gap prediction value of the coordinated control strategy based on the temperature change rate to generate a control instruction; synchronously sending the control instruction to the photovoltaic inverter and the flywheel motor, adjusting the photovoltaic inverter parameters and the flywheel speed response data; and dynamically updating a control parameter set based on the adjusted photovoltaic inverter parameters and the flywheel speed response data.

[0007] Optionally, the acquisition of photovoltaic power generation output forecast data and the real-time load power of the data center includes: the output forecast data is preliminarily dynamically corrected through a cloud movement speed prediction model and photovoltaic component temperature monitoring values; the acquisition of real-time load power includes actively synchronizing the data center's server cluster scheduling plan and cooling system temperature control instructions.

[0008] Optionally, the real-time speed state of the flywheel energy storage is used as a dynamic constraint variable to correct the time series matching relationship between the photovoltaic output prediction data and the load power to generate a power gap prediction value with a speed constraint, including: using the real-time speed state of the flywheel energy storage as a dynamic constraint variable to construct a speed constraint coefficient and a trend constraint condition; aligning the preliminary dynamically corrected photovoltaic output prediction data with the load power in time series, and calculating a theoretical power gap value; and using the speed constraint coefficient, the trend constraint condition, and the theoretical power gap value to calculate a power gap prediction value.

[0009] Optionally, generating the flywheel weight according to the magnitude of the voltage fluctuation rate includes: when the voltage fluctuation rate is greater than or equal to a preset first threshold, increasing the flywheel weight to greater than or equal to the preset first weight threshold; when the voltage fluctuation rate is between the first threshold and a preset second threshold, adjusting the flywheel weight by linear interpolation according to the duration of the fluctuation; when the voltage fluctuation rate is less than or equal to the second threshold, increasing the AC power weight to a preset second weight threshold; wherein, the first weight threshold is less than or equal to the second weight threshold.

[0010] Optionally, the power gap prediction value combined with the speed constraint generates a flywheel-mains coordinated control strategy including: if the change rate of the power gap prediction value exceeds a preset change threshold, triggering a step adjustment mode of power distribution; if it is the first trigger, amplifying the flywheel weight according to preset conditions; if it is a continuous trigger, attenuating the flywheel weight according to an exponential curve.

[0011] Optionally, the gradient constraint on the power gap prediction value of the collaborative control strategy based on the temperature change rate includes: preprocessing the temperature change rate; compensating the voltage amplitude of the photovoltaic inverter according to the magnitude of the temperature change rate; adjusting the upper limit of the trend constraint condition according to the magnitude of the temperature change rate; and correcting the power gap prediction value according to the magnitude of the temperature change rate.

[0012] Optionally, the adjusting photovoltaic inverter parameters and flywheel speed response data comprises: monitoring each harmonic of the photovoltaic inverter output in real time, and extracting a specific harmonic amplitude; when the specific harmonic amplitude exceeds a safety amplitude threshold, reducing a harmonic current response coefficient of the flywheel motor; calculating a harmonic severity index according to the specific harmonic amplitude, and adjusting a power grid weight in the flywheel power grid cooperative control strategy according to the harmonic severity index.

[0013] Optionally, the dynamically updating the control parameter set comprises: when a deviation between an actual response speed of the flywheel and a predicted response speed exceeds a preset response range, switching to a standby residual error compensation algorithm to correct a torque of the flywheel motor; performing stability rating according to the power grid voltage fluctuation parameter, and dynamically adjusting an update period of the control parameter set.

[0014] Optionally, the dynamically updating the control parameter set further comprises: generating a reverse power feedback strategy for the power grid according to a relationship between a residual capacity of the current flywheel energy storage and the power output prediction data; and adjusting an upper limit of the flywheel weight according to a size relationship between the residual capacity of the current flywheel energy storage and a preset critical value.

