Flywheel energy storage system control system for data center computing power load energy recovery

Abstract

The invention relates to the technical field of data center energy management, and discloses a flywheel energy storage system control system for data center computing power load energy recovery, which comprises an energy storage unit, a load monitoring unit, a flywheel control unit and an energy distribution unit. The energy storage unit is provided with a plurality of groups of flywheel energy storage devices capable of charging / discharging; the load monitoring unit monitors load data by covering key nodes of computing power equipment through a power sensor; the flywheel control unit collects flywheel rotating speed and power grid frequency fluctuation data; and the energy distribution unit decides charging and discharging based on load data, establishes a load and frequency coordinated alternating control strategy, and dynamically adjusts and predicts the running state of the flywheel. The system adopts time sequence synchronous control, the energy distribution unit comprises multiple modules, the alternate controller supports dual-mode switching and parameter optimization, and the flywheel unit adopts an annular redundancy or star-shaped concentrated structure. The system realizes efficient recovery of computing power load energy and flywheel energy storage optimization scheduling, and improves the energy utilization rate of a data center and the stability of a power grid.

Description

Technical Field

[0001] The present invention relates to the field of data center energy management technology, and specifically to a flywheel energy storage system control system for recovering computing load energy in a data center. Background Art

[0002] With the rapid development of information technology, data centers, as the infrastructure of the digital economy, have seen explosive growth in both scale and computing power. During data center operation, the load on computing equipment exhibits significant volatility and intermittent characteristics. This unstable power demand not only places increased pressure on the power grid but also results in significant energy waste. Traditional data center energy management systems struggle to effectively respond to dynamic changes in computing load, failing to achieve real-time energy recovery and rational allocation. This results in low energy utilization, increased data center operating costs, and increased environmental impact.

[0003] Among existing energy storage technologies, battery energy storage systems have problems such as short lifespan, high maintenance costs, and poor environmental adaptability. In addition, they are prone to performance degradation and safety hazards during frequent charging and discharging. Flywheel energy storage, as a mechanical energy storage method, has the advantages of high power density, fast charging and discharging speed, long cycle life, and environmental friendliness. It has unique advantages in dealing with short-term power fluctuations and energy recovery. However, the current flywheel energy storage system control system for the computing power load characteristics of data centers is not yet mature, and lacks an effective load monitoring, flywheel control and energy distribution coordination mechanism, making it impossible to achieve efficient recovery of computing power load energy and optimized scheduling of flywheel energy storage devices.

[0004] Specifically, the existing technology has the following deficiencies: First, in terms of load monitoring, traditional monitoring methods are unable to obtain the load data of each key node of the computing power equipment in a comprehensive and real-time manner, making it difficult to accurately reflect the true state of the computing power load, resulting in a lack of reliable basis for the charging and discharging decisions of the energy storage system. Secondly, in terms of flywheel control, there is a lack of coordinated monitoring and control of the flywheel speed and grid frequency, and the flywheel operating state cannot be adjusted in time according to grid frequency fluctuations, affecting the stability of the system and the quality of power. Furthermore, in terms of energy distribution, the existing system has failed to establish a control strategy for the coordination of load and frequency, and cannot achieve dynamic optimization scheduling of flywheel energy storage devices, resulting in low energy recovery efficiency and an inability to meet the data center's requirements for high reliability and high energy utilization.

[0005] Existing flywheel energy storage systems lack effective timing synchronization control mechanisms when multiple flywheels operate in concert, making it difficult to synchronize the operating states of the individual flywheel units, resulting in a decrease in overall system performance. Furthermore, the flywheel unit topology design lacks optimized configurations tailored to the specific characteristics of data center power networks, preventing the full utilization of flywheel energy storage devices.

[0006] Therefore, there is an urgent need for a flywheel energy storage system control system that can adapt to the computing load characteristics of data centers. Through efficient load monitoring, precise flywheel control and intelligent energy distribution, it can realize real-time recovery of computing load energy and optimized scheduling of flywheel energy storage devices, thereby improving the energy utilization rate of data centers and reducing operating costs and environmental impacts. Summary of the Invention

[0007] The purpose of the present invention is to provide a flywheel energy storage system control system for data center computing load energy recovery to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a flywheel energy storage system control system for data center computing load energy recovery, the system comprising: Energy storage unit, load monitoring unit, flywheel control unit and energy distribution unit; An energy storage unit includes a plurality of flywheel energy storage devices disposed in a data center power network; each flywheel energy storage device is in a charging or discharging state; The load monitoring unit includes multiple power sensors deployed on the power supply lines of the data center. Each power sensor monitors the power input nodes of the computing power equipment, so that at least one power sensor captures real-time load data of a group of computing power equipment. A flywheel control unit includes a flywheel speed sensor and a grid frequency detector, which respectively measure the speed data of the flywheel energy storage device and the grid frequency fluctuation data; The energy distribution unit uses the computing equipment load data collected by the load monitoring unit to make flywheel charging and discharging decisions; and establishes an alternating control strategy that coordinates load and frequency, and uses grid frequency fluctuation data to dynamically adjust charging and discharging decisions; based on the adjustment results, it predicts the operating status of the flywheel energy storage device in the next time period for rapid response to the next cycle of power dispatch.

[0009] Preferably, the energy storage unit, the load monitoring unit and the flywheel control unit are controlled by sequential synchronization; specifically including: The flywheel energy storage devices in the energy storage unit are numbered, and one of the flywheels is used as the master flywheel to generate the synchronization signal Fly SYNC; the remaining flywheels switch modes synchronously after receiving the synchronization signal Fly SYNC, and send the operating status data with the flywheel number to the energy distribution unit; The multiple power sensors of the load monitoring unit are numbered; the synchronization signal Fly SYNC adjusts the sampling frequency of each power sensor group and sends the adjusted sensor number information to the energy distribution unit; The synchronization signal Fly SYNC is simultaneously transmitted to the flywheel control unit to control the flywheel speed sensor and the grid frequency detector to synchronously collect data.

[0010] Preferably, the energy distribution unit includes a load prediction module, a speed regulation module, an alternating controller and a state estimation module; wherein, The load forecasting module is used to analyze the historical load data of computing power equipment and obtain the load change characteristics in the future period; The speed regulation module is used to calculate the difference between the measurement data of the flywheel speed sensor and the grid frequency detector to determine the matching parameters between the flywheel speed and the grid frequency; An alternating controller is used to construct a state-space model of alternating control based on load variation characteristics and matching parameters, and to modify controller parameters using real-time load data as constraints; the modified flywheel charge and discharge rate is used as output; The state estimation module is used to dynamically integrate the corrected charge and discharge rates in combination with the real-time matching parameters received by the speed regulation module to predict the operating mode threshold of the flywheel energy storage device in the next period.

[0011] Preferably, the alternating controller adopts a dual-mode switching strategy, in which: When the grid frequency fluctuation exceeds the set threshold, the flywheel speed compensation is calculated based on the matching parameters and a charge and discharge rate correction instruction is generated; When the load change rate exceeds the set threshold, a flywheel charging and discharging priority sequence is generated based on the load forecast results, and the coordinated scheduling mechanism of the energy storage device is triggered.

[0012] Preferably, during the dual-mode switching process, a dynamic weight allocation algorithm is used to optimize the control parameters.

