Transient voltage treatment control method based on data center flywheel energy storage system

Abstract

The invention relates to the technical field of data center power supply, and discloses a transient voltage treatment control method based on a data center flywheel energy storage system. The method comprises the following steps: firstly, acquiring three-phase voltage, current and frequency signals of a power supply bus of a data center and rotating speed and energy storage power signals of a flywheel energy storage system, and calculating a transient voltage characteristic matrix; then, a dynamic threshold value is quantified, a multi-layer cascade evaluation network structure is constructed to carry out voltage transient evaluation, and energy storage control evaluation parameters are generated; then establishing a transient constraint function, and calculating a dynamic stability boundary of the flywheel energy storage system; and finally, iteratively optimizing energy storage control parameters according to the result to obtain a flywheel energy storage power regulation sequence. Through closed-loop adjustment, energy storage power is dynamically adjusted according to load requirements and voltage deviation, transient fluctuation of power supply bus voltage is effectively treated, and power supply stability and reliability of the data center are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power supply for data centers, and specifically to a transient voltage governance control method based on a flywheel energy storage system in a data center. Background Art

[0002] In the current rapid development of modern information technology, data centers, as the key hubs for information storage, processing, and transmission, are increasing in scale and importance day by day. There are a large number of electronic devices in data centers, which have extremely high requirements for power supply reliability and power quality. Once there is a power supply problem, even a short-term voltage fluctuation may cause equipment failures and data loss, bringing huge losses to enterprises and society.

[0003] The transient voltage problem is a major challenge faced by the power supply of data centers. There are many complex factors in the power system, such as power grid short-circuit faults, start-stop of large-capacity equipment, lightning strikes, etc., which will all cause transient voltage fluctuations. When a short-circuit fault occurs in the power grid, the instantaneous current change will cause the voltage of the power supply bus to drop sharply, affecting the normal operation of the data center equipment. When a large-capacity equipment starts, it will draw a large amount of current in a short time, causing a voltage sag; when the equipment stops running, it may cause a voltage swell. Natural disasters such as lightning strikes will also impact the power grid, thereby affecting the stability of the power supply voltage of the data center.

[0004] Traditional voltage governance means, such as using static var compensators (SVC) and static synchronous compensators (STATCOM), have certain limitations in dealing with transient voltage problems. The response speed of SVC is relatively slow, and it is difficult to make timely and effective compensation at the moment of rapid voltage fluctuations. Although STATCOM has a relatively fast response speed, in dealing with complex transient conditions, its control strategy may not be able to accurately adapt, resulting in poor compensation effects. Moreover, when facing the characteristics of rapid load changes in data centers, these traditional devices cannot flexibly adjust their working states and are difficult to meet the requirements of high-quality transient voltage governance in data centers.

[0005] As a new type of energy storage technology, the flywheel energy storage system has shown unique advantages in the field of data center power supply. It has the characteristics of fast response speed, high charging and discharging efficiency, and long service life. It can quickly release or absorb energy when transient voltage fluctuates, providing stable power support for data centers. However, the control strategy of the flywheel energy storage system in the current application of data centers is not perfect enough. The existing control methods often find it difficult to accurately collect and analyze various signals of the power supply bus of the data center, and cannot accurately judge the characteristics and changing trends of transient voltages. When evaluating the effect of the flywheel energy storage system on transient voltage control, there is also a lack of effective evaluation system and precise control parameter optimization method, which leads to the flywheel energy storage system cannot give full play to its advantages and cannot solve the transient voltage problem of the data center well. Therefore, it is urgent to develop an efficient and accurate transient voltage control method based on the flywheel energy storage system of the data center. Summary of the invention

[0006] The purpose of the present invention is to provide a transient voltage management control method based on a flywheel energy storage system in a data center to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: a transient voltage management control method based on a flywheel energy storage system in a data center, the method comprising: Collect the three-phase voltage signal, current signal, frequency signal of the data center power supply bus, the speed signal and energy storage power signal of the flywheel energy storage system, and calculate the transient voltage characteristic matrix; Performing dynamic threshold quantization on the transient voltage characteristic matrix to generate a dynamic parameter matrix, and constructing a multi-layer cascade evaluation network structure to perform voltage transient evaluation to generate energy storage control evaluation parameters; Establishing a transient constraint function according to the energy storage control evaluation parameter, and calculating the dynamic stability boundary of the flywheel energy storage system; The energy storage control parameters are iteratively optimized according to the transient voltage characteristic matrix and the dynamic stability boundary to obtain a flywheel energy storage power adjustment sequence.

[0008] Preferably, the three-phase voltage signal, current signal, frequency signal of the power supply bus of the data center, the speed signal and energy storage power signal of the flywheel energy storage system are collected, and the transient voltage characteristic matrix is ​​calculated, including: The three-phase voltage signal, current signal, frequency signal of the power supply bus, the speed signal of the flywheel energy storage system and the energy storage power signal are synchronously sampled to generate a transient observation data matrix; Performing interval normalization processing on the transient observation data matrix to obtain a normalized data matrix, and calculating the normalized data matrix based on covariance analysis to generate initial element values ​​of a dynamic disturbance characteristic matrix; Perform phase compensation on the initial element values of the dynamic disturbance feature matrix, and generate a corrected dynamic disturbance feature matrix by introducing a dynamic correction coefficient for tuning. Based on the feature fusion algorithm, perform coupling analysis on the corrected dynamic disturbance feature matrix and the system response matrix to generate a disturbance response matrix, and perform multi-dimensional optimization calculations on the disturbance response matrix to generate transient voltage feature matrices under different working conditions.

[0009] Preferably, perform dynamic threshold quantization on the transient voltage feature matrix to generate a dynamic parameter matrix, and construct a multi-level cascade evaluation network structure for voltage transient evaluation to generate energy storage control evaluation parameters, including: Perform segmented quantization configuration on the transient voltage feature matrix to generate quantization interval parameters, and based on the quantization interval parameters, set dynamic thresholds for the transient feature information in the transient voltage feature matrix to generate a dynamic threshold matrix. Divide the transient feature information in the transient voltage feature matrix into multiple segments according to the dynamic threshold matrix to generate a segmented feature matrix. Perform hysteresis quantization processing on the segmented feature matrix and set a dynamic holding interval to generate a quantized feature matrix. Suppress high-frequency noise of the quantized feature matrix through a hysteresis filter to generate a smoothed feature matrix, and perform normalization transformation on the smoothed feature matrix to generate a dynamic parameter matrix. Based on the dynamic parameter matrix, construct a multi-level cascade evaluation network structure, and perform real-time evaluation on the voltage transient state of the flywheel energy storage system to generate energy storage control evaluation parameters.

[0010] Preferably, based on the dynamic parameter matrix, construct a multi-level cascade evaluation network structure, and perform real-time evaluation on the voltage transient state of the flywheel energy storage system to generate energy storage control evaluation parameters, including: Divide the dynamic parameter matrix into a control feature matrix and an evaluation feature matrix. Construct a control network based on the control feature matrix. The control network includes an input layer, a hidden layer, and an output layer. The input layer contains 5 neurons corresponding to voltage, current, frequency, speed, and power signals. The hidden layer uses a piecewise linear activation function, and the output layer generates a control instruction value to obtain the output data of the control network. Construct an evaluation network based on the evaluation feature matrix. The evaluation network includes an input layer, a hidden layer, and an output layer. The input layer receives the output data of the control network. The hidden layer uses an S-shaped activation function, and the output layer generates a state evaluation value to obtain the output data of the evaluation network. Jointly train the control network and the evaluation network to generate a multi-level cascaded evaluation network structure, calculate the error function and analyze the network adaptability for the multi-level cascaded evaluation network structure, and generate a network evaluation result; Based on the network evaluation result, dynamically update the parameters of the control network and the evaluation network in the multi-level cascaded evaluation network structure, adjust the network weights through the gradient optimization algorithm, and generate energy storage control evaluation parameters.