[0015] Compared with the prior art, the present application has the following advantages:

[0016] 1. The present application dynamically corrects the matching relationship between photovoltaic output and load by real-time detection of flywheel speed as a dynamic constraint variable, so that the power gap prediction value is always limited by the physical state of the flywheel, avoiding the control failure problem caused by over-limiting of the capacity of the energy storage system in the traditional method; the cooperative control strategy generated in combination with the voltage fluctuation rate realizes millisecond-level power complementation between the power grid and the flywheel, and significantly reduces the risk of voltage sag;

[0017] 2. The present application gradient-constrains the power gap by the temperature change rate, actively reduces the output demand before the flywheel motor overheats, and solves the problem of accelerated mechanical wear caused by sudden temperature rise in the traditional control; at the same time, the dynamic harmonic suppression mechanism reduces the loss of power electronic devices under high-frequency modulation, prolonging the service life of the photovoltaic inverter and the flywheel drive module;

[0018] 3. The present application dynamically corrects the control parameter set by using the residual error compensation algorithm, and adjusts the flywheel response deviation and the power grid stability rating, breaking through the adaptability limitations of traditional fixed parameter control to complex working conditions; through the dynamic association of the reverse power feedback strategy and the upper limit of the weight, the intelligent balance of the energy supply and demand relationship is realized.

[0019] Other features and advantages of the present application will be set forth in the specification, and in part will become apparent from the specification, or can be learned by practice of the present application. The objectives and other advantages of the present application can be realized and attained by the structure particularly pointed out in the specification, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 This is a flow chart of a method for controlling a flywheel energy storage system based on photovoltaic power supply for a data center according to an embodiment of the present invention.

[0022] Figure 2 2 is a photovoltaic sag curve diagram according to an embodiment of the present invention.

[0023] Figure 3 1 is a load mutation curve diagram of an embodiment of the present invention.

[0024] Figure 4 4 is a harmonic suppression curve diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0026] Reference Figure 1 One embodiment of the present invention proposes a control method for a flywheel energy storage system based on photovoltaic power supply for a data center. This method adopts a dual regulation mechanism of dynamic flywheel speed constraint and voltage fluctuation weight distribution, combined with a temperature gradient harmonic collaborative suppression strategy. This method can achieve millisecond-level power compensation response, reduce grid dependence, and extend the life of the energy storage system.

[0027] The method of this embodiment specifically includes:

[0028] Obtain output forecast data for photovoltaic power generation, real-time load power of data centers, voltage fluctuation parameters of the mains grid, and real-time speed status of flywheel energy storage;

[0029] Specifically, the photovoltaic power generation prediction system collects output forecast data for future time periods, monitors the total power consumption of the data center server cluster and cooling system in real time, and simultaneously measures the voltage fluctuation range of the mains power grid and the speed value of the flywheel energy storage device. The output forecast data must be combined with the light intensity and temperature parameters in the meteorological data. The real-time load power includes the energy consumption of the server computing load and the refrigeration equipment. The mains voltage fluctuation parameters are sampled at the millisecond level by the power monitoring module and the fluctuation rate is calculated. The flywheel speed status is collected through the speed sensor. Among them, the output forecast data is the power generation capacity prediction value based on meteorological conditions and photovoltaic module performance, the real-time load power is the current total power demand of the data center, the mains voltage fluctuation parameter is the variation range of the grid voltage within the set time window, and the real-time speed status of the flywheel energy storage is the instantaneous angular velocity of the energy storage flywheel rotor.

[0030] Taking the real-time speed state of the flywheel energy storage as the dynamic constraint variable, the time series matching relationship between the photovoltaic output forecast data and the load power is corrected to generate a power gap forecast value with speed constraint.

[0031] The voltage fluctuation rate is obtained according to the mains voltage fluctuation parameter, and a flywheel weight is generated according to the magnitude of the voltage fluctuation rate. The flywheel-mains coordinated control strategy is generated in combination with the power gap prediction value constrained by the speed.

[0032] Specifically, in the flywheel-mains coordinated control strategy, the power gap prediction value ,have:

[0033] ;

[0034] ;

[0035] ;

[0036] Where, is the flywheel weight, is the mains power weight, is the flywheel power, is the mains power.

[0037] detecting the temperature change rate of the flywheel motor in real time, and performing gradient constraint on the power gap prediction value of the collaborative control strategy according to the temperature change rate to generate a control instruction;

[0038] Synchronously sending the control instructions to the photovoltaic inverter and the flywheel motor to adjust the photovoltaic inverter parameters and the flywheel speed response data;

[0039] Specifically, when insufficient flywheel speed is detected, an instruction to increase the power factor is sent to the photovoltaic inverter, increasing its output by 3%-5% to compensate for the power gap; at the same time, a torque instruction is sent to the flywheel motor to control its speed increase rate within a threshold of 500 rpm.