[0013] Preferably, the alternating controller uses load change characteristics and matching parameters to construct a state vector consisting of a flywheel speed compensation amount, a charge and discharge priority coefficient, and a remaining capacity ratio of the energy storage device under the grid frequency reference, and establishes a state transfer equation in combination with a load forecast error correction factor, a frequency fluctuation compensation factor, and a mode switching delay parameter; uses real-time load data as boundary conditions to perform parameter correction on the controller; and uses the corrected flywheel charge and discharge rate as output.

[0014] Preferably, During the control process, the rapid response process when the flywheel operating mode or load monitoring object switches includes: 1) Acquiring charge and discharge rate data output by the alternating controller of the flywheel or sensor at the switching moment according to the synchronization signal Fly SYNC; 2) calculating the available energy reserve of each group of flywheels based on the charge and discharge rate data and the energy capacity parameters of each flywheel energy storage device; 3) Based on the real-time data from the load monitoring unit, the matching degree between the energy reserve and the load demand is calculated to determine the dispatchable energy threshold of each flywheel energy storage device; 4) Obtaining the actual operating status parameters of each flywheel energy storage device at the switching moment, and calculating the deviation between the actual operating status parameters of each flywheel energy storage device and the dispatchable energy threshold of each flywheel energy storage device; when the deviation of a flywheel is lower than the set threshold, the flywheel is included in the current scheduling queue and the charge and discharge rate adjustment instruction is executed.

[0015] Preferably, each group of flywheel energy storage devices includes a plurality of independent flywheel units; and the plurality of independent flywheels in each group of flywheel units are connected to the power grid in a set topology.

[0016] Preferably, the independent flywheels in each group of flywheel units are connected in a ring redundant structure or a star centralized structure; wherein, In the ring redundant structure, three flywheels are connected in series to form a closed loop, with the central node connected to the power distribution bus; In the star-type centralized structure, four flywheels are connected in parallel to the same power node, and the central node is connected to the output of the load monitoring unit.

[0017] Preferably, the power sensor is arranged so as to cover the AC input terminal, DC power distribution cabinet and uninterruptible power supply output port of the computing power equipment.

[0018] Compared with the prior art, the present invention has the following beneficial effects: In terms of load monitoring and data collection, the load monitoring unit, through the deployment of multiple power sensors along the data center's power supply lines, comprehensively covers key nodes such as the AC input terminals, DC distribution cabinets, and uninterruptible power supply output ports of computing power equipment. This allows for real-time and accurate capture of real-time load data for each group of computing power equipment. This comprehensive monitoring coverage ensures the integrity and accuracy of load data, providing a reliable basis for subsequent charging and discharging decisions, and enabling the system to promptly respond to dynamic changes in computing power load.

[0019] The design of the energy storage unit is highly innovative, employing multiple flywheel energy storage devices installed in the data center's power network. Each flywheel energy storage device includes multiple independent flywheel units connected to the power network in a ring-shaped redundant structure or a star-shaped centralized structure. In the ring-shaped redundant structure, three flywheels are connected in series to form a closed loop, with the central node connected to the power distribution bus. This structure improves the reliability and redundancy of the system. Even if a flywheel fails, the other flywheels can still maintain normal operation of the system. In the star-shaped centralized structure, four flywheels are connected in parallel to the same power node, and the central node is connected to the output of the load monitoring unit, facilitating centralized monitoring and control and improving the system's collaborative working capabilities. The coordinated operation of multiple flywheels and the flexible topology design significantly enhance the system's energy storage capacity and regulation capabilities, and can better adapt to the energy recovery needs of computing loads of different sizes in data centers.

[0020] The flywheel control unit uses a flywheel speed sensor and a grid frequency detector to measure the flywheel energy storage device's speed and grid frequency fluctuations in real time, providing the system with key operating status parameters. Based on these parameters, the system achieves coordinated control of flywheel speed and grid frequency. When grid frequency fluctuations exceed a set threshold, the system promptly calculates flywheel speed compensation and generates charge and discharge rate correction instructions, ensuring grid frequency stability and improving power quality and system stability.

[0021] The energy distribution unit is the core control module of the system, and its innovations are significant. By establishing an alternating control strategy that coordinates load and frequency, and combining the load forecasting module with trend analysis of historical load data of computing power equipment, it is possible to accurately obtain the load change characteristics of future time periods and formulate charging and discharging plans in advance. The speed regulation module calculates the difference between the measured data of flywheel speed and grid frequency to determine the matching parameters, providing an important adjustment basis for the alternating controller. The alternating controller adopts a dual-mode switching strategy. When the load change rate exceeds the set threshold, it generates a flywheel charging and discharging priority sequence based on the load forecast results, triggering the coordinated scheduling mechanism of the energy storage device, realizing dynamic optimization scheduling of the flywheel energy storage device, and improving energy recovery efficiency and system response speed. At the same time, the alternating controller uses load change characteristics and matching parameters to construct a state vector, and combines multiple correction factors to establish a state transition equation. Using real-time load data as boundary conditions for parameter correction, it makes charging and discharging decisions more accurate and scientific, further improving the system's control performance and energy utilization.

[0022] Another highlight of this invention is the sequential synchronization control mechanism. This mechanism employs sequential synchronization control across the energy storage unit, load monitoring unit, and flywheel control unit. The master flywheel generates a synchronization signal, Fly SYNC, which enables synchronized switching of the operating states of each flywheel energy storage device, synchronized adjustment of the power sensor sampling frequency, and synchronized data collection from the flywheel control unit. This synchronization ensures consistent data exchange and action execution across all system units, improving the system's collaborative efficiency and overall performance while reducing control delays and errors caused by asynchronous operation.

[0023] In terms of the system's rapid response capability, when the flywheel's operating mode or load monitoring target switches, the system obtains the charge and discharge rate data at the switching moment through a synchronization signal. Combined with the flywheel's energy capacity parameters and real-time load data, it quickly calculates the available energy reserve, schedulable energy threshold, and deviation, promptly identifies the schedulable flywheel, and executes the charge and discharge rate adjustment instructions. This rapid response mechanism ensures the system's stability and reliability during dynamic changes, effectively responding to sudden changes in data center computing power and load, and further improving the system's practicality and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a working principle diagram of the flywheel energy storage system control system for data center computing load energy recovery according to the present invention; Figure 2 This is a working principle diagram of the energy distribution unit; Figure 3 This is the working principle diagram of the dual-mode switching strategy of the alternating controller. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 are within the scope of protection of the present invention.

[0026] See also Figure 1-Figure 3 The present invention relates to a flywheel energy storage system control system for data center computing load energy recovery, which includes: an energy storage unit, a load monitoring unit, a flywheel control unit, and an energy distribution unit. The specific implementation steps are as follows: The energy storage unit includes multiple groups of flywheel energy storage devices installed in the data center power network, and each group of flywheel energy storage devices can be in a charging or discharging state. The load monitoring unit includes multiple power sensors deployed on the power supply line of the data center. The monitoring direction of each power sensor covers the power input node of the computing power equipment, so that at least one power sensor can capture the real-time load data of a group of computing power equipment. The flywheel control unit includes a flywheel speed sensor and a grid frequency detector, which are used to measure the speed data of the flywheel energy storage device and the grid frequency fluctuation data respectively. The energy distribution unit uses the computing power equipment load data collected by the load monitoring unit to make flywheel charging and discharging decisions, establishes an alternating control strategy for coordinated load and frequency, uses grid frequency fluctuation data to dynamically adjust the charging and discharging decisions, and predicts the operating status of the flywheel energy storage device in the next time period based on the adjustment results, for rapid response to the next cycle of power dispatch.