[0011] Preferably, establishing a transient constraint function according to the energy storage control evaluation parameters and calculating the dynamic stability boundary of the flywheel energy storage system includes: Construct a symmetric positive definite matrix for the energy storage control evaluation parameters, and generate a transient state matrix through non-linear transformation; Design a dynamic integral term based on the transient state matrix, perform piecewise function operations on the state variables to generate a dynamic constraint term, and combine the transient state matrix and the dynamic constraint term to construct an energy state constraint function to generate a transient constraint function; Calculate the time-domain derivative of the transient constraint function to generate a derivative analysis result, and perform a comparison operation between the derivative analysis result and the modulus of the state vector to generate a transient constraint condition; Partition the rated operating range of the flywheel energy storage system according to the transient constraint conditions, set a dynamic stability coefficient to generate a constraint range, and perform boundary analysis based on the constraint range to generate the dynamic stability boundary of the flywheel energy storage system.

[0012] Preferably, iteratively optimizing the energy storage control parameters according to the transient voltage characteristic matrix and the dynamic stability boundary to obtain a flywheel energy storage power regulation sequence includes: Construct a control weight matrix and an error compensation matrix according to the transient voltage characteristic matrix, and substitute the control weight matrix and the error compensation matrix into the dynamic programming function to generate an optimization objective function; Apply power regulation constraint conditions and power change rate limit conditions to the optimization objective function, and convert the dynamic stability boundary into an energy derivative constraint to generate optimization constraint conditions; Input the optimization objective function and the optimization constraint conditions into an online iterative solver to calculate the regulation parameters at discrete time sequences and generate an initial sequence of regulation parameters; Set an adaptive step size for the initial sequence of regulation parameters, and perform change rate constraint processing in combination with the operating mode of the flywheel energy storage system to generate a flywheel energy storage power regulation sequence.

[0013] Preferably, the method further includes: Perform signal conversion on the flywheel energy storage power regulation sequence to generate a drive control signal, and perform gain amplification on the drive control signal to generate an energy storage execution signal; Calculate the target voltage compensation amount according to the load demand of the data center, and perform real-time monitoring on the power supply bus voltage to generate voltage sampling data; Perform deviation calculation on the voltage sampling data and the target voltage compensation amount to generate a voltage deviation value, and compare the voltage deviation value with a preset adjustment threshold to generate a deviation determination result; Dynamically correct the adjustment parameters based on the deviation determination result to generate updated adjustment parameters, and perform power sequence conversion on the updated adjustment parameters. Convert the adjustment parameters into power adjustment values through flywheel characteristic mapping to generate a new power regulation sequence; Input the new power regulation sequence into the flywheel energy storage controller for closed-loop regulation, dynamically adjust the energy storage power, and complete the transient governance of the power supply bus voltage.

[0014] Preferably, the present invention further includes a transient voltage governance control system based on a data center flywheel energy storage system, used to implement the steps of the above transient voltage governance control method for a data center flywheel energy storage system. The system includes: An acquisition unit, used to acquire three-phase voltage signals, current signals, frequency signals of the power supply bus of the data center, the rotational speed signal and the energy storage power signal of the flywheel energy storage system, and calculate to obtain a transient voltage characteristic matrix; An evaluation unit, used to perform dynamic threshold quantization on the transient voltage characteristic matrix to generate a dynamic parameter matrix, and construct a multi-level cascade evaluation network structure for voltage transient evaluation to generate energy storage control evaluation parameters; A calculation unit, used to establish a transient constraint function according to the energy storage control evaluation parameters and calculate the dynamic stability boundary of the flywheel energy storage system; An iterative optimization unit, used to perform iterative optimization of energy storage control parameters based on the transient voltage characteristic matrix and the dynamic stability boundary to obtain a flywheel energy storage power regulation sequence.

[0015] Preferably, the acquisition unit includes: A synchronous sampling module, used to perform multi-channel synchronous acquisition on three-phase voltage signals, current signals, frequency signals of the power supply bus, the rotational speed signal and the energy storage power signal of the flywheel energy storage system; A normalization processing module, used to perform interval normalization processing on the acquired transient observation data matrix to eliminate the dimension difference; A covariance analysis module, used to calculate the initial element values of the dynamic disturbance characteristic matrix based on covariance analysis; A phase compensation module for dynamically compensating the phase of the initial element value and tuning the correction coefficient; A coupling analysis module for generating a perturbation response matrix through a feature fusion algorithm and performing multi-dimensional optimization.

[0016] Preferably, the iterative optimization unit includes: A weight configuration module for constructing a control weight matrix and an error compensation matrix according to the transient voltage feature matrix; A constraint conversion module for converting the dynamic stability boundary into an energy derivative constraint; An online solution module for calculating the adjustment parameters at discrete time sequences through an online iterative solver; An adaptive processing module for setting an adaptive step size and performing a change rate constraint process in combination with the operation mode.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: At the level of data acquisition and analysis, by synchronously sampling the three-phase voltage signal, current signal, frequency signal of the power supply bus, as well as the rotational speed signal and energy storage power signal of the flywheel energy storage system, and performing a series of complex data processing, such as interval normalization, covariance analysis, phase compensation, etc., a transient voltage feature matrix is finally generated. This comprehensive and accurate data processing method can deeply mine the key information in the data and accurately reflect the actual situation of the transient voltage. Compared with the traditional simple data acquisition method, the data features obtained by the present invention are richer and more accurate, providing a solid and reliable basis for subsequent control decisions and greatly improving the perception accuracy of transient voltage fluctuations.

[0018] In the voltage transient assessment link, a multi-level cascade assessment network structure is constructed, and after operations such as dynamic threshold quantization on the transient voltage feature matrix, energy storage control assessment parameters are generated. This network structure uses different activation functions to construct a control network and an assessment network respectively. After joint training and dynamic parameter update, it can evaluate the voltage transient state of the flywheel energy storage system in real time and accurately. Compared with the traditional assessment method, it can consider various factors in the system operation more comprehensively and capture the subtle changes of the transient voltage more sensitively, thereby providing more targeted and forward-looking guidance for the control of the energy storage system and significantly improving the accuracy and timeliness of the transient voltage assessment.

[0019] Calculating the dynamic stability boundary of the flywheel energy storage system is one of the key advantages of the present invention. Through a series of complex operations on the energy storage control evaluation parameters, a transient constraint function is established, and then the dynamic stability boundary of the system is determined. This process fully considers various operating conditions and constraints of the system, providing a clear boundary for the safe and stable operation of the flywheel energy storage system. In actual operation, the system can reasonably adjust the operating parameters according to this dynamic stability boundary, avoid system instability caused by overcharging or over-discharging or other improper operations, effectively ensure the stable operation of the flywheel energy storage system, and reduce the risk of system failures.