[0040] The control parameter set is dynamically updated based on the adjusted PV inverter parameters and flywheel speed response data.

[0041] Specifically, the control system's core parameter library is updated in real time based on the output characteristics of the modified PV inverter parameters and the dynamic response of the flywheel speed. In response to flywheel response speed deviations, a residual compensation algorithm is switched to correct the flywheel torque. The parameter update frequency is adaptively adjusted based on the utility grid stability level. Furthermore, the backfeed strategy and weight upper limit constraints are dynamically generated based on the matching degree between the flywheel's remaining capacity and the predicted PV output.

[0042] This method constructs a collaborative control system for photovoltaic flywheels and power grids through the dynamic coupling of multi-source data. Its core is to use flywheel speed as an immediate constraint and adaptively adjust the power allocation strategy based on grid conditions. This reduces reliance on utility power and leverages the flywheel's rapid response to smooth photovoltaic fluctuations. A temperature gradient constraint mechanism is introduced to prevent equipment overload risks. Dynamic harmonic suppression improves power quality and ensures stable operation of highly sensitive loads in data centers. For example, in the combined scenario of a sudden drop in photovoltaic output and grid instability, the flywheel's output priority can be automatically adjusted to restore power balance within a short period of time.

[0043] Optionally, the obtaining of photovoltaic power generation output forecast data and real-time load power of the data center includes:

[0044] The output forecast data is initially dynamically corrected using a cloud movement speed prediction model and photovoltaic module temperature monitoring values;

[0045] Specifically, cloud trajectory data from meteorological satellites is used as input into a cloud movement speed prediction model to calculate the future time window of photovoltaic panels being blocked. At the same time, combined with the measured value of the photovoltaic module backplane temperature sensor, based on the negative correlation between photovoltaic efficiency and temperature, the original predicted output is reduced and corrected to obtain the output prediction data. The correction formula is as follows:

[0046] ;

[0047] Where, Contribute to the original forecast, is the photovoltaic temperature attenuation coefficient (usually 0.004~0.005 / ℃), is the difference between the current temperature and the standard test temperature, is the cloud occlusion attenuation factor, The cloud movement speed prediction model is an algorithm that analyzes cloud motion vectors based on satellite cloud images, and the PV module temperature monitoring value is the module surface temperature measured by infrared sensors or thermocouples.

[0048] The acquisition of the real-time load power includes actively synchronizing the server cluster scheduling plan and the refrigeration system temperature control instructions of the data center.

[0049] Specifically, the data center management system sends the server cluster's batch task startup schedule and CPU utilization plan to the energy storage control unit, simultaneously obtains the compressor frequency instructions generated by the refrigeration system based on room temperature feedback, and comprehensively calculates the future minute-level load change trend.

[0050] Among them, the server cluster scheduling plan is the computing task batch execution schedule pre-set by the data center, and the cooling system temperature control instruction is the equipment power control signal generated based on the thermodynamic model.

[0051] Optionally, the method of using the real-time speed state of the flywheel energy storage as a dynamic constraint variable, correcting the time series matching relationship between the photovoltaic output prediction data and the load power, and generating a power gap prediction value with a speed constraint includes:

[0052] Taking the real-time speed state of the flywheel energy storage as the dynamic constraint variable, the speed constraint coefficient and trend constraint condition are constructed;

[0053] Specifically, the speed constraint coefficient is a real-time constraint, and the trend constraint condition is a trend constraint. A dual speed constraint model is constructed using the real-time constraint and the trend constraint. ,have:

[0054] ;

[0055] Where, is the current flywheel speed, To allow the maximum speed, is the threshold coefficient; the trend constraint condition uses ARIMA time series modeling of historical speed data to predict the speed change rate within a preset future time period. The speed constraint coefficient is the core parameter for dynamically adjusting the power allocation ratio based on the remaining flywheel energy storage. The trend constraint condition is a rule limiting the speed change rate predicted based on historical data. The theoretical power gap value is the original difference between the load power and the predicted PV output value.

[0056] Align the initial dynamically corrected photovoltaic output forecast data with the load power in time series, and calculate the theoretical power gap value;

[0057] Specifically, the preliminary dynamic corrected photovoltaic power prediction data and the load power are aligned by minute-level timestamps; the theoretical power gap value is calculated for each time point , wherein

[0058] ;

[0059] , wherein is the load power. Time series alignment is a preprocessing method for integrating multi-source data based on a unified time reference, and the theoretical power gap value is the calculated value of the power difference without considering the physical limitations of the flywheel.