[0027] The technical solution of the present invention is further described in detail below with reference to specific embodiments. Example

[0028] This embodiment uses sequential synchronization control for the energy storage unit, load monitoring unit, and flywheel control unit to ensure the coordination and consistency of the entire system during data collection, state switching, and instruction execution. The specific implementation method is as follows: For the timing synchronization control of energy storage units. The energy storage unit includes multiple groups of flywheel energy storage devices installed in the data center power network. Each group of devices is composed of multiple independent flywheel units, and each flywheel energy storage device needs to switch between charging and discharging states. To achieve synchronous control, each flywheel energy storage device needs to be uniquely numbered. The numbering rule can use a combination of "group number + single unit serial number", such as "Group1-01" and "Group2-02", to ensure that each flywheel has a unique identification in the system.

[0029] Based on the numbering, one of the flywheels is selected as the master flywheel. This selection can be based on pre-set rules, such as prioritizing the flywheel initially charged and with the highest remaining capacity, or through a self-selection mechanism during system initialization. The core function of the master flywheel is to generate the synchronization signal Fly SYNC, which contains key information such as a timestamp, command type (such as a mode switch command or data acquisition command), and synchronization period. The synchronization signal can be transmitted using optical fiber or high-speed Ethernet to ensure low latency and high reliability.

[0030] The remaining flywheels, acting as slaves, must perform the following operations upon receiving the Fly SYNC synchronization signal: First, they parse the command type and timestamp in the signal to confirm whether it is a valid synchronization command. If so, they then synchronously switch operating modes based on the command type, for example, from charging mode to discharging mode, or vice versa. After the switch is complete, the slave flywheel must package its own numbered operating status data (such as current speed, remaining capacity, charge and discharge rate, etc.) into a data frame and send it to the energy distribution unit via the established communication link. The data frame format must adhere to a unified communication protocol, for example, including fields such as a frame header, flywheel number, status parameters, and a checksum, to ensure that the energy distribution unit can accurately parse the data.

[0031] This system implements timing synchronization control for load monitoring units. Load monitoring units are deployed with multiple power sensors located at key nodes in the data center's power supply circuits, such as the AC input terminals of computing equipment, DC distribution cabinets, and uninterruptible power supply output ports. To achieve synchronized sampling, multiple groups of power sensors must be numbered. The numbers correspond to the areas or groups of devices they monitor. For example, "Sensor-A1" corresponds to the sensor monitoring the first group of computing equipment in area A.

[0032] The synchronization signal Fly SYNC adjusts the sampling frequency of each group of power sensors. Specifically, the synchronization signal generated by the master flywheel contains sampling frequency control parameters, such as setting the sampling period to 100ms or 50ms. After receiving the synchronization signal, each power sensor adjusts its sampling period based on the sampling frequency parameters and synchronously collects data at the beginning of each sampling period. For example, if the synchronization signal sets the sampling period to 100ms, all sensors will collect the voltage, current, power, and other parameters of the monitoring node at the beginning of every 100ms (such as 0ms, 100ms, 200ms, etc.).

[0033] After sampling is complete, the sensor must bind the collected data to its own serial number information, forming a data packet containing the sensor number, sampling time, and monitoring data, and then send it to the energy distribution unit. To avoid data transmission conflicts, communication mechanisms such as time division multiplexing (TDM) or carrier sense multiple access (CSMA) can be used to ensure that data from each sensor is transmitted in an orderly manner. Furthermore, synchronization signals can be used to calibrate the sensor clocks. By periodically sending time synchronization commands, the internal clocks of all sensors are kept consistent, avoiding inconsistent timestamps of sampled data due to clock deviation.

[0034] Timing synchronization control for the flywheel control unit. The flywheel control unit includes a flywheel speed sensor and a grid frequency detector, which measure the speed of the flywheel energy storage device and grid frequency fluctuations, respectively. The synchronization signal FlySYNC must be transmitted simultaneously to the flywheel control unit to control the synchronous data collection of these two sensors.

[0035] The flywheel speed sensor must be installed close to the flywheel's axis of rotation and utilize a non-contact measurement method (such as a photoelectric encoder or magnetoelectric sensor) to obtain the flywheel's real-time speed. Upon receiving a synchronization signal, the speed sensor triggers sampling at a specified moment (e.g., the rising edge of the synchronization signal), records the current flywheel speed, and combines this data with information such as the flywheel number and sampling time before transmitting it to the energy distribution unit.

[0036] The grid frequency detector monitors frequency fluctuations in the data center's power network. Its sampling points are typically located at the power network's ingress or at key busbar nodes. When a synchronization signal is transmitted to the frequency detector, it samples the grid frequency at the synchronization moment, obtaining the current frequency value and the rate of change (df / dt). This data is packaged into a frequency monitoring data packet and sent to the energy distribution unit.

[0037] To ensure timing synchronization among the three units (energy storage unit, load monitoring unit, and flywheel control unit), the Fly SYNC synchronization signal must possess a high-precision time base. The master flywheel can incorporate a high-precision clock module (such as a GPS-synchronized clock or atomic clock) to generate timestamps with nanosecond accuracy. The transmission delay of the synchronization signal must be controlled within the microsecond range and can be calibrated using hardware synchronization mechanisms (such as a dedicated synchronization bus) or software compensation algorithms (such as the NTP time synchronization protocol).

[0038] During system initialization, synchronization parameters must be configured, including the synchronization period, the master control flywheel, and the device number mapping table for each unit. The synchronization period can be flexibly adjusted based on the system's response requirements. For example, during periods of large load fluctuations, the synchronization period can be shortened to improve system real-time performance; during periods of stable load, the synchronization period can be extended to reduce communication overhead.

[0039] If a master flywheel fails during system operation, a master-slave switchover mechanism is required. For example, a new master flywheel is automatically elected from the slave flywheels. Before the new master flywheel takes over the synchronization signal generation task, it must synchronize its clocks with the other slave flywheels to ensure the continuity and accuracy of the synchronization signal during the switchover process.

[0040] At the data transmission level, the energy distribution unit must have a dedicated synchronous data receiving interface capable of simultaneously processing multiple sources of synchronous data (such as flywheel status data, sensor monitoring data, and frequency detection data). This interface must have data caching and timing alignment capabilities. For example, this interface can align data from different sources to the same point in time based on the timestamps in the data to facilitate subsequent collaborative analysis and control decisions.