[0020] In terms of optimizing the energy storage control parameters, iterative optimization is carried out based on the transient voltage characteristic matrix and the dynamic stability boundary to obtain the flywheel energy storage power regulation sequence. By constructing a control weight matrix and an error compensation matrix, combining with the dynamic programming function to generate an optimization objective function, and imposing various constraint conditions, the optimization result can better meet the actual operation requirements. This optimization method can quickly and accurately adjust the power output of the flywheel energy storage system according to different transient voltage conditions, and effectively control the transient voltage. When facing a voltage sag, it can quickly increase the discharge power of the energy storage system to boost the bus voltage; when the voltage surges, it can timely absorb the excess energy to stabilize the voltage. Compared with the traditional control method, it greatly improves the response speed and control accuracy of the energy storage system, effectively reduces the impact of transient voltage fluctuations on the data center equipment, ensures the normal operation of the data center equipment, and reduces the risk of data loss and equipment damage caused by voltage problems.

[0021] In addition, the present invention also sets up a closed-loop regulation mechanism. By performing signal conversion and gain amplification on the flywheel energy storage power regulation sequence, calculating the target voltage compensation amount according to the data center load demand, and real-time monitoring the power supply bus voltage, comparing the voltage deviation value with the preset regulation threshold, and dynamically correcting the regulation parameters, the dynamic adjustment of the energy storage power is realized. This closed-loop regulation mechanism enables the system to continuously optimize the control strategy according to the actual operation situation, further improve the effect of transient voltage control and the adaptive ability of the system, and ensure that the power supply of the data center is always in a stable state. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is the working principle diagram of the transient voltage control method for the data center flywheel energy storage system described in the present invention; Figure 2 is the working principle diagram of the calculation of the transient voltage characteristic matrix; Figure 3 is the working principle diagram of the construction of the multi-level cascade evaluation network structure and the generation of parameters; Figure 4 is the working principle diagram of the generation of the flywheel energy storage power regulation sequence; Figure 5It is the working principle diagram for the transient voltage governance of the power supply bus. Specific implementation manners

[0023] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0024] Please refer to Figures 1 - 5 , the present invention provides a transient voltage governance control method based on a flywheel energy storage system in a data center, and the specific implementation steps are as follows: Collect the three-phase voltage signals, current signals, frequency signals, rotational speed signals and energy storage power signals of the power supply bus of the data center. In actual operation, high-precision sensors are respectively connected to the corresponding positions of the power supply bus of the data center and the flywheel energy storage system to ensure the accuracy and real-time nature of signal collection. After collecting these signals, a transient voltage feature matrix is obtained through a specific calculation method, and this matrix comprehensively reflects various characteristics of the power supply system of the data center during the transient process.

[0025] Perform dynamic threshold quantization on the obtained transient voltage feature matrix. In this process, by reasonably setting quantization rules and threshold ranges, a dynamic parameter matrix is generated. Then, a multi-level cascade evaluation network structure is constructed. This network structure consists of multiple levels, and each level has its specific function. Use this network structure to evaluate the voltage transient state, thereby generating energy storage control evaluation parameters, and these parameters provide important bases for subsequent control decisions.

[0026] Establish a transient constraint function according to the generated energy storage control evaluation parameters. In the establishment process, fully consider the operating characteristics and limiting conditions of the flywheel energy storage system, and obtain this function through the processing and operation of relevant parameters. Based on this function, further calculate the dynamic stability boundary of the flywheel energy storage system, and clarify the stable operating range of the system under different operating conditions.

[0027] Iteratively optimize the energy storage control parameters based on the transient voltage feature matrix and the dynamic stability boundary. In the optimization process, continuously adjust the control parameters to make them meet the performance requirements and constraint conditions of the system, and finally obtain a flywheel energy storage power adjustment sequence. Through this adjustment sequence, the power output of the flywheel energy storage system can be effectively controlled to achieve the governance of the transient voltage of the power supply bus of the data center and ensure the stability and reliability of the power supply of the data center.

[0028] The following further illustrates the present invention in conjunction with Embodiments 1 to 5: Embodiment 1: In the process of collecting the three-phase voltage signals, current signals, frequency signals of the power supply bus in the data acquisition center, the rotational speed signal and energy storage power signal of the flywheel energy storage system, and calculating to obtain the transient voltage characteristic matrix, multi-channel synchronous sampling technology is adopted, and a specially designed synchronous sampling device is used to ensure that the three-phase voltage signals, current signals, frequency signals, rotational speed signal and energy storage power signal of the flywheel energy storage system can be collected at the same moment, avoiding data errors caused by asynchronous sampling times. After obtaining the transient observation data matrix, interval normalization processing is carried out. By using a specific normalization algorithm, data with different dimensions are unified into a specific interval, eliminating the influence of dimension differences on subsequent calculations and making the data comparable.

[0029] Based on covariance analysis, the initial element values of the dynamic disturbance characteristic matrix are calculated. By performing covariance analysis on the normalized data, the correlation and potential characteristics between the data are explored to obtain the initial element values of the dynamic disturbance characteristic matrix. Then, phase compensation is performed on these initial element values. According to the possible phase shift situations during the signal transmission process, a dynamic correction coefficient is introduced for tuning to accurately compensate the phase and generate the corrected dynamic disturbance characteristic matrix.

[0030] The feature fusion algorithm is used to perform coupled analysis on the corrected dynamic disturbance characteristic matrix and the system response matrix. This algorithm can fully integrate the effective information in the two matrices to generate a disturbance response matrix. Then, multi-dimensional optimization calculation is performed on the disturbance response matrix, processing and optimizing the matrix from multiple angles, and finally generating the transient voltage characteristic matrix under different working conditions, providing comprehensive and accurate data support for subsequent voltage transient evaluation and control.

[0031] In a large data center, an advanced flywheel energy storage system is equipped to ensure power supply stability. The three-phase voltage signals, current signals, frequency signals of the power supply bus in this data center, as well as the rotational speed signal and energy storage power signal of the flywheel energy storage system are respectively collected by corresponding high-precision sensors. For example, the three-phase voltage signals are collected through a voltage transformer, which has high precision and fast response characteristics and can obtain the changes in the three-phase voltage of the bus in real time; the current signals are collected using a Rogowski coil current sensor, which can accurately measure large currents with minimal impact on the original circuit; the frequency signals are extracted from the bus voltage signals with the help of a frequency measurement module. The rotational speed signal of the flywheel energy storage system is collected by an optoelectronic encoder installed on the flywheel shaft, and the energy storage power signal is calculated based on the measured current and voltage of the energy storage system through a specific algorithm. These sensors convert the collected analog signals into digital signals and transmit them to the data acquisition device, thereby generating a transient observation data matrix.

[0032] In order to eliminate the impact of the difference in signal dimensions on subsequent analysis, the collected transient observation data matrix needs to be normalized. Taking the three-phase voltage signal as an example, assuming that its normal fluctuation range is 375V-425V, the range is mapped to the [0,1] interval during normalization. For the A-phase voltage collected at a certain moment of 390V, it is converted into a value in the corresponding interval according to the normalization formula, such as (390-375) ÷ (425-375) = 0.3. Other signals are also processed similarly according to their respective fluctuation ranges, and finally a normalized data matrix is ​​obtained, so that the data of each signal is in a unified and comparable scale.

