A state prediction method for a centralized flywheel energy storage system

Abstract

The application discloses a state prediction method for a centralized flywheel energy storage system, and belongs to the technical field of power system energy storage, which comprises the following steps: acquiring real-time operation parameters of each flywheel unit in the centralized flywheel energy storage system at a current time; synchronously inputting the real-time operation parameters into a short-time high-frequency processing channel and a long-time trend processing channel for processing in different domains to generate real-time abnormal probability values and cumulative degradation indexes; generating comprehensive prediction state indexes based on the real-time abnormal probability values and the cumulative degradation indexes; and dynamically adjusting an operation control strategy of the energy storage system according to the comprehensive prediction state indexes. The application adopts a double-channel domain analysis technology of short-time high-frequency processing and long-time trend processing, combines with dynamic correction of abnormal evaluation parameters of an electrical mechanical coupling model, can realize minute-level accurate prediction of the flywheel health state, prolongs the service life of the equipment and guarantees the frequency modulation stability of the power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of power system energy storage technology, and in particular to a state prediction method for a centralized flywheel energy storage system. Background Art

[0002] Currently, centralized flywheel energy storage systems face the challenge of coupling mechanical condition monitoring with grid dynamic response in grid frequency regulation applications. Frequent charging and discharging operations lead to increased wear on flywheel bearings, while the accumulation of mechanical stress caused by grid frequency fluctuations can easily lead to sudden failures, compromising the reliability of the energy storage system.

[0003] Traditional solutions independently monitor mechanical vibration signals and grid frequency regulation performance, using fixed thresholds to identify equipment anomalies. For example, static alarm values ​​are set for bearing temperature, or maintenance decisions are made based on the number of frequency regulation delays. These approaches rely on single-parameter early warnings and fail to incorporate the cross-influence characteristics of electrical machinery.

[0004] However, traditional methods have serious drawbacks. Short-term sudden failures and long-term performance degradation cannot be evaluated collaboratively, resulting in delayed maintenance strategies. The correlation path between electrical load mutations and mechanical damage is difficult to quantify, and the root cause analysis of faults is insufficient. Single sensor signals are easily affected by noise and may falsely trigger protection actions, affecting the continuity of grid frequency regulation. Summary of the Invention

[0005] To solve the above problems, the present invention provides a state prediction method for a centralized flywheel energy storage system. It adopts a dual-channel domain analysis technology of short-term high-frequency processing and long-term trend processing, combined with an electrical-mechanical coupling model to dynamically correct abnormal assessment parameters. It can achieve minute-level accurate prediction of the flywheel health status, extend the equipment life and ensure the stability of grid frequency regulation.

[0006] The above objectives can be achieved through the following solutions:

[0007] A state prediction method for a centralized flywheel energy storage system comprises obtaining the real-time operating parameters of each flywheel unit in the centralized flywheel energy storage system at the current moment, including speed, temperature, adjacent flywheel vibration energy spectrum and grid frequency modulation signal parameters; synchronously inputting the real-time operating parameters into a short-time high-frequency processing channel and a long-time trend processing channel for domain processing; wherein the short-time high-frequency processing channel dynamically updates the real-time operating parameters in seconds based on a preset time window to generate a real-time abnormality probability value; the long-time trend processing channel updates the bearing friction coefficient and efficiency change on a monthly basis according to a historical degradation database to generate a cumulative degradation index; based on the real-time abnormality probability value and the cumulative degradation index, a comprehensive prediction state index is generated; and the operation control strategy of the energy storage system is dynamically adjusted according to the comprehensive prediction state index.

[0008] Optionally, the short-time high-frequency processing channel performs second-level dynamic update on the real-time operation parameter based on a preset time window, and generating the real-time abnormal probability value includes: extracting a flywheel rotating speed fluctuation curve updated every second and a peak value sequence of adjacent flywheel vibration energy spectrum by using a sliding window mechanism; calculating an axis system vibration energy attenuation rate according to the peak value sequence of the vibration energy spectrum, and combining the rotating speed fluctuation curve to generate a physical collision risk coefficient; comparing the physical collision risk coefficient with a preset risk threshold, and outputting a real-time abnormal probability value according to a comparison result.

[0009] Optionally, the long-time trend processing channel updates the bearing friction coefficient and the efficiency change amount monthly according to a historical degradation database, and generating the cumulative degradation index includes: collecting bearing friction coefficient measurement values of each flywheel unit monthly, and calculating a sliding average value of a change rate of the friction coefficient measurement values of each unit; counting a cumulative number of response delays of a power grid frequency modulation instruction in a corresponding time period to generate a response delay degradation coefficient; and calculating the cumulative degradation index according to the sliding average value and the response delay degradation coefficient, and inversely correcting a weight of the sliding average value according to a storage energy efficiency threshold.

[0010] Optionally, generating the comprehensive prediction state index based on the real-time abnormal probability value and the cumulative degradation index includes: constructing an electrical machinery influence factor matrix, mapping a power grid frequency fluctuation amplitude to a dynamic adjustment parameter of a bearing temperature rise rate; weighting and correcting the real-time abnormal probability value by using the dynamic adjustment parameter to obtain a corrected real-time abnormal probability value; and performing non-linear superposition on the corrected real-time abnormal probability value and the cumulative degradation index to generate the comprehensive prediction state index.

[0011] Optionally, the constructing the electrical machinery influence factor matrix and mapping the power grid frequency fluctuation amplitude to the dynamic adjustment parameter of the bearing temperature rise rate includes: collecting an absolute deviation of a power grid power actual measurement value and a planned value, a rated output power of a flywheel unit, a real-time monitoring frequency of a power grid, a rated power of a power grid, a bearing temperature rise rate, and a vibration energy spectrum peak value, and constructing the electrical machinery influence factor matrix; calculating the dynamic adjustment parameter by using parameters in the electrical machinery influence factor matrix, and ,

[0012] ,

[0013] wherein, is a weight coefficient associated with a fatigue characteristic of the flywheel material, is an attenuation adjustment coefficient based on a bearing operation life, is an absolute deviation of a power grid power actual measurement value and a planned value, is a rated output power of a flywheel unit, is a real-time monitoring frequency of a power grid, for the grid rated frequency, for the synthetic parameter of bearing temperature rise rate and vibration energy spectrum peak value.