[0041] Furthermore, timing synchronization control must consider electromagnetic compatibility (EMC) to prevent synchronization signals from being affected by electromagnetic interference from high-frequency equipment within the data center (such as servers and switches). Synchronization signal transmission cables must use shielded twisted-pair cables or optical fibers, and the wiring path must be away from strong electromagnetic sources to ensure signal integrity. Example

[0042] In this embodiment, the energy distribution unit includes a load prediction module, a speed regulation module, an alternating controller, and a state inference module. The load prediction module performs trend analysis on the historical load data of the computing power equipment to obtain the load change characteristics of the future period and provide forward-looking load information for charging and discharging decisions. The speed regulation module performs difference calculation on the measurement data of the flywheel speed sensor and the grid frequency detector to determine the matching parameter between the flywheel speed and the grid frequency, which reflects the degree of coordination between the flywheel speed and the grid frequency. The alternating controller constructs a state space model of alternating control based on the load change characteristics and matching parameters, uses real-time load data as a constraint condition to modify the controller parameters, and outputs the modified flywheel charge and discharge rate. The state inference module dynamically integrates the modified charge and discharge rate in combination with the real-time matching parameters received by the speed regulation module to predict the operating mode threshold of the flywheel energy storage device in the next period. Through the collaborative work of each module, the energy distribution unit can make more accurate charging and discharging decisions and operating state predictions, thereby achieving effective control of the flywheel energy storage device.

[0043] When building a load forecasting module, you first need to establish a complete historical load database. This database should contain computing equipment load data for different time periods (such as daily, weekly, monthly, and quarterly), along with metadata such as timestamps, equipment type, and business scenarios. The data collection frequency should be determined based on system requirements and equipment characteristics. For example, for computing equipment with high-frequency fluctuations, a collection frequency of seconds or milliseconds can be adopted. During the data preprocessing stage, the collected raw data needs to be cleaned to remove outliers and noise. Outlier detection can be performed using statistical methods (such as the 3σ principle) or machine learning algorithms (such as isolation forest and DBSCAN clustering). Missing data points can be filled using linear interpolation, spline interpolation, or time series forecasting methods based on historical data.

[0044] To characterize load fluctuations, the load forecasting module employs various time series analysis methods. First, trend analysis is performed, using methods such as moving average and exponential smoothing to identify long-term trends in load data. For example, for periodically fluctuating load data, seasonal decomposition can be used to decompose the data into trend, seasonal, and residual components. Simultaneously, the cyclical characteristics of the load data are analyzed to determine the length of the cycle and the amplitude of fluctuations. For load data with significant periodicity, a seasonal time series model (such as SARIMA) can be developed for forecasting. Furthermore, the volatility of the load data must be analyzed, and statistical indicators such as variance and standard deviation must be calculated to assess the severity of load fluctuations.

[0045] Based on trend analysis, the load forecasting module uses machine learning algorithms to build a forecasting model. Available algorithms include support vector regression (SVR), random forest, and long short-term memory (LSTM). The appropriate algorithm should be selected for different computing devices and business scenarios. For example, for load data with long-term dependencies, LSTM networks are better able to capture long-term patterns in time series. During model training, historical load data is divided into training and test sets. The model is trained using the training set, and its forecasting performance is evaluated using the test set. To improve the model's generalization capabilities, techniques such as cross-validation and regularization can be used.

[0046] The load forecasting module also needs to consider the impact of external factors on computing equipment load. These factors include weekdays / holidays, holidays, weather conditions, business activities, and more. Inputting these external factors into the forecasting model as features can further improve forecast accuracy. For example, computing equipment load patterns often differ significantly between weekdays and weekends. Including date type as a feature can help the model better capture these differences. For continuous external factors like weather conditions, discretization can be performed or feature engineering methods can be used to extract valuable information.

[0047] The core task of the speed regulation module is to calculate the matching parameters between the flywheel speed and the grid frequency. First, the flywheel speed sensor collects real-time flywheel speed data, while the grid frequency detector monitors grid frequency fluctuations in real time. These two sets of data require time synchronization to ensure consistent timestamps for accurate difference calculation. Time synchronization can be calibrated using a synchronization signal, FlySYNC, generated by the master control flywheel, similar to that used in Example 1.

[0048] To perform the difference calculation, the speed regulation module first converts the flywheel speed data to the same physical dimension as the grid frequency. For example, the flywheel speed (in revolutions per minute, RPM) is converted to frequency (in Hertz, Hz). The conversion formula is: Frequency (Hz) = Flywheel Speed (RPM) / 60 × Pole Pairs. Pole pairs are a key parameter of the flywheel motor and depend on the motor design. The difference between the converted flywheel frequency and the grid frequency is calculated to obtain the frequency deviation.

[0049] To more comprehensively reflect the match between flywheel speed and grid frequency, the speed regulation module calculates not only the frequency deviation but also the frequency rate of change (df / dt). The frequency rate of change reflects the dynamic characteristics of the grid frequency and is crucial for fast-response flywheel energy storage systems. The frequency rate of change is obtained by differentiating the grid frequency data. Simultaneously, the flywheel speed data is differentiated to obtain the flywheel frequency rate of change. These two rates of change are compared, and their difference and ratio are calculated as part of the matching parameter.

[0050] The speed control module also incorporates weighting factors to comprehensively assess the degree of matching. Different weights are assigned to the frequency deviation value and the frequency change rate difference value. The weightings are dynamically adjusted based on the system's operating status and control requirements. For example, when the grid frequency fluctuates significantly, the weighting of the frequency change rate difference value can be increased to prioritize the system's dynamic response performance. When the grid frequency is relatively stable, the weighting of the frequency deviation value can be increased to improve the system's steady-state accuracy. Ultimately, a comprehensive matching parameter is derived through weighted summation. The smaller the value of this parameter, the more closely matched the flywheel speed is to the grid frequency.

[0051] The alternating controller constructs a state-space model for alternating control based on the load variation characteristics provided by the load forecasting module and the matching parameters calculated by the speed regulation module. State-space models are a commonly used control system modeling method that comprehensively describes the dynamic characteristics of a system. In this system, the state variables of the state-space model include the speed, remaining energy, and charge / discharge rate of the flywheel energy storage device. Input variables include the load forecast value and matching parameters, and the output variable is the corrected flywheel charge / discharge rate.

[0052] The first step in building a state-space model is to determine the system's state equation and output equation. The state equation describes the relationship between the system's state variables and their time dependence, while the output equation describes the relationship between the system's output variables, the state variables, and the input variables. When developing these equations, it's important to consider the system's physical characteristics and control logic. For example, there's a physical relationship between the change in flywheel speed and the charge and discharge power, and corresponding equations can be established based on the law of conservation of energy. Furthermore, considering the nonlinear characteristics of the system, methods such as Taylor expansion can be used to linearize the nonlinear equations.

[0053] To improve model accuracy, the alternating controller uses real-time load data as constraints to modify the parameters of the state-space model. Specifically, the real-time load data is compared with the predicted values from the load forecasting module to calculate the prediction error. Based on the magnitude and direction of the prediction error, the parameters of the state-space model are adjusted to ensure that the model better reflects the actual system operation. Parameter modification can be performed using adaptive control algorithms (such as model reference adaptive control and self-tuning PID control) or optimization algorithms (such as least squares and gradient descent).

[0054] The alternating controller also needs to consider system constraints. These include the maximum charge and discharge power limits of the flywheel energy storage device, upper and lower speed limits, and the range of remaining energy. When solving the state-space model, these constraints are incorporated into the optimization problem to ensure that the charge and discharge rates output by the controller are within the system's permitted range. For example, when the remaining energy in the flywheel approaches its lower limit, the discharge rate is limited to prevent overdischarge; when the flywheel speed approaches its upper limit, the charge rate is limited to prevent overspeeding.