[0060] The power gap prediction value is calculated using the rotational speed constraint coefficient, the trend constraint condition, and the theoretical power gap value.

[0061] Specifically, considering the instantaneous output capability of the flywheel, the power gap prediction value is calculated using the rotational speed constraint coefficient, the trend constraint condition, and the theoretical power gap value. For the power gap prediction value , wherein

[0062] ;

[0063] , wherein is a normalization function used to suppress the risk of overshoot, is a trend constraint condition.

[0064] Illustratively, the photovoltaic power prediction of a certain data center suddenly drops, from 400kW to 220kW, and the load power increases to 350kW due to an emergency task. The flywheel rotational speed =14500rpm, =16000rpm, =0.85, the ARIMA model predicts that the future 3 minutes =200rpm / min, in an upward trend; the rotational speed constraint coefficient is calculated to be , and then the theoretical power gap value is calculated , j is calculated to be about 1, so the power gap prediction value . Through the double restriction of real-time rotational speed constraint and trend prediction, the flywheel output only needs to bear the gap of 121kW instead of the theoretical 130kW, actively reducing the load pressure of the flywheel in the working condition close to the upper limit, and slowing down the mechanical wear. After the trend prediction module detects the continuous upward trend of the rotational speed, it triggers the sigmoid function to suppress excessive power allocation, avoiding the risk of flywheel overspeed off-grid, and ensuring the safety response of the energy storage system. This method dynamically integrates the instantaneous physical state (rotational speed) and future trend of the flywheel, restricts the matching process of photovoltaic power prediction and load demand, and ensures that the generated power gap value is always within the safe operating boundary of the flywheel system.

[0065] Optionally, the generating the flywheel weight according to the size of the voltage fluctuation rate comprises:

[0066] When the voltage fluctuation rate is greater than or equal to a preset first threshold, the flywheel weight is increased to be greater than or equal to a preset first weight threshold.

[0067] Specifically, when the voltage fluctuation rate is greater than or equal to the first threshold, the flywheel weight is increased to a value not less than the first weight threshold, and for the new flywheel weight , there is:

[0068] ;

[0069] In the formula, is the first weight threshold, the first weight threshold is a larger value, and the value range can be set to be between 0.7-0.9, is an amplification coefficient, used for amplifying the current flywheel weight, and the value range can be set to be between 1.1-1.5, is the current flywheel weight, wherein the first weight threshold and the amplification coefficient are set as constraint conditions, that is, the product of the first weight threshold and the amplification coefficient is less than or equal to 0.9. is the voltage fluctuation rate, is the first threshold.

[0070] When the voltage fluctuation rate is between the first threshold and a preset second threshold, the flywheel weight is adjusted according to the fluctuation duration by linear interpolation;

[0071] Specifically, when the voltage fluctuation rate is detected to be between the first threshold and the second threshold, the timing is started, the fluctuation duration is calculated, and the flywheel weight is adjusted by using the fluctuation duration, and at this time, for the new flywheel weight , there is:

[0072] ;

[0073] In the formula, is a constant coefficient of the duration, is the second threshold. The linear interpolation adjustment strategy is a composite proportional adjustment method based on the fluctuation degree and the time.

[0074] When the voltage fluctuation rate is less than or equal to the second threshold, the power supply weight is increased to a preset second weight threshold.

[0075] The first weight threshold is less than or equal to the second weight threshold.

[0076] Specifically, when the voltage fluctuation rate is less than or equal to the second threshold, the mains power takes the lead in the weight value, and the flywheel weight is greatly reduced. ,have:

[0077] ;

[0078] Where, is the second weight threshold, Here, the second weight threshold is greater than the first weight threshold, so that the mains power takes a dominant position.