[0014] Optionally, the dynamically adjusting the operation control strategy of the energy storage system according to the synthetic prediction state index comprises: when it is detected that the grid frequency change rate exceeds a preset mutation threshold, the weight proportion of the real-time abnormal probability value is increased to a predetermined critical value; when there is no grid mutation, the weight proportion of the cumulative degradation index is increased based on the logarithmic function of the cumulative running time of the flywheel; the synthetic prediction state index is calculated according to the weight proportion distribution result, and the adjustment instruction of the charge and discharge power of each flywheel unit and the maintenance priority ranking are generated by using the size of the synthetic prediction state index.

[0015] Optionally, the method further comprises: calculating an abnormal propagation range evaluation value through the spatial distribution characteristics of the real-time abnormal probability value; if the abnormal propagation range evaluation value and the cumulative degradation index exceed a cooperative early warning threshold at the same time, triggering a joint protection mechanism across the flywheel unit group.

[0016] Optionally, the joint protection mechanism comprises: delimiting a number set of associated flywheel unit groups according to the abnormal propagation range evaluation value; applying a charge and discharge power synchronous restriction strategy to the flywheel units in the number set, while starting the standby capacity preloading to the units outside the set.

[0017] Optionally, the charge and discharge power synchronous restriction strategy comprises: taking the average energy storage efficiency of the target flywheel unit group as a reference value to generate a power deviation compensation coefficient of each unit; inputting the compensation coefficient into a grid frequency control controller to keep the output change rate of the flywheel units in the target group within a preset synchronous interval.

[0018] Based on the same inventive concept, the application also provides a state prediction system for a centralized flywheel energy storage system, which comprises: a data acquisition module for acquiring real-time operation parameters of each flywheel unit in the centralized flywheel energy storage system at the current time; a double-channel processing module comprising a short-time high-frequency processing unit and a long-time trend processing unit, for synchronously inputting the real-time operation parameters into the short-time high-frequency processing channel and the long-time trend processing channel for domain processing, and outputting a real-time abnormal probability value and a cumulative degradation index; a parameter coupling module for generating a synthetic prediction state index based on the real-time abnormal probability value and the cumulative degradation index; and a dynamic decision engine for dynamically adjusting the operation control strategy of the energy storage system according to the synthetic prediction state index.

[0019] Compared with the prior art, the application has the following advantages:

[0020] 1、The present application significantly improves the timeliness and accuracy of fault prediction through the divisional cooperation of short-time high-frequency processing channel and long-time trend processing channel; the short-time high-frequency channel captures the second-level mechanical impact risk, the long-time channel tracks the month-level performance degradation trend, the dual-channel data fusion dynamically generates a comprehensive prediction index, solves the transient misjudgment or long delay missing report problem caused by single time scale analysis of traditional methods, and reduces the probability of unplanned shutdown;

[0021] 2、The present application introduces an electrical mechanical influence factor matrix to realize the implicit correlation modeling of power grid disturbance and mechanical damage, quantifies the dynamic influence of power grid frequency fluctuation on bearing temperature rise rate, maps the frequency deviation amplitude into a mechanical risk correction coefficient, makes the prediction model respond to the power grid frequency modulation demand and the flywheel health state synchronously, and reduces the false alarm situation caused by simply relying on mechanical sensor signals;

[0022] 3、The present application is based on the cooperative warning mechanism of spatial anomaly propagation evaluation and cumulative degradation index, effectively inhibits the cross-unit cascading failure risk, dynamically delimits the high-risk flywheel group through the multi-dimensional matching of regional distribution characteristics of abnormal probability value and historical degradation data, and starts the joint protection strategy to avoid the single device failure diffusion into a system-level accident.

[0023] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and achieved by the structure indicated in the specification, claims and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0025] Figure 1 is a structural schematic diagram of a state prediction method for a centralized flywheel energy storage system according to an embodiment of the present application.

[0026] Figure 2 is a curve diagram of a comprehensive prediction state index changing with time according to an embodiment of the present application.

[0027] Figure 3 is a structural schematic diagram of a state prediction system for a centralized flywheel energy storage system according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0029] With reference to Figure 1 One embodiment of the present application provides a state prediction method for a centralized flywheel energy storage system, which adopts a dual-channel sub-domain analysis technology of short-time high-frequency processing and long-time trend processing, combines with dynamic correction of abnormal evaluation parameters of an electrical mechanical coupling model, and can realize minute-level accurate prediction of a flywheel health state, prolong equipment life and guarantee power grid frequency modulation stability.

[0030] The method of the embodiment specifically comprises:

[0031] Real-time running parameters of each flywheel unit in the centralized flywheel energy storage system at the current moment are acquired, including rotating speed, temperature, adjacent flywheel vibration energy spectrum and power grid frequency modulation signal parameters;

[0032] Specifically, mechanical state parameters and power grid interaction parameters of each flywheel unit are collected in real time by a sensor array, wherein the rotating speed is sampled by a Hall effect sensor at a frequency of 10 times per second, the temperature is measured by a distributed thermocouple array at key nodes of bearings, the vibration energy spectrum is captured by an accelerometer to identify vibration frequency domain characteristics of adjacent flywheel components, and the power grid frequency modulation signal parameters include a frequency change rate and a frequency modulation instruction response time. The real-time running parameters are a set of physical quantities reflecting the dynamic running state of the flywheel unit, including synchronous monitoring data of mechanical performance indicators and power grid interaction parameters; the vibration energy spectrum is a graph of frequency energy distribution by decomposing time domain vibration signals into frequency domain through Fourier transform, and is used for identifying abnormal resonance frequencies.