[0033] The initial element values ​​of the dynamic disturbance characteristic matrix are calculated based on covariance analysis. Taking voltage signals and current signals as examples, covariance analysis can reveal the relationship between the two in transient processes. When the load in the data center changes, if the voltage fluctuates, the current will also change accordingly. By calculating the covariance of the normalized voltage and current data, we can obtain values ​​that reflect the degree of coordinated change between the two. These values ​​are part of the initial element values ​​of the dynamic disturbance characteristic matrix. Similarly, covariance analysis is performed on the frequency signal, the speed signal of the flywheel energy storage system, the energy storage power signal and other signals to obtain more initial element values, thereby constructing the prototype of the dynamic disturbance characteristic matrix.

[0034] Since the signal may produce phase shift during transmission and acquisition, it is necessary to perform phase compensation on the initial element value of the dynamic disturbance characteristic matrix. Assuming that under a certain working condition, it is found through detection that the voltage signal has a 5° phase lag relative to the current signal, which may affect the accurate judgment of the transient characteristics of the system. At this time, the dynamic correction coefficient is introduced, and the coefficient is determined according to factors such as the signal transmission path and equipment characteristics. Through a complex tuning algorithm, the initial element value is phase adjusted so that the corrected dynamic disturbance characteristic matrix can more accurately reflect the true relationship between the signals and effectively restore the disturbance in the transient process of the system.

[0035] An advanced feature fusion algorithm is used to couple the modified dynamic disturbance feature matrix with the system response matrix. For example, when a data center is subjected to sudden disturbances such as lightning strikes, the system will generate a series of responses, and these response data constitute the system response matrix. The feature fusion algorithm combines the disturbance information in the dynamic disturbance feature matrix with the response characteristics in the system response matrix to dig out the deep relationship between the two. Afterwards, the coupled disturbance response matrix is ​​optimized and calculated from multiple dimensions such as time dimension and frequency dimension. In the time dimension, the changing trends of disturbances and responses at different times are analyzed; in the frequency dimension, the characteristic performance under different frequency components is studied. Through multi-dimensional optimization calculation, the transient voltage feature matrix under different working conditions is finally generated, providing a reliable basis for the subsequent accurate evaluation of the transient voltage condition of the data center power supply system.

[0036] Embodiment 2: In the process of dynamically threshold quantifying the transient voltage feature matrix, generating a dynamic parameter matrix, and constructing a multi-level cascade evaluation network structure for voltage transient evaluation to generate energy storage control evaluation parameters. First, perform segmented quantization configuration on the transient voltage feature matrix. According to the distribution characteristics and change rules of the data in the transient voltage feature matrix, reasonably divide the quantization interval to generate quantization interval parameters. Based on these quantization interval parameters, set dynamic thresholds for the transient feature information in the transient voltage feature matrix to form a dynamic threshold matrix.

[0037] Perform multi-segment division processing on the transient feature information in the transient voltage feature matrix according to the dynamic threshold matrix. According to the set thresholds, divide the transient feature information into different paragraphs, and each paragraph represents a different feature state, thereby generating a segmented feature matrix. Perform hysteresis quantization processing on the segmented feature matrix. By setting the hysteresis characteristics, avoid the influence of frequent fluctuations of data near the threshold on the quantization result, and set a dynamic holding interval to make the quantization result more stable and reliable, generating a quantized feature matrix.

[0038] Suppress high-frequency noise of the quantized feature matrix through a hysteresis filter. The hysteresis filter can effectively remove high-frequency noise components in the quantized feature matrix, ensure the accuracy and stability of the data, and generate a smoothed feature matrix. Then perform standardization conversion on the smoothed feature matrix to make it conform to a specific standard format, generating a dynamic parameter matrix.

[0039] Based on the dynamic parameter matrix, construct a multi-level cascade evaluation network structure. Divide the dynamic parameter matrix into a control feature matrix and an evaluation feature matrix, which are used to construct a control network and an evaluation network respectively. The control network includes an input layer, a hidden layer, and an output layer. The input layer contains 5 neurons, corresponding to voltage, current, frequency, rotational speed, and power signals respectively. The hidden layer uses a piecewise linear activation function, which can perform flexible linear transformations according to different ranges of input data. The output layer generates a control instruction value to obtain the output data of the control network. The evaluation network also includes an input layer, a hidden layer, and an output layer. The input layer receives the output data of the control network. The hidden layer uses an S-shaped activation function. The output layer generates a state evaluation value to obtain the output data of the evaluation network. Jointly train the control network and the evaluation network. By continuously adjusting the parameters and weights of the network, make the network accurately evaluate the voltage transient state, generating a multi-level cascade evaluation network structure. Calculate the error function and analyze the network adaptability of the multi-level cascade evaluation network structure. According to the calculation results, dynamically update the parameters of the control network and the evaluation network in the network, and adjust the network weights through the gradient optimization algorithm to finally generate energy storage control evaluation parameters.

[0040] Taking a medium-sized data center as an example, the data center has a flywheel energy storage system for power supply, and the transient voltage characteristic matrix of the data center has been obtained through previous acquisition and calculation. This matrix contains a large amount of transient characteristic information related to the three-phase voltage signal, current signal, frequency signal of the power supply bus, and the rotational speed signal and energy storage power signal of the flywheel energy storage system.

[0041] Perform segmented quantization configuration on the transient voltage characteristic matrix. For example, according to previous research and experience on the data center power supply system, divide the transient characteristics of the voltage signal into different intervals according to the voltage fluctuation range. Assume that the bus voltage is between 380V and 400V during normal operation. When the voltage fluctuation exceeds this range, divide it into intervals with every 5V. Below 375V - 380V is an interval, 380V - 385V is an interval, and so on. Based on these divisions, generate quantization interval parameters. Then, based on these quantization interval parameters, set dynamic thresholds for the transient characteristic information in the transient voltage characteristic matrix. For example, for the case of voltage drop, when the voltage is in the range of 375V - 380V, set a relatively low threshold to indicate that the voltage transient characteristic is in a mild abnormal state at this time; when the voltage is below 375V, set a higher threshold to indicate that the voltage transient characteristic is in a severe abnormal state, thereby generating a dynamic threshold matrix.

[0042] Perform multi-segment division processing on the transient characteristic information in the transient voltage characteristic matrix according to the dynamic threshold matrix. Taking the transient characteristic of the voltage signal at a certain moment as an example, if the voltage is 378V at this time, according to the dynamic threshold matrix, it is divided into the interval of 375V - 380V, then the related transient characteristic information is classified into the corresponding paragraph, and so on, divide all the transient characteristic information in the matrix to generate a segmented characteristic matrix.

[0043] Perform hysteresis quantization processing on the segmented characteristic matrix and set a dynamic holding interval. Assume that the voltage signal fluctuates around 378V within a certain period of time. To avoid frequent changes in the quantization result due to small fluctuations, set the hysteresis width to 2V and the dynamic holding interval to ±1V. When the voltage rises from 378V to 380V, since it does not exceed the hysteresis upper limit (380 + 1 = 381V), the quantization result remains unchanged; only when the voltage rises above 381V, the quantization result will change. Similarly, when the voltage drops, only when it is below 377V (378 - 1 = 377V), the quantization result will be adjusted accordingly. Generate a quantization characteristic matrix in this way to make the quantization result more stable and reliable.