[0055] The state estimation module's primary function is to predict the flywheel energy storage device's operating mode threshold for the next time period. This module dynamically integrates the corrected charge and discharge rates output by the alternating controller with the real-time matching parameters received by the speed regulation module. This dynamic integration process accounts for the system's inertia and delay characteristics. By integrating the charge and discharge rates over time, it derives the energy change of the flywheel energy storage device over the next period.

[0056] During the integration process, the state inference module introduces an attenuation factor to simulate the system's energy loss. The magnitude of the attenuation factor depends on factors such as the flywheel energy storage device's efficiency and heat dissipation. The specific value of the attenuation factor is determined through experimental or theoretical analysis and dynamically adjusted during the integration process. Furthermore, to account for the uncertainty of grid frequency fluctuations, the state inference module performs a probabilistic analysis of the matching parameters and calculates the energy variation range under different matching conditions.

[0057] Based on the dynamic integration results, the state estimation module calculates the operating mode thresholds for the flywheel energy storage device for the next period. These operating mode thresholds include charging, discharging, and standby thresholds. When the predicted energy change exceeds the charging threshold, the system switches to charging mode; when the predicted energy change falls below the discharging threshold, the system switches to discharging mode; and when the predicted energy change is within the standby threshold, the system maintains its current state or enters standby mode.

[0058] The state inference module also needs to consider the system's response time and safety margin. To ensure the system can promptly respond to load changes and grid frequency fluctuations, a safety margin is reserved when calculating thresholds. For example, the charging threshold is set slightly higher than the actual required value, while the discharging threshold is set slightly lower than the actual required value to avoid frequent system mode switching. Furthermore, the threshold value is adjusted based on the system's response time characteristics to ensure that the system can complete mode switching within the specified time.

[0059] During actual operation, the state inference module continuously updates its predictions and adjusts operating mode thresholds based on the latest forecasts. When the system's operating state changes, such as a sudden increase in load or a significant fluctuation in grid frequency, the state inference module quickly recalculates the thresholds, providing timely and accurate decision-making for the energy distribution unit. This dynamic prediction and adjustment mechanism enables the system to proactively prepare for changes in load and grid fluctuations, improving its responsiveness to them and enabling more efficient energy distribution and control.

[0060] Through the collaborative work of the load forecasting module, speed regulation module, alternating controller, and state inference module, the energy distribution unit can achieve precise control of the flywheel energy storage device. The load forecasting module provides forward-looking load information to help the system plan charging and discharging strategies in advance; the speed regulation module calculates the matching parameters between the flywheel speed and the grid frequency, providing key inputs for the controller; the alternating controller makes parameter corrections and control decisions based on the state space model and real-time data; the state inference module predicts future operating mode thresholds and guides the dynamic adjustment of the system. This multi-level, multi-dimensional control architecture enables the system to adapt to different operating conditions and external environmental changes, achieves efficient recovery and utilization of data center computing load energy, and improves the stability and reliability of the entire system. Example

[0061] The alternating controller of this embodiment adopts a dual-mode switching strategy. When the grid frequency fluctuation exceeds the set threshold, the flywheel speed compensation is calculated based on the matching parameter, and a charge and discharge rate correction instruction is generated to cope with the impact of the grid frequency fluctuation on the system and ensure the stability of the system. When the load change rate exceeds the set threshold, a flywheel charge and discharge priority sequence is generated based on the load forecast result, and the collaborative scheduling mechanism of the energy storage device is triggered, so that the energy storage device can adjust the charge and discharge strategy in time according to the rapid change of the load, thereby improving the system's responsiveness to load changes. During the dual-mode switching process, a dynamic weight allocation algorithm is used to optimize the control parameters, and the weights of the control parameters are dynamically adjusted according to different operating conditions and parameter changes, thereby further improving the performance and adaptability of the controller.

[0062] When building a dual-mode switching strategy for an alternating controller, it's first necessary to set the grid frequency fluctuation threshold and the load change rate threshold. These thresholds must be set based on the grid's stability requirements, the performance limitations of the flywheel energy storage device, and the operational needs of the data center's computing equipment. The grid frequency fluctuation threshold is typically determined based on the grid's nominal frequency (e.g., 50Hz or 60Hz) and the permissible fluctuation range, for example, set to ±0.5Hz or ±1Hz. The load change rate threshold is determined based on the data center's business characteristics and statistical analysis of historical load data, for example, setting the load change per minute to no more than 10% or 20% of the rated load.

[0063] When grid frequency fluctuations exceed a set threshold, the alternating controller enters frequency compensation mode. In this mode, the flywheel speed compensation is further calculated based on the matching parameters calculated by the speed regulation module. The matching parameters reflect the degree of coordination between the flywheel speed and the grid frequency, including information such as the frequency deviation and the frequency rate of change difference. By analyzing these parameters, the extent of flywheel speed adjustment required to compensate for grid frequency fluctuations is determined.

[0064] When calculating the flywheel speed compensation, several factors need to be considered. The first is the amplitude and direction of the grid frequency fluctuation. The greater the amplitude, the greater the compensation required; the direction of the fluctuation determines whether the flywheel speed needs to be increased or decreased. The second factor is the current state of the flywheel energy storage device, including speed and remaining energy. If the flywheel speed is approaching the upper or lower limit, or if the remaining energy is insufficient, the compensation may need to be limited to ensure safe system operation. Furthermore, the system's response time and inertia characteristics must be considered to avoid system oscillations caused by over- or under-compensation.

[0065] Based on the calculated flywheel speed compensation, the alternating controller generates a charge / discharge rate correction command. This command is sent to the flywheel control unit, which controls the flywheel energy storage device to adjust the charge / discharge rate, thereby regulating the flywheel speed. When generating the correction command, the flywheel's charge / discharge efficiency and energy conversion losses must be considered. Different charge / discharge rates may correspond to different efficiencies, so it is necessary to select a charge / discharge rate that meets the speed regulation requirements while ensuring high energy conversion efficiency.

[0066] When the load change rate exceeds a set threshold, the alternating controller enters load response mode. In this mode, a priority sequence for flywheel charging and discharging is generated based on the load forecast provided by the load forecasting module. The load forecast includes information such as load change trends, peak and valley values for future periods. Based on this information, the flywheel energy storage device's ability and adaptability to load changes is evaluated, thereby determining its charging and discharging priority.

[0067] When generating a charge and discharge priority sequence, several factors need to be considered. The first is the remaining energy level of the flywheel energy storage device. Flywheels with higher remaining energy levels can provide more energy support in discharge mode and therefore have a higher priority. In charging mode, flywheels with lower remaining energy levels require priority charging to restore their energy storage capacity. The second factor is the flywheel's response speed and adjustability. Different flywheels may have different response times and adjustment ranges, so flywheels with faster response speeds and stronger adjustability have higher priority. Furthermore, the flywheel's historical operating status and health status must be considered to avoid premature damage caused by overuse of certain flywheels.

[0068] Based on the generated charge and discharge priority sequence, the alternating controller triggers the coordinated scheduling mechanism for the energy storage devices. This mechanism coordinates the charging and discharging operations of multiple flywheel energy storage devices according to the priority sequence, enabling them to work together to cope with rapid load fluctuations. During the coordinated scheduling process, it is necessary to ensure that the charging and discharging operations of the individual flywheels do not interfere with each other, while maximizing the overall system responsiveness and energy efficiency.