[0079] Optionally, generating a flywheel-mains coordinated control strategy by combining the power gap prediction value with the speed constraint includes:

[0080] If the change rate of the power gap prediction value exceeds a preset change threshold, a step adjustment mode of power allocation is triggered;

[0081] If it is the first trigger, the flywheel weight will be amplified according to the preset conditions;

[0082] Specifically, if it is the first trigger, the flywheel needs to bear a larger proportion of the power gap prediction value, and the influence of the flywheel temperature needs to be considered. ,have:

[0083] ;

[0084] Where, To obtain the minimum function, To obtain the maximum value function, is the adjustment coefficient of temperature constraint, is the safe temperature threshold of the flywheel, is the current temperature of the flywheel.

[0085] If it is continuously triggered, the flywheel weight will be attenuated according to the exponential curve.

[0086] Specifically, the flywheel weight is adjusted using an exponential decay allocation strategy. ,have:

[0087] ;

[0088] Where, The duration of the trigger.

[0089] Optionally, performing gradient constraint on the power gap prediction value of the collaborative control strategy according to the temperature change rate includes:

[0090] Preprocessing the temperature change rate;

[0091] Specifically, a sliding window check was used to calculate the temperature change rate five times in a row to eliminate abnormal mutations, such as instantaneous ±50°C / s exceeding the limit, and the smoothed temperature change rate data was retained.

[0092] Compensating the voltage amplitude of the photovoltaic inverter according to the magnitude of the temperature change rate;

[0093] Specifically, the temperature change rate is determined. When the temperature change rate is greater than a preset first temperature threshold, the voltage amplitude of the photovoltaic inverter is compensated. ,have:

[0094] ;

[0095] Where, is the compensation coefficient, is the temperature change rate. Through voltage amplitude compensation, the output forecast data of photovoltaic power generation can be adjusted.

[0096] adjusting the upper limit of the trend constraint condition according to the magnitude of the temperature change rate;

[0097] Specifically, the upper limit of the trend constraint condition is adjusted according to the temperature change rate and the preset second temperature threshold, wherein the second temperature threshold can be equal to the first temperature threshold. ,have:

[0098] ;

[0099] Where, is the speed upper limit adjustment coefficient, By limiting the trend constraint condition, the power gap prediction value is limited.

[0100] The power gap prediction value is corrected according to the magnitude of the temperature change rate.

[0101] Specifically, the duration of the temperature change rate being greater than the first temperature threshold or the second temperature threshold is collected, and the duration is not calculated repeatedly. When the temperature change rate is greater than the first temperature threshold and the second temperature threshold at the same time, the duration is calculated for the longest duration. When the duration is greater than the preset load shedding time threshold, the current power gap prediction value is corrected, and the corrected power gap prediction value is ,have:

[0102] .

[0103] Optionally, adjusting photovoltaic inverter parameters and flywheel speed response data includes:

[0104] Real-time monitoring of the harmonics output by the photovoltaic inverter and extraction of the specific harmonic amplitude;

[0105] Specifically, the voltage and current waveforms at the inverter output are collected in real time, and the 3rd / 5th / 7th harmonic components are extracted through an improved fast Fourier transform (FFT) algorithm to obtain the specific harmonic amplitude.

[0106] When the amplitude of the specific subharmonic exceeds a safe amplitude threshold, reducing the harmonic current response coefficient of the flywheel motor;

[0107] Specifically, the current loop PI controller parameters are dynamically adjusted. For example, when the 5th harmonic amplitude is detected to exceed 110% of the IEC61000-3-2 standard limit, the proportional gain coefficient in the PI controller is adjusted. ,have:

[0108] ;

[0109] Where, 为当前比例增益系数, For the 次谐波幅值, For the 次谐波电流限值, is the gain attenuation coefficient. The proportional gain coefficient is negatively correlated with the flywheel response delay. When high-frequency harmonics are large, the proportional gain coefficient decreases, suppressing the rapid response of the control loop and avoiding the risk of oscillation caused by overmodulation of the flywheel motor. When the amplitude of a specific subharmonic is less than the safe amplitude threshold, the proportional gain coefficient returns to a higher value to maintain the dynamic tracking capability of the control system.

[0110] A harmonic severity index is calculated according to the specific subharmonic amplitude, and a mains weight in the flywheel-mains coordinated control strategy is adjusted according to the harmonic severity index.