[0044] Suppress high-frequency noise in the quantized feature matrix through a hysteresis filter. During the operation of the data center, electromagnetic interference and other factors generated by electrical equipment will cause the collected signals to contain high-frequency noise. The hysteresis filter can effectively identify and remove these high-frequency noise components. For example, there are some noise signals with frequencies higher than 10 kHz in the quantized feature matrix. The hysteresis filter can filter out these high-frequency noises according to its set frequency characteristics, thereby generating a smoothed feature matrix. After that, perform a normalization transformation on the smoothed feature matrix. Assuming that the element value range in the smoothed feature matrix is 0 - 100, through the normalization transformation formula, it is transformed into the range [-1, 1] to make the data conform to a specific standard format and generate a dynamic parameter matrix.

[0045] Based on the dynamic parameter matrix, construct a multi-level cascaded evaluation network structure. Divide the dynamic parameter matrix into a control feature matrix and an evaluation feature matrix. For example, select some data directly related to voltage, current, frequency, rotational speed, and power signals as the control feature matrix, and the remaining data used to reflect the overall state and trend of the system as the evaluation feature matrix.

[0046] Construct a control network based on the control feature matrix. The input layer of the control network contains 5 neurons, corresponding to voltage, current, frequency, rotational speed, and power signals respectively. When the bus voltage of the data center power supply is 390 V, the current is 50 A, the frequency is 50.1 Hz, the rotational speed of the flywheel energy storage system is 5000 r / min, and the energy storage power is 50 kW at a certain moment, these data are transmitted as input signals to the neurons in the input layer. The hidden layer uses a piecewise linear activation function. Assuming the breakpoints are set at 0 and 1, when the input value is less than 0, the function output is twice the input value; when the input value is between 0 and 1, the function output is equal to the input value; when the input value is greater than 1, the function output is 0.5 times the input value. After being processed by the hidden layer, the output layer generates a control instruction value, such as generating a control instruction value for adjusting the charge and discharge power of the flywheel energy storage system, to obtain the output data of the control network.

[0047] Construct an evaluation network based on the evaluation feature matrix. The input layer of the evaluation network receives the output data of the control network, and the hidden layer uses an S-shaped activation function. The S-shaped activation function has the characteristic of mapping the input value to the range between 0 and 1. It can further process the output data of the control network and extract more representative features. For example, when the output data of the control network is 0.8, after being processed by the S-shaped activation function, a value that can better reflect the current state of the system is output. The output layer generates a state evaluation value according to the output result of the hidden layer, such as evaluating the stability of the transient voltage state of the current data center power supply system, to obtain the output data of the evaluation network.

[0048] The control network and the evaluation network are jointly trained. During the training process, the parameters and weights of the network are continuously adjusted. For example, set the initial weight values, and then calculate the error between the network output result and the actual situation based on a large amount of historical data and actual operation data. Through the error backpropagation algorithm, gradually adjust the connection weights between neurons in each layer of the control network and the evaluation network, so that the network output is closer to the actual situation. After multiple iterative trainings, a multi-level cascaded evaluation network structure is generated. Then, perform error function calculation and network adaptability analysis on the multi-level cascaded evaluation network structure. Calculate the error function value after each training to evaluate the accuracy of the network; at the same time, analyze the adaptability of the network under different working conditions, such as the response of the network when the load of the data center suddenly increases or decreases significantly. According to the calculation and analysis results, perform dynamic parameter updates on the control network and the evaluation network in the multi-level cascaded evaluation network structure, adjust the network weights through the gradient optimization algorithm, and finally generate energy storage control evaluation parameters to provide a decision-making basis for subsequent control of the flywheel energy storage system.

[0049] Embodiment 3: Regarding establishing a transient constraint function based on the energy storage control evaluation parameters and calculating the dynamic stability boundary of the flywheel energy storage system. Construct a symmetric positive definite matrix for the energy storage control evaluation parameters. Through a specific matrix construction method, transform the energy storage control evaluation parameters into the form of a symmetric positive definite matrix to facilitate subsequent calculations and analyses. Generate a transient state matrix through a non-linear transformation, and this transformation method can more accurately reflect the state characteristics of the system during the transient process.

[0050] Design a dynamic integral term based on the transient state matrix, consider the change of the system state over time, and perform piecewise function operations on the state variables. According to the characteristics of the system in different operating stages, use different function forms for operations to generate dynamic constraint terms. Combine the transient state matrix and the dynamic constraint terms to construct an energy state constraint function to form a transient constraint function, which comprehensively considers the energy state and constraint conditions of the system.

[0051] Calculate the time-domain derivative of the transient constraint function. By taking the derivative of the function, analyze the change trend of the function over time to generate a derivative analysis result. Perform a comparison operation between the derivative analysis result and the modulus value of the state vector, and generate transient constraint conditions according to the comparison result. Partition the rated operating range of the flywheel energy storage system according to the transient constraint conditions, divide the rated operating range into different regions, and each region corresponds to a different operating state. Set a dynamic stability coefficient, reasonably determine the value of the dynamic stability coefficient according to the actual requirements and operating characteristics of the system, and generate a constraint range. Perform boundary analysis based on the constraint range. Through in-depth research and calculation of the boundaries of the constraint range, generate the dynamic stability boundary of the flywheel energy storage system, providing an important reference basis for the stable operation of the system.

[0052] Suppose in a large data center, the capacity of the flywheel energy storage system equipped is 10 MW / 20 MWh. When the data center encounters a sudden power failure, such as a short - circuit trip of a certain power supply line, it will cause a severe voltage fluctuation in the power supply bus. At this time, the method of this embodiment is required to determine the dynamic stability boundary of the flywheel energy storage system and ensure the stable operation of the system.

[0053] When establishing a transient constraint function based on energy storage control evaluation parameters and calculating the dynamic stability boundary of the flywheel energy storage system, first, a symmetric positive - definite matrix needs to be constructed for the energy storage control evaluation parameters. For example, within a period of time after a fault occurs, a series of energy storage control evaluation parameters are collected. These parameters may include the rate of change of the flywheel's rotational speed, the change in the charge - discharge power of the energy storage system, etc. Arrange these parameters according to specific rules to form a matrix . To construct a symmetric positive - definite matrix, the method of (in the formula, is the original matrix composed of energy storage control evaluation parameters, is the transpose matrix of matrix , and is the constructed symmetric positive - definite matrix) can be used to obtain the symmetric positive - definite matrix . Through non - linear transformation, using a suitable non - linear function (such as logarithmic function, exponential function, etc.), the symmetric positive - definite matrix is transformed into a transient state matrix . This transformation can more effectively reflect the actual state of the system during the transient process.

[0054] Based on the transient state matrix , a dynamic integral term is designed. Suppose an element in the transient state matrix represents the rotational speed of the flywheel. Taking time as a variable, integrate the rotational speed of the flywheel to obtain the angular change of the flywheel rotation within a period of time. At the same time, perform a piece - wise function operation on the state variables. For example, according to the characteristics of the flywheel energy storage system, when the rotational speed of the flywheel is higher than 90% of its rated rotational speed, one function form is used to calculate the dynamic constraint term; when the rotational speed is lower than 90% of the rated rotational speed, another function form is used. Combine the transient state matrix and the dynamic constraint term obtained through the piece - wise function operation to construct an energy state constraint function , thereby generating a transient constraint function.

[0055] Calculate the time - domain derivative of the transient constraint function . Through mathematical derivation, the derivative analysis result is obtained. Suppose the state vector is , and its modulus value is . The derivative analysis result Compare with the magnitude of the state vector Perform a comparison operation. For example, when occurs, set one condition; when occurs, set another condition, and thus generate transient constraint conditions.