[0069] During the dual-mode switching process, the alternating controller uses a dynamic weight allocation algorithm to optimize control parameters. This algorithm dynamically adjusts the weights of control parameters based on various operating conditions and parameter changes. Operating conditions include factors such as grid frequency fluctuations, load change rate, and the status of the flywheel energy storage device. By monitoring these factors in real time, the current system operating status is assessed and the control parameter weights are dynamically adjusted based on pre-set rules or learning algorithms.

[0070] The core of the dynamic weight allocation algorithm is to establish a weight adjustment model. This model can be constructed based on rules, machine learning, or optimization algorithms. Rule-based models adjust weights according to preset conditions and rules. For example, when the grid frequency fluctuates significantly, the weight of parameters related to the frequency compensation mode is increased; when the load rate of change is high, the weight of parameters related to the load response mode is increased. Machine learning models, by learning from historical data, automatically discover the optimal weight allocation scheme under different operating conditions. Optimization algorithm models transform the weight adjustment problem into an optimization problem and determine the weights by solving the optimal solution.

[0071] In practical applications, dynamic weight allocation algorithms require real-time access to system operating status data for rapid calculations and decision-making. To improve the algorithm's efficiency and real-time performance, a distributed computing architecture or hardware acceleration can be employed. Furthermore, the algorithm must be regularly updated and optimized to adapt to changing operating environments and system requirements.

[0072] Alternating controllers also need to consider the smooth transition during mode switching. If the system switches from one mode to another too abruptly, this can lead to system instability or significant fluctuations. To avoid this, alternating controllers employ a smooth transition strategy, gradually adjusting control parameters and commands during the mode switching process to ensure a smooth transition from one operating state to another.

[0073] Smooth transition strategies can be implemented in a variety of ways. One approach is to precalculate and prepare the incoming mode before switching, then gradually apply the new control parameters and instructions at the moment of transition. Another approach is to use an interpolation algorithm to interpolate between the control parameters of the two modes, generating a smooth transition curve along which the system's response gradually changes. Furthermore, a transition period can be set during which the influence of both modes is considered simultaneously, gradually reducing the influence of the original mode and increasing the influence of the new mode until the system fully switches to the new mode.

[0074] The alternating controller also needs to have fault detection and handling capabilities. During the dual-mode switching process, various faults may occur, such as sensor failure, communication failure, and actuator failure. The alternating controller needs to monitor the system's operating status in real time, promptly detect faults, and take appropriate measures. For example, when a sensor failure is detected, data from a redundant sensor or model-based estimates can be used as a replacement. When a communication failure occurs, attempts can be made to reestablish the communication connection or switch to an alternate communication channel. When an actuator fails, the faulty device can be promptly shut down and the operating status of other devices adjusted to ensure safe and stable system operation.

[0075] To improve the reliability and maintainability of the alternate controller, comprehensive logging and diagnostic capabilities are also required. Logging can record system operating status, control parameters, fault information, and more, providing a basis for subsequent analysis and maintenance. Diagnostics can assess the system's operating status in real time, identify potential problems, and provide recommendations and solutions. These features allow operations and maintenance personnel to promptly understand system performance, quickly locate and resolve issues, and reduce system downtime and maintenance costs.

[0076] Through the alternating controller's dual-mode switching strategy and dynamic weight allocation algorithm, the system automatically adjusts its control strategy based on varying operating conditions, improving its responsiveness to grid frequency fluctuations and load changes. The frequency compensation mode effectively addresses grid frequency fluctuations and maintains system stability; the load response mode rapidly responds to load changes, ensuring the normal operation of data center computing equipment. The dynamic weight allocation algorithm further optimizes control parameters, enabling the system to achieve optimal control results under various operating conditions. This intelligent, adaptive control strategy provides strong technical support for flywheel energy storage systems used for data center computing load energy recovery, enabling them to better meet the energy efficiency and stability requirements of modern data centers. Example

[0077] The alternating controller of this embodiment uses the load change characteristics and matching parameters to construct a state vector, which is composed of the flywheel speed compensation amount under the grid frequency reference, the charge and discharge priority coefficient, and the remaining capacity ratio of the energy storage device. At the same time, the state transition equation is established in combination with the load forecast error correction factor, the frequency fluctuation compensation factor and the mode switching delay parameter. The controller parameters are corrected using real-time load data as boundary conditions, and the corrected flywheel charge and discharge rate is used as output. During the control process, when the flywheel operating mode or the load monitoring object is switched, a rapid response is achieved through the following steps: First, the construction of the state vector of the alternating controller requires the integration of multi-dimensional information. The flywheel speed compensation under the grid frequency reference is generated based on the frequency deviation and change rate calculated by the speed regulation module, reflecting the dynamic matching requirements of the flywheel speed and the grid frequency. The charge and discharge priority coefficient is determined by the load change trend output by the load forecasting module and the current state of the energy storage device (such as remaining capacity and speed). It is used to quantify the priority order of different flywheels in the charge and discharge scheduling. The remaining capacity ratio of the energy storage device is directly related to the available energy of the flywheel and is a key indicator for determining whether it can participate in the current charge and discharge task. These three components constitute the core elements of the state vector. The dimensional consistency is ensured through normalization processing to facilitate subsequent mathematical modeling.

[0078] The establishment of the state transition equation must take into account the dynamic characteristics of the system and external interference factors. The load forecast error correction factor is used to compensate for the deviation between the predicted value and the real-time load data, and the equation parameters are dynamically adjusted by real-time monitoring of the load fluctuation amplitude; the frequency fluctuation compensation factor is based on the measured data of the grid frequency detector, reflecting the impact of frequency fluctuations on the flywheel speed and energy conversion efficiency; the mode switching delay parameter is used to simulate the time lag from the controller receiving the switching command to completing the state adjustment, and the specific value is determined through historical data statistics or hardware characteristic testing. The state transition equation describes the evolution of the state vector over time in the form of a differential equation or a difference equation. For example, the delay characteristics of the flywheel speed change are simulated by a first-order inertia link, or the coupling relationship between the various parameters is reflected by a linear combination model.

[0079] When real-time load data is input, the alternating controller uses it as a boundary condition to modify the parameters of the state transition equation. The specific process is as follows: first, the absolute error and relative error between the load prediction value and the measured value are calculated. If the error exceeds a preset threshold (such as 5% of the rated load), the correction mechanism is triggered. Correction methods include adjusting the weight coefficients of each component in the state vector, updating the gain parameters of the state transition equation, or reinitializing the integral term of the equation. For example, when the measured load is significantly higher than the predicted value, the weight of the charge and discharge priority coefficient is increased to accelerate the discharge rate of the high-priority flywheel; if the grid frequency fluctuation intensifies, the value of the frequency fluctuation compensation factor is increased to enhance the dynamic regulation of the flywheel speed.

[0080] In the flywheel operation mode or load monitoring object switching scenario, the rapid response mechanism is implemented through the following steps: Data Acquisition and Synchronization: Based on the timestamp of the Fly SYNC synchronization signal, the charge and discharge rate data output by the alternating controller at the switching moment is captured. This data, including the current charge and discharge rate command values and state vector parameters of each flywheel, is transmitted to the energy distribution unit's cache module via a dedicated communication channel with microsecond latency, ensuring real-time and complete data.