[0033] The real-time running parameters are synchronously input to a short-time high-frequency processing channel and a long-time trend processing channel for sub-domain processing; wherein the short-time high-frequency processing channel performs second-level dynamic updating on the real-time running parameters based on a preset time window to generate real-time abnormal probability values; the long-time trend processing channel updates bearing friction coefficients and efficiency change amounts according to a historical degradation database on a monthly basis to generate cumulative degradation indexes;

[0034] Specifically, the short-time high-frequency processing channel adopts a 5-second sliding window to continuously update the rotation speed and vibration energy spectrum parameters, and the long-time trend processing channel extracts bearing friction coefficient and frequency modulation response efficiency data from the historical database monthly. After parallel processing of the two channels, real-time abnormal probability values and cumulative degradation indexes are output respectively. Among them, the short-time high-frequency processing channel is an algorithm module for processing transient abnormal signals with a second-level time base, which compresses data delay through a sliding window mechanism; the long-time trend processing channel is an algorithm module for analyzing device degradation based on a monthly time scale, which discovers the progressive failure trend through historical data correlation.

[0035] Based on the real-time abnormal probability value and the cumulative degradation index, a comprehensive prediction state index is generated;

[0036] According to the comprehensive prediction state index, the operation control strategy of the energy storage system is dynamically adjusted.

[0037] Specifically, when the comprehensive index exceeds the safety threshold, the charging and discharging power of the flywheel with high degradation rate is preferentially limited, and the response priority of the standby unit is improved. At the same time, when the grid frequency suddenly changes, the decision weight of the short-time high-frequency channel is expanded in real time to more than 70%, ensuring the rapid suppression of transient abnormalities. Among them, the dynamic adjustment is a closed-loop control logic based on the real-time optimization of power distribution and maintenance scheduling according to the operation risk level; the weight ratio increase is to adjust the influence proportion of the final decision by the adaptive algorithm of the double-channel output, balancing the instantaneous response and long-term reliability demand.

[0038] The application strips the transient abnormality and progressive degradation characteristics of the flywheel system state through the sub-domain processing mechanism, reveals the hidden damage path of the mechanical components caused by power grid disturbance by using the electrical and mechanical coupling model, and finally realizes early warning and precise regulation through dynamic weight decision. Its beneficial effects are: solving the defect that short-term sudden failure and long-term performance degradation are difficult to evaluate cooperatively in traditional methods; avoiding misoperation caused by false alarm of a single sensor; suppressing the risk of multiple device cascading failure through abnormal propagation evaluation, prolonging the overall life of the flywheel cluster, and at the same time ensuring the real-time demand of power grid frequency regulation.

[0039] Optionally, the short-time high-frequency processing channel performs second-level dynamic update on the real-time operation parameters based on a preset time window, and the generation of the real-time abnormal probability value comprises:

[0040] The sliding window mechanism is adopted to extract the flywheel rotation speed fluctuation curve and the adjacent flywheel vibration energy spectrum peak value sequence updated every second;

[0041] Specifically, a sliding window with a width of 5 seconds is set to perform segmented interception of the speed signal every second, and the standard deviation of the speed within the window is used as the fluctuation curve feature. At the same time, a fast Fourier transform is performed on the frequency domain signals collected by the adjacent flywheel vibration sensors, and the maximum amplitude of the 50-200Hz frequency band in the energy spectrum within a 0.1-second interval is extracted to form a peak sequence. Among them, the sliding window mechanism is a dynamic interception and refresh method for fixed time interval data, which maintains the priority calculation of the latest data through second-by-second shifting. The vibration energy spectrum peak sequence is a discrete set of values ​​of the maximum vibration energy within a specific frequency band, arranged in time, and is used to characterize the evolution of mechanical shock risk.

[0042] Calculating the shaft system vibration energy attenuation rate based on the vibration energy spectrum peak sequence, and generating a physical collision risk coefficient in combination with the speed fluctuation curve;

[0043] Specifically, the shaft vibration energy attenuation rate is defined as the relative rate of change of the peak values ​​of two adjacent sampling periods, and is calculated using the first-order difference formula. ,have:

[0044] ,

[0045] Where, For the The peak value of the vibration energy spectrum of seconds, The absolute value of the shaft vibration energy attenuation rate is multiplied by the standard deviation of the speed fluctuation curve, and then normalized to the range of 0-1 to obtain the physical collision risk coefficient. ,have:

[0046] ,

[0047] Where, Compress the data range for the logical function, is the standard deviation of the speed fluctuation curve, is the baseline experience value, is a normalized adjustment parameter. The shaft vibration energy attenuation rate quantifies the rate at which vibration energy dissipates over time, with negative values ​​indicating an intensified vibration state. The physical collision risk coefficient is a probabilistic assessment parameter that combines mechanical impact intensity and the degree of motion instability.

[0048] The physical collision risk coefficient is compared with a preset risk threshold, and a real-time abnormality probability value is output according to the comparison result.

[0049] Specifically, set dynamic risk thresholds , for real-time abnormal probability value ,have:

[0050] ,

[0051] The calculation results are updated every second on the status monitoring interface. The preset risk threshold is a hierarchical trigger boundary value determined based on historical failure case statistics, while the real-time abnormality probability value is a normalized assessment result reflecting the possibility of mechanical collision at the current moment.