[0111] 具体的,对于谐波严重指数 ,have:

[0112] ;

[0113] Where, For the 次谐波幅值, For the Subharmonic current limit, using the harmonic severity index to correct the mains weight, for the corrected mains weight ,have:

[0114] ;

[0115] Where, 为常系数, 为当前的市电权重。

[0116] Optionally, the dynamic update control parameter set includes:

[0117] When the deviation between the actual response speed of the flywheel and the predicted response speed exceeds the preset response range, the system switches to the backup residual compensation algorithm to correct the torque of the flywheel motor;

[0118] Specifically, the absolute value of the deviation of the speed change rate is calculated by collecting the flywheel speed change rate in real time and the expected speed change rate calculated by the preset flywheel dynamics model; it is judged whether the absolute value of the deviation is greater than the preset response range. If so, it is determined that the response is abnormal and the backup residual compensation algorithm is switched to; the residual compensation algorithm records the historical data of the actual speed change rate of recent cycles, calculates the mean of the absolute value of the deviation, and obtains the residual mean; the residual mean is used to correct the torque instruction, and the corrected torque instruction ,have:

[0119] ;

[0120] Where, 为当前的转矩指令, 为修正增益, The first 10ms after triggering adopts the linear superposition of the current torque command and the corrected torque command. ,have:

[0121] ;

[0122] Where, is the smoothing coefficient, which increases from 0 to 1 with a step size of 0.1 / ms to achieve smooth transition.

[0123] A stability rating is performed according to the voltage fluctuation parameters of the mains power grid, and an update period of the control parameter set is dynamically adjusted.

[0124] Specifically, the voltage fluctuation rate, frequency deviation and harmonic distortion rate are extracted according to the voltage fluctuation parameters of the mains power grid, and the stability evaluation value is calculated. ,have:

[0125] ;

[0126] Where, 为电压波动率的归一化数值, 为频率偏移度的归一化数值, 为谐波畸变率, 、 、 The weight coefficient is used to classify the stability rating into levels, such as high stability, medium stability, and low stability. The stability evaluation value range is divided into intervals based on the level, for example, 0-0.4 for high stability, 0.4-0.7 for medium stability, and 0.7-1 for low stability. An update cycle is set for each interval. The update cycle is updated based on the stability evaluation value range.

[0127] Optionally, the dynamic update control parameter set further includes:

[0128] generating a reverse feeding strategy for the utility grid based on a relationship between the current remaining capacity of the flywheel energy storage and the output forecast data;

[0129] Specifically, define the relationship threshold between photovoltaic predicted output and flywheel remaining capacity :

[0130] ;

[0131] Where, 为当前飞轮储能的剩余容量, 为飞轮储能的最大容量, The rated value of the output prediction data. When the value is greater than the preset standard value, it is allowed to feed power to the grid, and the priority is set according to the electricity price time-sharing strategy. When x is less than or equal to the preset standard value, power feeding is prohibited and local loads are prioritized.

[0132] The upper limit of the flywheel weight is adjusted according to the relationship between the current remaining capacity of the flywheel energy storage and the preset critical value.

[0133] Specifically, the critical value can be set to 20% of the maximum capacity of the flywheel energy storage, so the upper limit of the flywheel weight is ,have:

[0134] ;

[0135] Where, 为预设的临界值。

[0136] In order to verify the feasibility of the present invention in implementation, the present invention was applied to a large cloud computing data center. Located in the northwest region with abundant sunlight resources, the data center has deployed a megawatt-level distributed photovoltaic power generation system and is equipped with a flywheel energy storage array to smooth out photovoltaic fluctuations and load mutations. The data center load includes a server cluster, a liquid cooling system, and a backup power supply. The total power demand peak is 800kW, and power supply continuity and voltage stability must be guaranteed. Since photovoltaic output fluctuates frequently due to cloud cover, the traditional mains power plus energy storage model cannot achieve millisecond-level power compensation, resulting in a significant risk of voltage sag at critical loads.

[0137] In this embodiment, a data center employs the photovoltaic-flywheel coordinated control method proposed in this invention, collecting real-time photovoltaic output forecast data (based on a cloud trajectory model and component temperature correction), server load scheduling plans, flywheel speed, and utility voltage fluctuation parameters. The system uses a dynamic speed constraint model to correct power shortfall predictions, incorporates voltage fluctuations to generate a flywheel weight allocation strategy, and incorporates temperature gradient constraints and harmonic suppression mechanisms to optimize control instructions.