[0081] Calculation of Available Energy Reserve: Based on charge and discharge rate data and the energy capacity parameters of each flywheel energy storage device (such as rated capacity and energy conversion efficiency), the available energy reserve of each flywheel group at the switching moment is calculated through integration. The calculation formula is: Available Energy = Initial Remaining Energy + Charge and Discharge Rate × Switching Time Interval × Efficiency Factor. The efficiency factor takes into account energy losses during the charge and discharge process (such as mechanical and electromagnetic losses) and is determined based on device nameplate parameters or historical operating data.

[0082] Matching Calculation and Threshold Determination: Combined with real-time data from load monitoring units (such as the current active and reactive power requirements of computing equipment), the matching degree between energy reserves and load demand is calculated. This matching metric can be defined as the ratio of available energy to load demand, or the Euclidean distance can be used to measure the degree of deviation between energy supply and demand. Based on the matching results, a dispatchable energy threshold is set for each flywheel energy storage device. This threshold must include a safety margin (e.g., 10% of rated capacity) to prevent damage to the device due to overcharging and discharging.

[0083] Deviation Calculation and Scheduling Queue Generation: The actual operating parameters of each flywheel at the time of switching (such as real-time speed, remaining energy, and charge / discharge current) are obtained and compared with the adjustable energy measurement threshold to calculate the deviation. The deviation calculation formula is: Deviation = |Actual Operating Parameter - Threshold Parameter| / Threshold Parameter × 100%. When a flywheel's deviation falls below the set threshold (e.g., 15%), it is considered dispatchable and added to the current scheduling queue. A rate adjustment instruction is generated based on the charge / discharge priority factor. If the deviation exceeds the threshold, an early warning mechanism is triggered, prompting operations and maintenance personnel to check the equipment status.

[0084] During rapid response, time synchronization must be ensured across all links. The Fly SYNC synchronization signal not only triggers data acquisition but also serves as the time reference for each computational step. Global clock calibration mechanisms (such as the IEEE 1588 precision clock protocol) ensure that timestamp errors between modules are less than 1 microsecond. Furthermore, the energy distribution unit must possess high-speed data processing capabilities, accelerating the computational process through field-programmable gate arrays (FPGAs) or graphics processing units (GPUs), ensuring that the total latency from data acquisition to command output is kept within milliseconds.

[0085] The parameters of the state vector and state transition equations require regular offline optimization. By reviewing historical operating data and employing intelligent optimization algorithms such as particle swarm optimization (PSO) and genetic algorithms (GA), the weight coefficients of the state vector and the gain parameters of the state transition equations are iteratively optimized to accommodate long-term changes in data center load characteristics (such as changes in load patterns caused by business expansion and equipment upgrades). Constraints must be set during the optimization process to ensure that parameter values remain within the safe operating range of the equipment and to avoid over-optimization that could reduce system robustness.

[0086] In summary, Example 4 realizes the refined modeling and control of the flywheel energy storage device by constructing a state vector containing multi-dimensional parameters and a dynamically adjusted state transfer equation. The rapid response mechanism ensures that the system can quickly adjust the scheduling strategy and maintain the balance of energy supply and demand when the operating mode or monitoring object is switched through synchronous data acquisition, energy reserve calculation, matching analysis and deviation determination. This solution deeply couples load changes, grid frequency fluctuations and flywheel states to form a closed-loop control system, which provides key technical support for the efficient and stable operation of the data center computing load energy recovery system. Example

[0087] In this embodiment, each flywheel energy storage device comprises multiple independent flywheel units, each connected to the data center power network in a predefined topology. Each flywheel unit is interconnected using either a ring-shaped redundant structure or a star-shaped centralized structure. Power sensors are positioned across key power nodes of the computing equipment to enable comprehensive load monitoring.

[0088] 1. Flywheel unit topology design ① Ring redundant structure In this structure, three independent flywheels (denoted as F1, F2, and F3) are connected in series via a power electronics interface to form a closed loop. A central node (denoted as Node_C) is connected to the data center's power distribution busbar. Specifically, the output of F1 is connected to the input of F2, the output of F2 is connected to the input of F3, and the output of F3 is reversely connected to the input of F1, forming a ring circuit. The central node, Node_C, extends from the connection point between F1 and F2 and connects to the power grid via a busbar cable. The advantage of this structure lies in its redundancy: if any flywheel fails (e.g., F2 shuts down), the ring loop automatically disconnects the fault point, allowing the remaining two flywheels (F1 and F3) to continue supplying power to the busbar via Node_C, maintaining system operation. Furthermore, the ring structure balances the charge and discharge loads of each flywheel, preventing overload on any single device.

[0089] ②Star-shaped centralized structure The structure consists of four independent flywheels (denoted as F4, F5, F6, and F7), all connected in parallel to the same power node (denoted as Node_S). The input / output of each flywheel is directly connected to Node_S via an independent cable. The central node is led out from Node_S and is also connected to the output of the load monitoring unit. The star structure is characterized by centralized control: the load monitoring unit can obtain the voltage and current data of all flywheels in real time through Node_S, facilitating the rapid calculation of the total energy storage capacity and charge and discharge rate. When the number of flywheels needs to be expanded, it is only necessary to connect the new flywheel in parallel to Node_S without changing the original topology, and it has strong scalability. However, this structure relies on the reliability of the central node Node_S, and redundant power switches need to be configured to prevent single point failures.

[0090] 2. Power Sensor Layout Power sensors are deployed at the third-level monitoring nodes of the data center's power supply lines, covering the entire power input chain of computing equipment: Level 1 node: AC input terminal Micro power sensors are deployed at the AC input ports (such as IEC C13 / C14 interfaces) of computing servers. Each sensor monitors the real-time power consumption (active power P, reactive power Q, apparent power S) of a single server. The sensors are magnetically or snap-on to the outside of the terminals, supporting plug-and-play operation. The sampling frequency is at least 1kHz to capture high-frequency load fluctuations (such as instantaneous peak CPU power consumption).

[0091] Secondary node: DC distribution cabinet Hall-effect current sensors and voltage transmitters are installed in the output branches of the DC distribution cabinet to monitor the DC current I_dc and voltage U_dc of each branch. The sensors are connected to the load monitoring unit via an RS-485 bus with a 100ms communication cycle. This unit is used to calculate regional load (for example, the total power consumption of all servers in a cabinet).

[0092] Level 3 node: Uninterruptible power supply (UPS) output port A high-precision power analyzer is deployed on the UPS's AC output bus to monitor the total output power P_total and frequency f. This sensor, equipped with harmonic analysis capabilities, detects voltage distortion rate (THD_u) and current distortion rate (THD_i), with a sampling period of 1 second. This is used to assess the power quality of the entire data center and the compensation effectiveness of the flywheel energy storage system.

[0093] The synergy between topology and monitoring nodes The topology of the flywheel unit and the arrangement of the power sensors form a closed data loop: ① In a ring-based redundant architecture, when a load fluctuation in a computing device group is detected by a primary node sensor, the load monitoring unit analyzes real-time power data to determine whether to trigger flywheel charging or discharging. If discharge is required, the flywheel with the highest remaining capacity in the ring (such as F1) is prioritized, adjusting its speed to release energy while simultaneously monitoring the status of the other flywheels (F2 and F3) as backup.