[0052] Specifically, the short-term, high-frequency processing solution dynamically couples vibration energy dissipation characteristics with motion stability to construct a physically meaningful collision risk assessment model. Its beneficial effects include: overcoming the hysteresis inherent in traditional threshold alarms responding to slowly accumulating anomalies, proactively capturing potential collision risks through spectral decay rate calculations; distinguishing between normal operating disturbances and structural anomalies by combining speed fluctuations, reducing the probability of false alarms; and employing an adaptive threshold mapping mechanism to ensure predictions are more aligned with engineers' risk perception habits, enhancing alert credibility.

[0053] Optionally, the long-term trend processing channel updates the bearing friction coefficient and efficiency change on a monthly basis according to the historical degradation database to generate a cumulative degradation index, including:

[0054] Collect the bearing friction coefficient measurement values ​​of each flywheel unit on a monthly basis and calculate the sliding average of the change rate of the friction coefficient measurement values ​​of each unit;

[0055] Specifically, the bearing friction coefficient is measured by a torque sensor under fixed load conditions on the first day of each month, and the historical data of the last 12 months is stored to form a time series. The sliding window length L = 6 months is used to calculate the sliding average of the friction coefficient change rate. ,have:

[0056] ,

[0057] Where, For the The friction coefficient change rate of the first month, where Monthly rate of change of friction coefficient ,have:

[0058] ,

[0059] Where, For the The bearing friction coefficient is a dimensionless parameter that reflects the change in the bearing's operating resistance and is calculated in real time using the ratio of input torque to speed. The sliding average is a statistical processing method that averages the rate of change data within a limited time window to smooth out occasional fluctuations.

[0060] The number of response delay accumulations of the grid frequency modulation instruction in the corresponding time period is counted to generate a response delay degradation coefficient;

[0061] Specifically, in the 6-month period corresponding to the sliding window, the number of times that the output adjustment delay of the flywheel unit after receiving the frequency modulation instruction exceeds 500 ms, i.e., the number of response delay accumulations, is counted . For the response delay degradation coefficient , there is:

[0062] ,

[0063] wherein the cubic root correction term is enabled when the number of response delay accumulations is greater than or equal to 3, to prevent excessive punishment caused by high delay times. The number of response delay accumulations is a key event counter reflecting the influence of flywheel mechanical aging on dynamic response capability; and the degradation coefficient is a mapping function output that converts the delay frequency into a numerical evaluation parameter.

[0064] An accumulated degradation index is calculated according to the sliding average and the response delay degradation coefficient, and the weight of the sliding average is inversely corrected according to a storage energy efficiency threshold.

[0065] Specifically, for the accumulated degradation index , there is:

[0066] ,

[0067] wherein, is the weight of the sliding average. The actual storage energy efficiency of the monitoring system is monitored, and when the actual storage energy efficiency is greater than 95% of the rated storage energy efficiency, the weight of the sliding average , when the actual storage energy efficiency decreases to 95% of the rated storage energy efficiency, the weight of the sliding average is adjusted, and the weight of the sliding average , there is:

[0068] ,

[0069] wherein, is the actual storage energy efficiency, is the rated storage energy efficiency. The final accumulated degradation index is subjected to [0, 1] interval truncation processing. Among them, the accumulated degradation index is a normalized evaluation index representing the comprehensive wear degree of the flywheel unit; and the inverse correction degradation weight is a parameter adjustment mechanism that dynamically balances mechanical wear and response delay according to the current performance loss.

[0070] Exemplarily, the sliding average value of the friction coefficient change rate of a certain flywheel unit in the past 6 months is 0.15 / month, and the number of cumulative times of the statistical frequency modulation response delay in the same period is 8. When the actual energy storage efficiency is 92%, the rated energy storage efficiency is 100%, and the actual energy storage efficiency is lower than 95%, the correction is triggered, and the weight of the sliding average value , the response delay degradation coefficient , and thus the cumulative degradation index . Through the dual-factor fusion of mechanical friction degradation and power grid response performance, the implicit correlation between component wear and control system aging is accurately captured, and the problem of insufficient sensitivity caused by traditional methods relying on only a single parameter for early warning is solved; the weight is dynamically corrected using the energy storage efficiency feedback to ensure that the evaluation results truly reflect the actual health status of the flywheel, and to avoid excessive maintenance or maintenance delay caused by the lagging nature of historical data.

[0071] Optionally, generating a comprehensive prediction state index based on the real-time abnormal probability value and the cumulative degradation index comprises:

[0072] An electrical-mechanical influence factor matrix is constructed to map the power grid frequency fluctuation amplitude to a dynamic adjustment parameter of the bearing temperature rise rate;

[0073] Specifically, the electrical-mechanical influence factor matrix is a mathematical mapping model representing the correlation between power grid disturbance and mechanical component thermal stress, integrating the additional temperature rise effect of power grid power fluctuation on the flywheel bearing; the bearing temperature rise rate is the amount of change in bearing temperature per unit time, which is obtained by high-frequency sampling and difference calculation of the thermocouple array.

[0074] The real-time abnormal probability value is weighted and corrected using the dynamic adjustment parameter to obtain a corrected real-time abnormal probability value;

[0075] Specifically, the real-time abnormal probability value output per second of the short-time high-frequency channel is dynamically coupled with the electrical-mechanical influence factor for calculation, and the correction formula is:

[0076] ,

[0077] wherein, is the corrected real-time abnormal probability value, is a system preset reference coupling coefficient, which is set according to the flywheel installation base stiffness, and a typical value is , is a dynamic adjustment parameter. When the power grid frequency fluctuates dramatically, causing , a smoothing filter is automatically enabled to limit the correction amplitude to no more than twice the original value. Weighted correction is the operation process of correcting the transient abnormality evaluation value through electrical-mechanical coupling effect, which strengthens the aggravating effect of power grid disturbance on mechanical risk; the reference coupling coefficient is a standardization adjustment parameter for the sensitivity difference of different flywheel structures to power grid disturbance.