[0056] Partition the rated operating range of the flywheel energy storage system according to the transient constraint conditions. For example, divide the rated speed range of the flywheel [8000 r / min, 12000 r / min] into three regions: the speed range of 8000 r / min - 9500 r / min is the low-speed region, 9500 r / min - 10500 r / min is the normal-speed region, and 10500 r / min - 12000 r / min is the high-speed region. Set the dynamic stability coefficient , and according to the actual situation and safety requirements of the system, determine to be 0.9. Combine the transient constraint conditions and the dynamic stability coefficient to generate a constraint range. Finally, perform boundary analysis based on the constraint range. By studying the boundaries of different partitions, determine the dynamic stability boundaries of the flywheel energy storage system under different operating conditions. For example, in the low-speed region, when the speed approaches 8000 r / min, determine the maximum charge-discharge power boundary of the energy storage system at this time, so as to ensure the stable operation of the flywheel energy storage system during the transient process and provide stable and reliable power support for the data center.

[0057] Example 4: Iteratively optimize the energy storage control parameters based on the transient voltage characteristic matrix and the dynamic stability boundary to obtain the implementation process of the flywheel energy storage power regulation sequence. Construct a control weight matrix and an error compensation matrix according to the transient voltage characteristic matrix. By analyzing and processing the data in the transient voltage characteristic matrix, determine the element values of the control weight matrix and the error compensation matrix, and these two matrices can reflect the control requirements and error conditions of the system. Substitute the control weight matrix and the error compensation matrix into the dynamic programming function, and use the method of dynamic programming to generate an optimization objective function, which can comprehensively consider the performance indicators and control requirements of the system.

[0058] Apply power regulation constraint conditions and power change rate limit conditions to the optimization objective function. According to the actual performance and safe operation requirements of the flywheel energy storage system, set reasonable power regulation ranges and power change rate limits to ensure that the system will not exceed the safe range during the regulation process. Convert the dynamic stability boundary into an energy derivative constraint to enable the optimization process to fully consider the dynamic stability characteristics of the system and generate optimization constraint conditions.

[0059] Input the optimization objective function and optimization constraints into the online iterative solver. The online iterative solver adopts an efficient algorithm, which can continuously perform iterative calculations during real-time operation, solve the adjustment parameters under discrete time series, and generate an initial sequence of adjustment parameters. Set an adaptive step size for the initial sequence of adjustment parameters. According to the operating state and convergence of the system, automatically adjust the step size to improve the efficiency and accuracy of iterative calculations. Combine the operating mode of the flywheel energy storage system to handle the rate-of-change constraint. Consider the characteristics of the flywheel energy storage system under different operating modes, constrain the rate of change of the adjustment parameters, and finally generate a flywheel energy storage power adjustment sequence to achieve precise control of the power of the flywheel energy storage system.

[0060] Suppose in a large-scale data center equipped with a flywheel energy storage system to address voltage fluctuations during power supply. When a large server cluster in the data center suddenly starts, it will cause a transient drop in the voltage of the power supply bus. At this time, the method in this embodiment is required to control the flywheel energy storage system to stabilize the voltage.

[0061] First, construct a control weight matrix and an error compensation matrix based on the transient voltage characteristic matrix. The transient voltage characteristic matrix is calculated by collecting the three-phase voltage signals, current signals, frequency signals of the power supply bus of the data center, as well as the rotational speed signal and energy storage power signal of the flywheel energy storage system in the early stage. For example, within 1 second after the server cluster starts, a series of relevant signal data are collected and processed to obtain the transient voltage characteristic matrix. The control weight matrix is constructed by analyzing the relationship between these data and the desired stable voltage state. Suppose it is found that the change in the voltage signal has a greater impact on system stability, then in the control weight matrix, the element weight corresponding to the voltage signal will be relatively high. The error compensation matrix is constructed based on the deviation between the current transient voltage characteristic and the ideal state. If the current voltage is lower than the ideal voltage value, the error compensation matrix will set the element values accordingly for subsequent adjustment of the control strategy to compensate for the voltage deviation.

[0062] Substitute the control weight matrix and the error compensation matrix into the dynamic programming function to generate the optimization objective function. The dynamic programming function comprehensively considers various factors of the system, such as voltage stability and the charge-discharge efficiency of the energy storage system. Taking voltage stability as an example, the optimization objective function hopes to adjust the voltage to the stable range as soon as possible while meeting other conditions. In this process, a performance index is set, such as minimizing the sum of the squares of the voltage deviations, to construct the optimization objective function.

[0063] Power regulation constraints and power change rate limit conditions are imposed on the optimization objective function. Since the power regulation ability of the flywheel energy storage system is limited, there is a certain power regulation range. Suppose the maximum charging power of this flywheel energy storage system is -5 MW (the negative sign indicates charging), and the maximum discharging power is 5 MW. This is the power regulation constraint condition. At the same time, in order to protect the energy storage system and ensure the stable operation of the system, there are also restrictions on the power change rate. For example, it is stipulated that the power change rate cannot exceed 1 MW / s, which limits the change speed of the charging and discharging power of the flywheel energy storage system. In addition, the dynamic stability boundary is converted into an energy derivative constraint. The dynamic stability boundary is obtained through previous calculations, and it reflects the stable operation range of the system under different states. Converting it into an energy derivative constraint ensures that during the optimization process, the energy change of the system is always within the stable range, generating optimization constraint conditions.

[0064] The optimization objective function and optimization constraint conditions are input into the online iterative solver. The online iterative solver adopts an efficient algorithm to continuously iterate and calculate the adjustment parameters at discrete time series. For example, taking 0.1 second as a time interval, the corresponding adjustment parameters are calculated within each time interval. After multiple iterative calculations, an initial sequence of adjustment parameters is generated. During the iteration process, the solver will continuously adjust the adjustment parameters according to the optimization objective function and constraint conditions to make them gradually approach the optimal value.

[0065] An adaptive step size is set for the initial sequence of adjustment parameters, and the change rate constraint is processed in combination with the operation mode of the flywheel energy storage system. At the initial stage of the system voltage drop, in order to increase the voltage as soon as possible, the adaptive step size will be relatively large, enabling the adjustment parameters to be adjusted quickly. As the voltage gradually approaches the stable value, the step size will gradually decrease to avoid over-adjustment. At the same time, according to the current operation mode of the flywheel energy storage system, such as the charging mode, discharging mode, or standby mode, the change rate of the adjustment parameters is constrained. For example, in the discharging mode, since too fast discharging speed may affect the life and safety of the energy storage system, the change rate of the adjustment parameters will be more strictly restricted. After these processes, a flywheel energy storage power regulation sequence is finally generated. Through this sequence, the charging and discharging power of the flywheel energy storage system is precisely controlled to stabilize the voltage of the power supply bus of the data center and ensure the normal operation of the data center equipment.

[0066] Embodiment 5: In the complete process of the entire transient voltage governance control method, in addition to the above key steps, the following links are also included. The signal conversion is performed on the flywheel energy storage power regulation sequence to convert the power regulation sequence into a signal form suitable for drive control, generating a drive control signal. The drive control signal is amplified in gain. The amplitude of the signal is amplified through an amplifier so that it can meet the actual control requirements, generating an energy storage execution signal.

[0067] Calculate the target voltage compensation amount according to the load demand of the data center. Through real-time monitoring and analysis of the data center load, combined with the voltage condition of the power supply bus, calculate the voltage amount that needs to be compensated to ensure the stability of the power supply voltage. Monitor the power supply bus voltage in real time, use a voltage sensor to continuously measure the power supply bus voltage, and generate voltage sampling data.