[0138] To verify its effectiveness, a test period from March to September 2024 was selected to compare the performance of traditional energy storage control (lithium iron phosphate batteries with static weight distribution) with the proposed method. Both the experimental and control groups were connected to the same PV array and load, with data collected at a millisecond frequency. The analysis scenarios included three typical events: PV sag, sudden load changes, and utility voltage fluctuations.

[0139] Test scenario 1: Short-term sag in PV output

[0140] At 11:20 on July 15, Figure 2 As shown, cloud cover caused the PV output to drop from 650kW to 320kW within 30 seconds, with the load stabilizing at 720kW. The system of the present invention monitored the speed status in real time (currently 12,500 rpm, maximum 16,000 rpm), calculated the speed constraint coefficient as 0.88, and generated a predicted power shortfall of 148kW (theoretical value 170kW). Taking into account the mains voltage fluctuation (3.2%, exceeding the threshold of 2.5%), the flywheel weight was increased to 0.75, resulting in an instantaneous flywheel output of 111kW, supplemented by 37kW from the mains, with a response time of 1.2 seconds. Due to a battery response delay (4.5 seconds), the control group experienced a 280ms voltage sag (to 198V), while the experimental group maintained a voltage fluctuation range of 205V±2V.

[0141] Test scenario 2: Server cluster load burst

[0142] At 14:05 on August 3, Figure 3 As shown, the launch of a batch computing task caused the load to jump from 600 kW to 780 kW, while the PV output remained at 480 kW. The system predicted the load change by synchronizing with the server's scheduling plan and triggered a step-by-step adjustment mode: upon initial triggering, the flywheel weight was amplified to 0.85, then exponentially decayed to 0.6 within 10 seconds. The flywheel speed increased from 14,000 rpm to 15,200 rpm, compensating for the shortfall by 238 kW. The mains then supplied the remaining 42 kW after a 5-second delay. The experimental group experienced voltage recovery in 0.8 seconds. In the control group, due to the static weight distribution (flywheel:mains = 0.5:0.5), compensation was delayed, resulting in voltage recovery in 3.6 seconds.

[0143] Test scenario 3: Mains high harmonic interference

[0144] September 12, 10:45, Figure 4 As shown, when the fifth harmonic (amplitude exceeding the limit by 15%) was introduced to the grid, the system detected a real-time harmonic severity index (H) of 1.32, reducing the flywheel harmonic response coefficient (PI gain attenuation by 32%) while simultaneously increasing the utility power weight to 0.65. The experimental group's output voltage THD dropped from 7.8% to 3.1%. However, due to the lack of dynamic harmonic suppression strategy adjustment in the control group, THD remained as high as 6.9%, triggering protective shutdowns on some sensitive servers.

[0145] Experiments have shown that this method, through dynamic speed constraints and coordinated weight distribution between flywheel and mains, improves the power gap compensation response speed to 1.2 seconds, a 275% improvement over traditional methods. The temperature gradient constraint mechanism ensures that the flywheel operates within a safe speed range, reducing annual overspeed events by 92%, and adaptive harmonic suppression reduces THD by 56%. In terms of economic efficiency, the annual comprehensive cost savings reach 378,000 yuan, validating the significant advantages of this method in improving power supply quality, extending equipment life, and reducing operating costs.

[0146] It should be noted that the electrical connections between the above-mentioned units do not necessarily mean direct connections of lines. Indirect connections are applicable to the embodiments of the present invention as long as the purpose of the present invention is achieved. The above description is only an exemplary embodiment of the present invention and is not intended to limit the scope of the present invention.

[0147] That is, any equivalent changes and modifications made according to the teachings of the present invention are still within the scope of the present invention. Those skilled in the art will readily conceive of other embodiments of the present invention after considering the disclosure of the specification and practical truths. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary technical means in the art not described herein.