[0078] The corrected real-time abnormal probability value is nonlinearly superimposed with the cumulative degradation index to generate a comprehensive prediction status indicator.

[0079] Specifically, the corrected real-time abnormal probability value and cumulative degradation index are fused using piecewise functions to calculate the comprehensive prediction state index. ,have:

[0080] ,

[0081] Where, 、 is the weight of the corrected real-time abnormal probability value, 、 is the weight of the cumulative degradation index, and nonlinear superposition is used to distinguish the equipment degradation stages to select the calculation strategy of the fusion coefficient, focusing on sudden anomaly detection in the early stage and focusing on the cumulative degradation impact in the later stage.

[0082] For example, Figure 2 The graph shows the time-varying changes in the comprehensive prediction state indicators. This method uses electrical and mechanical influencing factors to reveal the thermal-mechanical coupling damage mechanism of the flywheel bearing caused by grid frequency regulation, and corrects the sensitivity of transient anomaly assessment in real time. It also uses nonlinear superposition to balance the decision weights between short-term risk and long-term performance degradation. This overcomes the drawbacks of traditional linear superposition, which can lead to early false triggering or late missed detection. A dynamic weighting mechanism driven by physical characteristics allows the assessment model to adapt to the changing patterns of mechanical stress under different grid disturbance scenarios, improving the accuracy of identifying complex fault modes.

[0083] Optionally, constructing an electrical machinery impact factor matrix and mapping the grid frequency fluctuation amplitude to a dynamic adjustment parameter of the bearing temperature rise rate includes:

[0084] Collect the absolute deviation between the actual measured value and the planned value of the grid power, the rated output power of the flywheel unit, the real-time monitoring frequency of the grid, the rated power of the grid, the bearing temperature rise rate, and the vibration energy spectrum peak value to build an electrical machinery influencing factor matrix;

[0085] The dynamic adjustment parameters are calculated by using the parameters in the electrical machinery influence factor matrix. ,have:

[0086] ,

[0087] in, is the weight coefficient associated with the fatigue properties of the flywheel material, is the attenuation adjustment coefficient based on the bearing’s service life, an absolute deviation of an actual measurement value of grid power from a planned value, a rated output power of the flywheel unit, a real-time monitoring frequency of the grid, a rated frequency of the grid, a combined parameter of bearing temperature rise rate and vibration energy spectrum peak value.

[0088] Specifically, based on the power deviation and frequency offset parameters in the grid frequency modulation process, a mathematical model of the influence of electrical side disturbance on the thermodynamics of mechanical components is established to obtain a mapping function of the dynamic adjustment parameter, wherein the dynamic adjustment parameter is a combined parameter of the electrical mechanical influence factor matrix, reflecting the quantitative value of the coupling effect of grid disturbance and mechanical wear; the weight coefficient associated with the fatigue characteristics of the flywheel material is dynamically looked up through the material fatigue coefficient table, the carbon fiber rotor takes 0.6, and the steel rotor takes 0.4; the rated output power of the flywheel unit is marked on the equipment nameplate; the attenuation adjustment coefficient based on the bearing operation life is calculated according to the formula , the initial value is 0.8, is the bearing operation life; for the combined parameter of the bearing temperature rise rate and the vibration energy spectrum peak value , there is:

[0089] ,

[0090] In the formula, is the bearing temperature rise rate, is the vibration energy spectrum peak value.

[0091] Wherein, the electrical mechanical influence factor matrix is a quantitative model of the thermal stress effect of the grid power fluctuation and frequency deviation on the flywheel bearing; the material fatigue coefficient table is a dynamic weight query table matched according to the material type, and the material fatigue coefficient is determined according to the fatigue resistance grade of the flywheel rotor material ; the bearing temperature rise rate is the bearing temperature rise rate per minute, which is calculated by difference through a temperature sensor array; the vibration energy spectrum peak value is the maximum value of the vibration acceleration signal in a preset frequency band after Fourier transform.

[0092] Exemplarily, a certain carbon fiber flywheel unit has a rated power , the real-time detected grid power deviation is , the grid frequency is , the bearing temperature rise rate is , the vibration energy spectrum peak value is , and the bearing operation life is years. Among them , the carbon fiber takes the corresponding value 0.6, and after calculation, it can be obtained that . The dynamic adjustment parameter will dynamically adjust the abnormal probability value. This example verifies the coupling mapping mechanism of power grid power fluctuation and mechanical state parameters, and shows the nonlinear amplification effect of continuous low-frequency disturbance on bearing heat accumulation. Through the dynamic combination of material matching weight (carbon fiber ) and annual attenuation coefficient ( ), the sensitivity rising characteristics of old bearings under frequency deviation are accurately reflected, and the problem of decreased evaluation accuracy of traditional fixed weight model in the aging stage of equipment is solved.

[0093] This method forms a cross-influence model of power grid frequency modulation action and flywheel health status by quantifying the hidden damage path of electrical disturbance to mechanical systems. The example shows that intermittent power fluctuations will be strengthened by the quadratic term to enhance their cumulative effect on rotor fatigue, while frequency deviation will accelerate bearing wear evaluation through temperature rise-vibration coupling parameters. Breakthrough the limitation of isolated analysis of electrical and mechanical parameters, capture the indirect damage mechanism of frequent power grid frequency modulation on flywheel life; use the differential adjustment of fatigue resistance characteristics of materials to adjust the weight, and realize the precise operation and maintenance strategy generation of different configuration flywheel clusters.