[0094] ② In a star-type centralized configuration, when the three-level node sensors detect UPS output frequency fluctuations (e.g., f deviates from 50Hz±0.2Hz), the flywheel control unit synchronizes the speeds of all parallel flywheels based on the frequency detector data. For example, when the frequency is below the nominal value, the flywheels are controlled to discharge faster, injecting active power into the grid; when the frequency is above the nominal value, the flywheels are controlled to decelerate and charge, absorbing excess energy.

[0095] Formula description (contains only 1 structure-related formula) In the ring redundant structure, the equivalent output voltage U_ring of the flywheel group can be expressed as:

[0096] in: is the output voltage of the central node Node_C of the ring structure (unit: V); They are the terminal voltages of flywheels F1, F2, and F3 (unit: V), and their polarity is determined by the connection direction (F1 and F2 are forward connected, and F3 is reverse connected).

[0097] This formula reflects the vector superposition characteristics of the flywheel voltages in the ring structure. By adjusting the speed of different flywheels (changing the back electromotive force), the output voltage of Node_C can be dynamically adjusted to match the grid demand or the supply voltage fluctuations of the computing power equipment.

[0098] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0099] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A flywheel energy storage system control system for data center computing load energy recovery, characterized in that: include: Energy storage unit, load monitoring unit, flywheel control unit and energy distribution unit; An energy storage unit, comprising a plurality of flywheel energy storage devices disposed in a data center power network; Each set of flywheel energy storage devices is in a charging or discharging state; The load monitoring unit includes multiple power sensors deployed on the power supply lines of the data center. Each power sensor monitors the power input nodes of the computing power equipment, so that at least one power sensor captures real-time load data of a group of computing power equipment. A flywheel control unit includes a flywheel speed sensor and a grid frequency detector, which respectively measure the speed data of the flywheel energy storage device and the grid frequency fluctuation data; The energy distribution unit uses the computing equipment load data collected by the load monitoring unit to make flywheel charging and discharging decisions; and establishes an alternating control strategy that coordinates load and frequency, and uses grid frequency fluctuation data to dynamically adjust charging and discharging decisions; based on the adjustment results, it predicts the operating status of the flywheel energy storage device in the next time period for rapid response to the next cycle of power dispatch.

2. A flywheel energy storage system control system for data center computing load energy recovery according to claim 1, characterized in that: The energy storage unit, load monitoring unit and flywheel control unit are controlled by sequential synchronization; specifically including: The flywheel energy storage devices in the energy storage unit are numbered, and one of the flywheels is used as the master flywheel to generate the synchronization signal Fly SYNC; the remaining flywheels switch modes synchronously after receiving the synchronization signal Fly SYNC, and send the operating status data with the flywheel number to the energy distribution unit; The multiple power sensors of the load monitoring unit are numbered; the synchronization signal Fly SYNC adjusts the sampling frequency of each power sensor group and sends the adjusted sensor number information to the energy distribution unit; The synchronization signal Fly SYNC is simultaneously transmitted to the flywheel control unit to control the flywheel speed sensor and the grid frequency detector to synchronously collect data.

3. A flywheel energy storage system control system for data center computing load energy recovery according to claim 1, characterized in that: The energy distribution unit includes a load prediction module, a speed regulation module, an alternating controller and a state estimation module; wherein, The load forecasting module is used to analyze the historical load data of computing power equipment and obtain the load change characteristics in the future period; The speed regulation module is used to calculate the difference between the measurement data of the flywheel speed sensor and the grid frequency detector to determine the matching parameters between the flywheel speed and the grid frequency; An alternating controller is used to construct a state-space model of alternating control based on load variation characteristics and matching parameters, and to modify controller parameters using real-time load data as constraints; the modified flywheel charge and discharge rate is used as output; The state estimation module is used to dynamically integrate the corrected charge and discharge rates in combination with the real-time matching parameters received by the speed regulation module to predict the operating mode threshold of the flywheel energy storage device in the next period.

4. A flywheel energy storage system control system for data center computing load energy recovery according to claim 3, characterized in that: The alternating controller adopts a dual-mode switching strategy, in which: When the grid frequency fluctuation exceeds the set threshold, the flywheel speed compensation is calculated based on the matching parameters and a charge and discharge rate correction instruction is generated; When the load change rate exceeds the set threshold, a flywheel charging and discharging priority sequence is generated based on the load forecast results, and the coordinated scheduling mechanism of the energy storage device is triggered.

5. A flywheel energy storage system control system for data center computing load energy recovery according to claim 4, characterized in that: During the dual-mode switching process, a dynamic weight allocation algorithm is used to optimize the control parameters.

6. A flywheel energy storage system control system for data center computing load energy recovery according to claim 3, characterized in that: The alternating controller uses load variation characteristics and matching parameters to construct a state vector consisting of a flywheel speed compensation amount, a charge and discharge priority coefficient, and a remaining capacity ratio of the energy storage device under a grid frequency reference. It also establishes a state transition equation by combining a load forecast error correction factor, a frequency fluctuation compensation factor, and a mode switching delay parameter. The controller is parameterized using real-time load data as boundary conditions, and the corrected flywheel charge and discharge rate is output.

7. A flywheel energy storage system control system for data center computing load energy recovery according to claim 6, characterized in that: During the control process, the rapid response process when the flywheel operating mode or load monitoring object switches includes: 1) Acquiring charge and discharge rate data output by the alternating controller of the flywheel or sensor at the switching moment according to the synchronization signal Fly SYNC; 2) calculating the available energy reserve of each group of flywheels based on the charge and discharge rate data and the energy capacity parameters of each flywheel energy storage device; 3) Based on the real-time data from the load monitoring unit, the matching degree between the energy reserve and the load demand is calculated to determine the dispatchable energy threshold of each flywheel energy storage device; 4) Obtaining the actual operating status parameters of each flywheel energy storage device at the switching moment, and calculating the deviation between the actual operating status parameters of each flywheel energy storage device and the dispatchable energy threshold of each flywheel energy storage device; when the deviation of a flywheel is lower than the set threshold, the flywheel is included in the current scheduling queue and the charge and discharge rate adjustment instruction is executed.

8. A flywheel energy storage system control system for data center computing load energy recovery according to any one of claims 1 to 7, characterized in that: Each group of flywheel energy storage devices includes multiple independent flywheel units; and the multiple independent flywheels of each group of flywheel units are connected to the power network in a set topology structure.

9. The flywheel energy storage system control system for data center computing load energy recovery according to claim 7, characterized in that: The connection mode of the independent flywheels in each flywheel unit is a ring redundant structure or a star centralized structure; In the ring redundant structure, three flywheels are connected in series to form a closed loop, with the central node connected to the power distribution bus; In the star-type centralized structure, four flywheels are connected in parallel to the same power node, and the central node is connected to the output of the load monitoring unit.

10. The flywheel energy storage system control system for data center computing load energy recovery according to claim 1, characterized in that: The power sensor is arranged at a position covering the AC input terminal, DC distribution cabinet and uninterruptible power supply output port of the computing power equipment.

Images