[0068] Calculate the deviation between the voltage sampling data and the target voltage compensation amount. By calculating the difference between the two, generate a voltage deviation value. Compare the voltage deviation value with a preset adjustment threshold, and generate a deviation determination result based on the comparison result. Dynamically correct the adjustment parameters based on the deviation determination result. If the deviation exceeds the preset threshold, adjust the adjustment parameters according to the magnitude and direction of the deviation to generate updated adjustment parameters. Perform power sequence conversion on the updated adjustment parameters. Through flywheel characteristic mapping, convert the adjustment parameters into power adjustment values to generate a new power adjustment sequence. Input the new power adjustment sequence into the flywheel energy storage controller for closed-loop adjustment. The flywheel energy storage controller dynamically adjusts the energy storage power according to the input power adjustment sequence, and continuously cycle this process until the power supply bus voltage reaches a stable state, completing the transient governance of the power supply bus voltage and ensuring the stability and reliability of the data center power supply.

[0069] In a data center with a large number of precision electronic devices, the daily operation has extremely high requirements for power supply stability. This data center is equipped with a flywheel energy storage system to cope with transient voltage fluctuations.

[0070] The flywheel energy storage system obtains the flywheel energy storage power adjustment sequence based on the previous calculation. This sequence contains a series of power adjustment instructions, which are generated based on the real-time condition of the data center power supply system. For example, when a large number of computing tasks suddenly increase in a certain area of the data center, causing the power consumption load in that area to rise instantaneously, the power supply bus voltage will experience a transient drop. At this time, the previously calculated flywheel energy storage power adjustment sequence begins to play a role. The power adjustment instructions in the sequence will first perform signal conversion, converting the instructions originally represented by power values into a signal form suitable for driving control, such as converting the power value into a pulse signal with a specific frequency and duty cycle to generate a drive control signal. The amplitude of this drive control signal is usually small and cannot directly drive the power adjustment device of the flywheel energy storage system. Therefore, it is necessary to amplify its gain. Through a dedicated signal amplifier, amplify the amplitude of the drive control signal to a level sufficient to drive the actuator to generate an energy storage execution signal.

[0071] The managers of the data center calculate the target voltage compensation amount according to the rated voltage requirements of various devices in the data center and the actual operating conditions. Suppose the rated operating voltage of the devices in the data center is 400V, and after the sudden change of the electrical load, the supply bus voltage drops to 380V. By analyzing the voltage-power characteristic curve of the devices and the overall power supply architecture of the data center, it is calculated that the voltage needs to be increased by 20V at this time to ensure the normal operation of the devices, and this 20V is the target voltage compensation amount. At the same time, the data center installs a high-precision voltage monitoring device on the supply bus to monitor the supply bus voltage in real time, and collects voltage data every certain period (such as 1 millisecond) to generate voltage sampling data.

[0072] Calculate the deviation between the voltage sampling data and the target voltage compensation amount. For example, at a certain moment, the voltage sampling data shows that the supply bus voltage is 385V, and the target voltage compensation amount is to increase the voltage to 400V, then the voltage deviation value is 400 - 385 = 15V. Compare this voltage deviation value with the preset adjustment threshold. Suppose the preset adjustment threshold is 5V. Since 15V is greater than 5V, the deviation determination result indicates that the current voltage deviation exceeds the allowable range, and the adjustment parameters need to be adjusted.

[0073] Dynamically correct the adjustment parameters based on the deviation determination result. According to the preset correction rules, adjust the adjustment parameters in combination with the magnitude and direction of the voltage deviation. For example, if the voltage deviation is positive (the actual voltage is lower than the target voltage) and the deviation is large, appropriately increase the discharge power adjustment parameter of the flywheel energy storage system; if the deviation is small, adjust the parameter slightly. After correction, generate the updated adjustment parameters. Then perform a power sequence conversion on the updated adjustment parameters, and convert the adjustment parameters into power adjustment values through the flywheel characteristic mapping. The flywheel characteristic mapping is a mathematical model established based on the physical characteristics of the flywheel energy storage system, which can accurately convert the adjustment parameters into actual power adjustment values and generate a new power adjustment sequence.

[0074] Finally, input the new power adjustment sequence into the flywheel energy storage controller for closed-loop regulation. After receiving the new power adjustment sequence, the flywheel energy storage controller will adjust the charge and discharge power of the flywheel energy storage system in real time according to these instructions. For example, if the new power adjustment sequence requires an increase in the discharge power, the controller will control the relevant circuit to make the flywheel release more energy and increase the supply bus voltage. During the regulation process, continuously repeat the above steps of collecting voltage data, calculating deviation, correcting parameters, and adjusting power to form a closed-loop control system. Continuously perform dynamic adjustment on the energy storage power until the supply bus voltage is stabilized within the target range, complete the transient governance of the supply bus voltage, ensure that the precision electronic devices in the data center can operate normally, and avoid problems such as device damage or data loss caused by voltage fluctuations.

[0075] It should be noted that, in this document, relational terms such as first and second are only used 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 "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0076] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A transient voltage governance control method based on a flywheel energy storage system in a data center, characterized in that It includes the following steps: Collect the three-phase voltage signal, current signal, frequency signal of the power supply bus of the data center, the rotational speed signal and energy storage power signal of the flywheel energy storage system, and calculate to obtain a transient voltage characteristic matrix; Perform dynamic threshold quantization on the transient voltage characteristic matrix, generate a dynamic parameter matrix, and construct a multi-level cascade evaluation network structure for voltage transient evaluation to generate energy storage control evaluation parameters; Establish a transient constraint function according to the energy storage control evaluation parameters, and calculate to obtain the dynamic stability boundary of the flywheel energy storage system; Iteratively optimize the energy storage control parameters based on the transient voltage characteristic matrix and the dynamic stability boundary to obtain the flywheel energy storage power regulation sequence.

2. The transient voltage governance control method based on the flywheel energy storage system of the data center according to claim 1, wherein, The step of collecting the three-phase voltage signal, current signal, frequency signal of the power supply bus of the data center, the rotational speed signal and energy storage power signal of the flywheel energy storage system, and calculating to obtain a transient voltage characteristic matrix includes: Synchronously sample the three-phase voltage signal, current signal, frequency signal of the power supply bus, the rotational speed signal and energy storage power signal of the flywheel energy storage system to generate a transient observation data matrix; Perform interval normalization processing on the transient observation data matrix to obtain a normalized data matrix, and calculate based on covariance analysis on the normalized data matrix to generate the initial element values of the dynamic disturbance characteristic matrix; Perform phase compensation on the initial element values of the dynamic disturbance characteristic matrix, and tune by introducing a dynamic correction coefficient to generate a corrected dynamic disturbance characteristic matrix; Based on the feature fusion algorithm, perform coupling analysis on the corrected dynamic disturbance characteristic matrix and the system response matrix to generate a disturbance response matrix, and perform multi-dimensional optimization calculation on the disturbance response matrix to generate transient voltage characteristic matrices under different working conditions.