[0094] Optionally, the dynamic adjustment of the operation control strategy of the energy storage system according to the comprehensive prediction state index comprises:

[0095] When it is detected that the grid frequency change rate exceeds the preset mutation threshold, the weight proportion of the real-time abnormal probability value in the comprehensive prediction state index is increased to a predetermined critical value;

[0096] Specifically, the grid frequency change rate is calculated every second. When the change rate is greater than the preset mutation threshold, it is triggered. Immediately increase the weight proportion of the real-time abnormal probability value output by the short-time high-frequency processing channel in the comprehensive prediction state index from the current value to a predetermined critical value, such as 0.7, to suppress the decision-making influence of the long-time degradation channel. Wherein, the preset mutation threshold is the judgment boundary value of the severity of power grid frequency transient, which is an empirical value set according to the transient stability standard of power system, such as 0.5 Hz / s; the predetermined critical value is the maximum weight limit of 70% to ensure the emergency frequency modulation demand of the power grid, that is, the predetermined critical value is 0.7.

[0097] When it is detected that the grid frequency change rate does not exceed the preset mutation threshold, the weight proportion of the cumulative degradation index is increased based on the logarithmic function of the cumulative running time of the flywheel;

[0098] Specifically, when the grid frequency change rate does not exceed the preset mutation threshold, the cumulative running time logarithmic function is started to calculate the weight adjustment amount, and the weight 、 of the cumulative degradation index is adjusted, and the weight of the cumulative degradation index plus the weight adjustment amount, so as to obtain a new weight 、 . For the weight adjustment amount , wherein:

[0099] ,

[0100] wherein, is an error coefficient, is a cumulative operating hours of the flywheel, is a standard operating hours of the flywheel. The cumulative operating hours logarithmic function is a nonlinear weight adjustment model based on the aggravation trend of equipment aging, which makes the long-term degradation evaluation gradually dominate the decision-making process.

[0101] The comprehensive prediction state index is calculated according to the weight distribution result, and the size of the comprehensive prediction state index is used to generate an adjustment instruction for the charge-discharge power of each flywheel unit and a maintenance priority ranking.

[0102] Exemplarily, the modified real-time abnormal probability value and the cumulative degradation index are respectively weighted and summed with a weight of 70% and 30%. For the flywheel unit whose comprehensive prediction state index exceeds 0.7, an instruction to reduce the discharge power by 20% is generated; for the unit whose comprehensive prediction state index exceeds 0.9, it is marked as a priority maintenance object.

[0103] Exemplarily, when the frequency of the power grid has a step mutation of 1.2 Hz, the frequency change rate reaches 1.8 Hz / s, which is much higher than the preset threshold of 0.5 Hz / s. The real-time abnormal probability weight is immediately increased to 70%. Assuming that the real-time abnormal probability value of a flywheel unit is 0.8 and the cumulative degradation index is 0.6, the comprehensive prediction state index is calculated as 0.74, triggering the power reduction instruction. At the same time, another flywheel unit with an operating time of 1800 hours does not have a mutation event, and the weight adjustment amount of the weight distribution is calculated as 0.217, so that the real-time abnormal probability weight becomes 0.483 and the weight of the cumulative degradation index becomes 0.517, so that the cumulative degradation index is 0.55 and the comprehensive prediction state index is 0.41, maintaining the original power operation. When the power grid has a sudden disturbance, high mechanical risk units are preferentially suppressed to avoid accelerated damage due to excessive frequency modulation; under stable working conditions, the weight of equipment aging evaluation is automatically enhanced to accurately match the preventive maintenance period. By dynamically allocating weights, both transient safety and life management are considered to reduce the probability of failure and outage.

[0104] Optionally, the method further comprises:

[0105] The abnormal propagation range evaluation value is calculated by the spatial distribution characteristics of the real-time abnormal probability value;

[0106] Specifically, the real-time abnormal probability values of adjacent units within a preset range, such as 3 meters, from the abnormal source flywheel unit are counted, and the abnormal propagation range evaluation value is calculated. For the abnormal propagation range evaluation value , wherein:

[0107] ,

[0108] wherein, is the real-time anomaly probability value of the th adjacent unit, is the distance between the th adjacent unit and the anomaly source, is the total number of affected units. When the anomaly propagation range evaluation value of at least 5 units is greater than a preset threshold value such as 0.4, it is determined that there is anomaly propagation. The spatial distribution feature is the geographic diffusion pattern of the anomaly signal in the flywheel cluster; the anomaly propagation range evaluation value is a topological parameter for quantifying the mechanical fault chain risk.

[0109] If the anomaly propagation range evaluation value and the cumulative degradation index exceed the cooperative warning threshold value at the same time, a joint protection mechanism for a group of flywheel units is triggered.

[0110] Specifically, the cooperative warning threshold value is that the anomaly propagation range evaluation value is greater than the evaluation threshold value and the cumulative degradation index is greater than the index threshold value, the evaluation threshold value can be 0.5, and the index threshold value can be 0.65. When both conditions are met, a secondary alarm is sent to the dispatch center and the state of the fault unit group is frozen.

[0111] Illustratively, a flywheel unit A suddenly overheats the bearing, the real-time anomaly probability is 0.9, and there are 6 units within a distance of 0.8-2.5 meters around the flywheel unit A, i.e., 6 units, the anomaly probabilities of which are [0.6, 0.55, 0.7, 0.4, 0.3, 0.2], respectively. The anomaly propagation range evaluation value of the flywheel unit A is calculated to be 0.457. If the cumulative degradation index of the flywheel unit A is 0.68, which is greater than the index threshold value 0.65, but the anomaly propagation range evaluation value 0.457 is less than the evaluation threshold value 0.5, the joint protection is not triggered. If the number of abnormal units around the flywheel unit A increases to 8 and the anomaly propagation range evaluation value is calculated to be 0.53, and the cumulative degradation index is 0.7, the cooperative warning condition is met. Through the dual criteria of space and performance, single false triggering is avoided, and the real fault source with group harm is effectively identified. The system-level shock risk caused by local anomalies is suppressed, and the fault tolerance is improved.