Claims

1. A flywheel energy storage system control method based on photovoltaic power supply in a data center, characterized in that: The method comprises: Obtaining photovoltaic power generation output forecast data, real-time load power of the data center, mains grid voltage fluctuation parameters, and real-time speed status of the flywheel energy storage; wherein obtaining photovoltaic power generation output forecast data and real-time load power of the data center includes: performing preliminary dynamic correction on the output forecast data using a cloud movement speed prediction model and photovoltaic module temperature monitoring values; obtaining real-time load power includes actively synchronizing the data center's server cluster scheduling plan and cooling system temperature control instructions; The real-time speed state of the flywheel energy storage is used as a dynamic constraint variable, and the time series matching relationship between the photovoltaic output forecast data and the load power is corrected to generate a power gap prediction value with a speed constraint. This includes: using the real-time speed state of the flywheel energy storage as a dynamic constraint variable, constructing a speed constraint coefficient and a trend constraint condition; aligning the preliminary dynamically corrected photovoltaic output forecast data with the load power in time series and calculating a theoretical power gap value; and calculating a power gap prediction value using the speed constraint coefficient, the trend constraint condition, and the theoretical power gap value. The voltage fluctuation rate is obtained according to the mains voltage fluctuation parameter, and a flywheel weight is generated according to the magnitude of the voltage fluctuation rate. The flywheel-mains coordinated control strategy is generated in combination with the power gap prediction value constrained by the speed. detecting the temperature change rate of the flywheel motor in real time, and performing gradient constraint on the power gap prediction value of the collaborative control strategy according to the temperature change rate to generate a control instruction; Synchronously sending the control instructions to the photovoltaic inverter and the flywheel motor to adjust the photovoltaic inverter parameters and the flywheel speed response data; The control parameter set is dynamically updated based on the adjusted PV inverter parameters and flywheel speed response data.

2. A flywheel energy storage system control method based on photovoltaic power supply for a data center according to claim 1, characterized in that: Generating a flywheel weight according to the magnitude of the voltage fluctuation rate includes: When the voltage fluctuation rate is greater than or equal to a preset first threshold, the flywheel weight is increased to be greater than or equal to the preset first weight threshold; When the voltage fluctuation rate is between the first threshold and a preset second threshold, adjusting the flywheel weight by linear interpolation according to the duration of the fluctuation; When the voltage fluctuation rate is less than or equal to the second threshold, the mains power weight is increased to a preset second weight threshold; The first weight threshold is less than or equal to the second weight threshold.

3. A flywheel energy storage system control method based on photovoltaic power supply for data centers according to claim 1, characterized in that: The generating of the flywheel-mains coordinated control strategy by combining the power gap prediction value with the speed constraint includes: If the change rate of the power gap prediction value exceeds a preset change threshold, a step adjustment mode of power allocation is triggered; If it is the first trigger, the flywheel weight will be amplified according to the preset conditions; If it is continuously triggered, the flywheel weight will be attenuated according to the exponential curve.

4. A flywheel energy storage system control method based on photovoltaic power supply for a data center according to claim 1, characterized in that: The step of performing gradient constraint on the power gap prediction value of the collaborative control strategy according to the temperature change rate includes: Preprocessing the temperature change rate; Compensating the voltage amplitude of the photovoltaic inverter according to the magnitude of the temperature change rate; adjusting the upper limit of the trend constraint condition according to the magnitude of the temperature change rate; The power gap prediction value is corrected according to the magnitude of the temperature change rate.

5. The flywheel energy storage system control method based on photovoltaic power supply for data center according to claim 1, characterized in that: The adjustment of photovoltaic inverter parameters and flywheel speed response data includes: Real-time monitoring of the harmonics output by the photovoltaic inverter and extraction of the specific harmonic amplitude; When the amplitude of the specific subharmonic exceeds a safe amplitude threshold, reducing the harmonic current response coefficient of the flywheel motor; A harmonic severity index is calculated according to the specific subharmonic amplitude, and a mains weight in the flywheel-mains coordinated control strategy is adjusted according to the harmonic severity index.

6. A flywheel energy storage system control method based on photovoltaic power supply for a data center according to claim 1, characterized in that: The dynamic update control parameter set includes: When the deviation between the actual response speed of the flywheel and the predicted response speed exceeds the preset response range, the system switches to the backup residual compensation algorithm to correct the torque of the flywheel motor; A stability rating is performed according to the voltage fluctuation parameters of the mains power grid, and an update period of the control parameter set is dynamically adjusted.

7. A flywheel energy storage system control method based on photovoltaic power supply for a data center according to claim 1, characterized in that: The dynamic update control parameter set also includes: generating a reverse feeding strategy for the utility grid based on a relationship between the current remaining capacity of the flywheel energy storage and the output forecast data; The upper limit of the flywheel weight is adjusted according to the relationship between the current remaining capacity of the flywheel energy storage and the preset critical value.

Images