3. The transient voltage governance control method for a flywheel energy storage system based on a data center according to claim 1, wherein The step of performing dynamic threshold quantization on the transient voltage characteristic matrix, generating a dynamic parameter matrix, and constructing a multi-level cascade evaluation network structure for voltage transient evaluation to generate energy storage control evaluation parameters includes: Perform segmented quantization configuration on the transient voltage characteristic matrix to generate quantization interval parameters, and set dynamic thresholds for the transient characteristic information in the transient voltage characteristic matrix based on the quantization interval parameters to generate a dynamic threshold matrix; Perform multi-segment division processing on the transient characteristic information in the transient voltage characteristic matrix according to the dynamic threshold matrix to generate a segmented characteristic matrix; Perform hysteresis quantization processing on the segmented characteristic matrix and set a dynamic holding interval to generate a quantized characteristic matrix; Suppress high-frequency noise of the quantized characteristic matrix through a hysteresis filter to generate a smoothed characteristic matrix, and perform standardization conversion on the smoothed characteristic matrix to generate a dynamic parameter matrix; Based on the dynamic parameter matrix, construct a multi-level cascade evaluation network structure, and perform real-time evaluation on the voltage transient state of the flywheel energy storage system to generate energy storage control evaluation parameters.

4. The transient voltage governance control method for a flywheel energy storage system based on a data center according to claim 3, characterized in that The step of constructing a multi-level cascade evaluation network structure based on the dynamic parameter matrix, and performing real-time evaluation on the voltage transient state of the flywheel energy storage system to generate energy storage control evaluation parameters includes: Divide the dynamic parameter matrix into a control characteristic matrix and an evaluation characteristic matrix; Construct a control network based on the control feature matrix. The control network includes an input layer, a hidden layer, and an output layer. The input layer contains 5 neurons corresponding to voltage, current, frequency, rotational speed, and power signals. The hidden layer uses a piecewise linear activation function, and the output layer generates a control instruction value to obtain the control network output data; Construct an evaluation network based on the evaluation feature matrix. The evaluation network includes an input layer, a hidden layer, and an output layer. The input layer receives the control network output data. The hidden layer uses an S-shaped activation function, and the output layer generates a state evaluation value to obtain the evaluation network output data; Jointly train the control network and the evaluation network to generate a multi-level cascaded evaluation network structure, and perform error function calculation and network adaptability analysis on the multi-level cascaded evaluation network structure to generate a network evaluation result; Based on the network evaluation result, perform dynamic parameter update on the control network and the evaluation network in the multi-level cascaded evaluation network structure, and adjust the network weights through the gradient optimization algorithm to generate energy storage control evaluation parameters.

5. The transient voltage governance control method based on the flywheel energy storage system of the data center according to claim 1, characterized in that, Establish a transient constraint function according to the energy storage control evaluation parameters and calculate the dynamic stability boundary of the flywheel energy storage system, including: Construct a symmetric positive definite matrix for the energy storage control evaluation parameters and generate a transient state matrix through non-linear transformation; Design a dynamic integral term based on the transient state matrix, perform piecewise function operations on the state variables to generate a dynamic constraint term, and combine the transient state matrix and the dynamic constraint term to construct an energy state constraint function to generate a transient constraint function; Calculate the time-domain derivative of the transient constraint function to generate a derivative analysis result, and perform a comparison operation between the derivative analysis result and the modulus value of the state vector to generate a transient constraint condition; Partition the rated operating interval of the flywheel energy storage system according to the transient constraint condition, set a dynamic stability coefficient to generate a constraint range, and perform boundary analysis based on the constraint range to generate the dynamic stability boundary of the flywheel energy storage system.

6. The transient voltage governance control method for a flywheel energy storage system based on a data center according to claim 1, characterized in that Iteratively optimize the energy storage control parameters based on the transient voltage feature matrix and the dynamic stability boundary to obtain a flywheel energy storage power regulation sequence, including: Construct a control weight matrix and an error compensation matrix according to the transient voltage feature matrix, and substitute the control weight matrix and the error compensation matrix into the dynamic programming function to generate an optimization objective function; Apply power regulation constraint conditions and power change rate limit conditions to the optimization objective function, and convert the dynamic stability boundary into an energy derivative constraint to generate optimization constraint conditions; Input the optimization objective function and the optimization constraint conditions into an online iterative solver to calculate the regulation parameters at discrete time sequences to generate an initial sequence of regulation parameters; Set an adaptive step size for the initial sequence of regulation parameters, and perform change rate constraint processing in combination with the operating mode of the flywheel energy storage system to generate a flywheel energy storage power regulation sequence.

7. The transient voltage governance control method based on the flywheel energy storage system of the data center according to claim 1, characterized in that, The method further includes: Perform signal conversion on the flywheel energy storage power regulation sequence to generate a drive control signal, and perform gain amplification on the drive control signal to generate an energy storage execution signal; Calculate the target voltage compensation amount according to the load demand of the data center, and monitor the voltage of the power supply bus in real time to generate voltage sampling data; Calculate the deviation between the voltage sampling data and the target voltage compensation amount to generate a voltage deviation value, and compare the voltage deviation value with a preset adjustment threshold to generate a deviation determination result; Dynamically correct the adjustment parameters based on the deviation determination result to generate updated adjustment parameters, and perform power sequence conversion on the updated adjustment parameters. Convert the adjustment parameters into power adjustment values through flywheel characteristic mapping to generate a new power adjustment sequence; Input the new power adjustment sequence into the flywheel energy storage controller for closed-loop adjustment to dynamically adjust the energy storage power and complete the transient governance of the power supply bus voltage.

8. A transient voltage governance control system based on a flywheel energy storage system in a data center, characterized in that, For implementing the steps of the method according to any one of claims 1 to 7, the system includes: An acquisition unit for acquiring three-phase voltage signals, current signals, frequency signals, rotational speed signals and energy storage power signals of the power supply bus of the data center, and calculating to obtain a transient voltage characteristic matrix; An evaluation unit for dynamically threshold quantifying the transient voltage characteristic matrix to generate a dynamic parameter matrix, and constructing a multi-level cascade evaluation network structure for voltage transient evaluation to generate energy storage control evaluation parameters; A calculation unit for establishing a transient constraint function according to the energy storage control evaluation parameters and calculating the dynamic stability boundary of the flywheel energy storage system; An iterative optimization unit for iteratively optimizing the energy storage control parameters based on the transient voltage characteristic matrix and the dynamic stability boundary to obtain a flywheel energy storage power adjustment sequence.

9. The transient voltage governance control system based on the flywheel energy storage system of the data center according to claim 8, characterized in that, The acquisition unit includes: A synchronous sampling module for multi-channel synchronous acquisition of three-phase voltage signals, current signals, frequency signals, rotational speed signals and energy storage power signals of the power supply bus; A normalization processing module for performing interval normalization processing on the acquired transient observation data matrix to eliminate the dimension difference; A covariance analysis module for calculating the initial element values of the dynamic disturbance characteristic matrix based on covariance analysis; A phase compensation module for dynamically compensating the phase and tuning the correction coefficient of the initial element values; A coupling analysis module for generating a disturbance response matrix through a feature fusion algorithm and performing multi-dimensional optimization.

10. The transient voltage governance control system based on the flywheel energy storage system of the data center according to claim 8, wherein The iterative optimization unit includes: A weight configuration module for constructing a control weight matrix and an error compensation matrix according to the transient voltage characteristic matrix; A constraint conversion module for converting the dynamic stability boundary into an energy derivative constraint; An online solution module for calculating the adjustment parameters at discrete time sequences through an online iterative solver; An adaptive processing module for setting an adaptive step size and performing change rate constraint processing in combination with the operating mode.

Images