[0112] Optionally, the joint protection mechanism comprises:

[0113] According to the anomaly propagation range evaluation value, a number set of associated flywheel unit groups is determined;

[0114] Specifically, a dynamic zoning is performed with the anomaly source as the center and a radius of 2.5 times the anomaly propagation range evaluation value, all unit numbers in the region and with a cumulative degradation index greater than or equal to 0.5 are screened, and an associated group list is formed.

[0115] A charge and discharge power synchronization restriction strategy is applied to the flywheel units in the numbered set, and a spare capacity preloading is started for the units outside the set.

[0116] Specifically, the upper limit of the charge and discharge power of all units in the numbered set is uniformly set to 85% of the lowest rated power in the group; at the same time, a preload instruction of increasing the power by 20% is sent to healthy units outside the set to maintain the balance of the total output of the power grid.

[0117] For example, when the anomaly propagation range is assessed at 0.6, the zoning radius is 1.5 meters. Assume there are five units within this radius, three of which have cumulative degradation indices of [0.7, 0.55, 0.6]. Two units are selected to satisfy a cumulative degradation index greater than or equal to 0.5. Their power is limited to 85% of the group's lowest rated power of 3 MW, or 2.55 MW. Simultaneously, the power of the 10 external backup units is increased to 3.6 MW, representing the original power of 3 MW x 1.2, while the total output remains unchanged at 45 MW. Dynamic zoning prevents over- or under-protection within a fixed range. Synchronous power limiting prevents accelerated wear of units within the group due to uneven output. A preload mechanism ensures that the grid's frequency regulation capability is not compromised by local power curtailment.

[0118] Optionally, the charging and discharging power synchronous limiting strategy includes:

[0119] Taking the average energy storage efficiency of the target flywheel unit group as the benchmark value, generate the power deviation compensation coefficient of each unit;

[0120] Specifically, the average energy storage efficiency within the associated group is calculated, and the compensation coefficient for each unit efficiency is calculated using the following formula:

[0121] ,

[0122] Where, is the compensation coefficient, For the The efficiency of each unit, is the average energy storage efficiency within the associated group. and When , which allows a 1.1% increase in output.

[0123] The compensation coefficient is input into the grid frequency regulation controller to keep the output change rate of the flywheel units in the target group within a preset synchronization range.

[0124] Specifically, the frequency modulation controller uses a compensation coefficient to constrain the output adjustment speed to ensure that the power change of each unit per minute does not exceed 1.5 times the rated value. For units with this feature, the maximum adjustment per minute increases from 1.5% to 1.5165%.

[0125] Exemplarily, the associated group average efficiency is 88%, and the three units have efficiencies of 85%, 88%, and 91%. Corresponding compensation coefficients are 0.983, 1, and 1.017. When the output needs to be increased by 5%, the three units are adjusted at the coefficient rate, and the actual change rates are 4.915%, 5%, and 5.085%, respectively. The deviation is controlled within ±0.085%. The differentiated compensation based on the efficiency difference avoids overloading of low-efficiency units, maintains the collaborative response speed within the group through the synchronization interval constraint, improves the frequency modulation accuracy, and reduces the mechanical stress.

[0126] Based on the same inventive concept, as shown in Figure 3 The application also provides a state prediction system for a centralized flywheel energy storage system, the system comprising:

[0127] a data acquisition module configured to acquire real-time operating parameters of each flywheel unit in the centralized flywheel energy storage system at a current time point;

[0128] a dual-channel processing module comprising a short-time high-frequency processing unit and a long-time trend processing unit, configured to input the real-time operating parameters into a short-time high-frequency processing channel and a long-time trend processing channel for domain processing, and output real-time abnormal probability values and cumulative degradation indexes;

[0129] a parameter coupling module configured to generate a comprehensive predicted state index based on the real-time abnormal probability values and the cumulative degradation indexes;

[0130] a dynamic decision engine configured to dynamically adjust an operating control strategy of the energy storage system according to the comprehensive predicted state index.

[0131] It should be noted that the electrical connection between the above-mentioned units does not necessarily represent a direct connection, and an indirect connection mode can also be used as long as the purpose of the application is achieved. The above-mentioned is only an exemplary embodiment of the application, and cannot limit the scope of the application.

[0132] That is, any equivalent changes and modifications made in accordance with the teachings of the present application are still within the scope of the present application. Other embodiments of the present application will be readily apparent to those skilled in the art upon considering the specification and practice of the true principles disclosed herein. The present application is intended to cover any variations, uses, or adaptive changes to the present application that follow the general principles of the present application and include common knowledge or conventional technical means in the art that are not described in the present application.

Claims

1. A state prediction method for a centralized flywheel energy storage system, characterized in that, The method comprises: acquiring real-time operating parameters of each flywheel unit in the centralized flywheel energy storage system at the current moment, including rotational speed, temperature, adjacent flywheel vibration energy spectrum and power grid frequency modulation signal parameters; synchronously inputting the real-time operating parameters into a short-time high-frequency processing channel and a long-time trend processing channel for domain processing; wherein the short-time high-frequency processing channel performs second-level dynamic updating on the real-time operating parameters based on a preset time window to generate a real-time abnormal probability value; the long-time trend processing channel updates the bearing friction coefficient and efficiency change amount according to a historical degradation database on a monthly basis to generate a cumulative degradation index; generating a comprehensive prediction state index based on the real-time abnormal probability value and the cumulative degradation index; dynamically adjusting the operation control strategy of the energy storage system according to the comprehensive prediction state index; wherein the generation of the real-time abnormal probability value comprises: extracting the flywheel rotational speed fluctuation curve and the adjacent flywheel vibration energy spectrum peak value sequence updated every second by using a sliding window mechanism; calculating the shaft vibration energy attenuation rate according to the vibration energy spectrum peak value sequence, and combining the rotational speed fluctuation curve to generate a physical collision risk coefficient; comparing the physical collision risk coefficient with a preset risk threshold value, and outputting a real-time abnormal probability value according to the comparison result; wherein the generation of the cumulative degradation index comprises: collecting the bearing friction coefficient measurement values of each flywheel unit on a monthly basis, calculating the sliding average of the change rate of each unit friction coefficient measurement value; counting the cumulative number of response delays of the grid frequency modulation instructions within the corresponding time period to generate a response delay degradation coefficient; calculating the cumulative degradation index according to the sliding average and the response delay degradation coefficient, and inversely correcting the weight of the sliding average according to the energy storage efficiency threshold.

2. A state prediction method for a centralized flywheel energy storage system according to claim 1, characterized in that, The generation of the comprehensive prediction state index based on the real-time abnormal probability value and the cumulative degradation index comprises: constructing an electrical and mechanical influence factor matrix to map the grid frequency fluctuation amplitude to a dynamic adjustment parameter of the bearing temperature rise rate; using the dynamic adjustment parameter to weight and correct the real-time abnormal probability value to obtain a corrected real-time abnormal probability value; nonlinearly superimposing the corrected real-time abnormal probability value and the cumulative degradation index to generate a comprehensive prediction state index.

3. A state prediction method for a centralized flywheel energy storage system according to claim 2, characterized in that, The construction of the electrical and mechanical influence factor matrix to map the grid frequency fluctuation amplitude to the dynamic adjustment parameter of the bearing temperature rise rate comprises: collecting the absolute deviation of the actual and planned values of the grid power, the rated output power of the flywheel unit, the real-time monitored frequency of the grid, the rated power of the grid, the bearing temperature rise rate and the vibration energy spectrum peak value to construct the electrical and mechanical influence factor matrix; The dynamic adjustment parameter is calculated by using the parameters in the electromechanical influence factor matrix, and the dynamic adjustment parameter is used to adjust the power grid dynamic adjustment parameter , ， wherein, is a weight coefficient associated with the fatigue properties of the flywheel material, is a decay adjustment coefficient based on the bearing operating age, is the absolute deviation of the grid power actual measurement from the planned value, is the rated output power of the flywheel unit, is the grid real-time monitoring frequency, is the grid rated frequency, is a synthetic parameter of the bearing temperature rise rate and the vibration energy spectrum peak value.

4. The state prediction method for a centralized flywheel energy storage system according to claim 1, wherein, The dynamic adjustment of the operation control strategy of the energy storage system according to the comprehensive prediction state index comprises: when detecting that the grid frequency change rate exceeds a preset mutation threshold, increasing the weight proportion of the real-time abnormal probability value to a predetermined critical value; when there is no grid mutation, increasing the weight proportion of the cumulative degradation index based on the logarithmic function of the cumulative running time of the flywheel; calculating the comprehensive prediction state index according to the weight proportion allocation result, and generating adjustment instructions for the charge and discharge power of each flywheel unit and maintenance priority sorting by using the size of the comprehensive prediction state index.

5. The state prediction method for a centralized flywheel energy storage system according to claim 1, wherein, The method further comprises: calculating an abnormal propagation range evaluation value through the spatial distribution characteristics of the real-time abnormal probability value; if the abnormal propagation range evaluation value and the cumulative degradation index exceed the cooperative early warning threshold value at the same time, triggering a joint protection mechanism across the flywheel unit group.

6. A state prediction method for a centralized flywheel energy storage system according to claim 5, characterized in that, The joint protection mechanism comprises: delineating a number set of associated flywheel unit groups according to the abnormal propagation range evaluation value; applying a charge-discharge power synchronous restriction strategy to the flywheel units in the number set, while starting the standby capacity preloading for the units outside the set.

7. A state prediction method for a centralized flywheel energy storage system according to claim 6, characterized in that, The charge-discharge power synchronous restriction strategy comprises: generating a unit power deviation compensation coefficient based on the average energy storage efficiency of the target flywheel unit group as a reference value; inputting the compensation coefficient into the grid frequency regulation controller to keep the output change rate of the flywheel units in the target group within the preset synchronous interval.

8. A state prediction system for a centralized flywheel energy storage system, applied to the state prediction method for a centralized flywheel energy storage system according to any one of claims 1-7, characterized in that, The system comprises: a data acquisition module for acquiring real-time operating parameters of each flywheel unit in the centralized flywheel energy storage system at the current time; a dual-channel processing module including a short-time high-frequency processing unit and a long-time trend processing unit, for synchronously inputting the real-time operating parameters into the short-time high-frequency processing channel and the long-time trend processing channel for domain processing, and outputting real-time abnormal probability values and cumulative degradation indexes; a parameter coupling module for generating a comprehensive prediction state index based on the real-time abnormal probability values and the cumulative degradation indexes; a dynamic decision engine for dynamically adjusting the operating control strategy of the energy storage system according to the comprehensive prediction state index; wherein the short-time high-frequency processing unit is configured to extract a flywheel speed fluctuation curve updated every second and a peak value sequence of adjacent flywheel vibration energy spectrum using a sliding window mechanism, calculate an axis vibration energy attenuation rate according to the peak value sequence of the vibration energy spectrum, and combine the speed fluctuation curve to generate a physical collision risk coefficient, and compare the physical collision risk coefficient with a preset risk threshold value to output a real-time abnormal probability value according to the comparison result; wherein the long-time trend processing unit is configured to collect bearing friction coefficient measurement values of each flywheel unit monthly, calculate a sliding average value of the change rate of the friction coefficient measurement values of each unit, count the cumulative number of response delays of the grid frequency regulation instructions within the corresponding time period to generate a response delay degradation coefficient, and calculate a cumulative degradation index according to the sliding average value and the response delay degradation coefficient, and inversely correct the weight of the sliding average value according to the energy storage efficiency threshold value.

Images