Flywheel energy storage and battery hybrid ups system for data centers

Abstract

The application relates to the technical field of power management, in particular to a flywheel energy storage and battery hybrid UPS system for a data center, which comprises a load identification module, a disturbance correction module, a scheduling screening module, a time reconstruction module and an instruction updating module. In the application, the current rotating speed and output current of the flywheel are acquired, slope change trend analysis is performed, the impact characteristics generated when the load starts can be identified in advance, and the transient instability risk caused by the large current impact can be avoided. The original access timing of the load and the power disturbance response of the power supply side are cross-analyzed, so that the load access behavior originally concentrated in timing and with intense current fluctuation can be orderly expanded in the time domain. The power supply parameters of the air conditioner group are extracted by the flywheel in the steady-state operation period as a reference, interval adaptation judgment is performed on the operation characteristics of the to-be-accessed operation and maintenance equipment, the equipment with high stability is preferentially scheduled, and the balance and controllability of the overall scheduling response are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power management, in particular to a flywheel energy storage and battery hybrid UPS system for data centers. BACKGROUND

[0002] The technical field of power management includes research and application of distribution control conversion and regulation of power resources, mainly used to ensure the continuous and stable operation of electrical equipment under various power supply conditions. It includes the construction of emergency response mechanism for unstable power supply, the guarantee measures for power supply continuity and the coordinated control of various backup power supply modes.

[0003] Among them, the flywheel energy storage and battery hybrid UPS system for data centers refers to a multi-energy storage joint power supply architecture constructed by connecting flywheel energy storage devices and battery packs in parallel to realize uninterrupted power supply switching control. It mainly aims at the problems of inconsistent power supply switching response time, single power supply redundancy configuration and insufficient instantaneous load support capacity of traditional battery UPS in data centers under sudden power failure or voltage fluctuation.

[0004] The traditional UPS system relies on single energy storage mode of battery to respond instantaneously when facing sudden power failure or voltage fluctuation. In the process of starting to deal with peak load, the response often lags behind, and it cannot match the instantaneous power demand of the sudden load increase. The load access timing depends on the preset strategy or static judgment, without considering the dynamic changes of current power supply capacity and load behavior characteristics, resulting in uneven current distribution in the power supply switching stage, frequent mis-triggering of relays. Lack of load access adjustment mechanism based on real-time power disturbance trajectory, when multiple loads are accessed at similar time points, it forms a current peak, aggravates the system instability factors, and may cause flywheel speed drop or battery overload discharge. The scheduling of stable devices such as air conditioners and fluctuating operation and maintenance devices is not distinguished, and the scheduling strategy lacks the basis for adapting to the running characteristics between devices, so the system cannot effectively filter incompatible targets and interfere with the scheduling order of other loads. The original control trigger logic cannot dynamically adjust to the real-time state of load access, resulting in that the control command cannot respond to the changes in load demand in time, causing relay logic delay in the power supply link to affect the overall response speed, and reducing the continuous power supply reliability of data centers in sudden scenarios. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art, and a flywheel energy storage and battery hybrid UPS system for data centers is proposed.

[0006] In order to achieve the above purpose, the present application adopts the following technical scheme: a flywheel energy storage and battery hybrid UPS system for data centers comprises:

[0007] The load identification module obtains the current rotating speed of the flywheel and the current output current and performs synchronous slope analysis, extracts a starting current threshold of the current to-be-connected load, screens devices with a starting current threshold greater than a flywheel power supply limit, and marks a load delay item set;

[0008] The disturbance correction module obtains a flywheel output power change and a battery voltage response change trajectory and performs clustering division, performs backward processing on an original access time point of a device in the load delay item set according to a division result, and generates a load delay adjustment sequence;

[0009] The scheduling screening module obtains a current mean of a data center air conditioner group in a steady-state running cycle of flywheel power supply and a standard deviation of apparent power output, compares the current to-be-connected operation and maintenance type device, screens an operation and maintenance device early scheduling list, and performs early scheduling on the operation and maintenance device.

[0010] The time reconstruction module reconstructs a time axis of a load in a data center main power supply channel based on the load delay adjustment sequence and the operation and maintenance device early scheduling list, and forms a load control time mapping index table.

[0011] The instruction updating module updates a trigger control process in a data center main current channel of a flywheel and a battery combination through the load control time mapping index table, and forms a trigger control scheduling instruction flow.

[0012] As a further scheme of the application, the load delay item set includes devices with a starting current threshold exceeding a flywheel power supply limit, to-be-connected loads that need to be delayed, and load targets identified through synchronous slope analysis, the load delay adjustment sequence includes an offset value of an original access time point, a disturbance grouping identifier corresponding to clustering division, and a load mapping item corresponding to a power disturbance peak, the operation and maintenance device early scheduling list includes devices meeting a current mean requirement, devices meeting a standard deviation of apparent power output condition, and operation and maintenance type devices consistent with air conditioner group running characteristics, the load control time mapping index table includes an access time update item in the operation and maintenance device early scheduling list, an access time adjustment item in the load delay adjustment sequence, and a time axis arrangement item after unified mapping, and the trigger control scheduling instruction flow includes a load access sequence update instruction, a trigger control signal in a main current channel, and a scheduling trigger action corresponding to a time mapping index.

[0013] As a further scheme of the application, the load identification module includes:

[0014] The rotating speed and current extraction submodule obtains continuous sampling data of the current rotating speed and the current output current of the flywheel, calculates a rotating speed change rate and a current change rate in each sampling period, and generates a rotating speed and current joint change rate set.

[0015] The slope mutation determination sub-module adopts a local weighted regression algorithm to perform trend fitting based on the set of speed-current combined change rates, performs continuous period slope gradient calculation on a fitting residual sequence, and judges whether the residual gradient continuously rises and exceeds the average background fluctuation amplitude of the trend section, to generate a trend slope mutation identification interval;

[0016] The power supply limit screening sub-module extracts the starting current threshold of the load according to the load to be accessed identified in the trend slope mutation identification interval, and compares it with the current power supply limit value of the flywheel, screens the target load whose starting current threshold is greater than the power supply limit value, and generates a load delay item set.

[0017] As a further scheme of the application, the disturbance correction module comprises:

[0018] The power feature extraction sub-module obtains the flywheel output power change and the battery voltage response change trajectory, extracts the flywheel power change rate, the battery voltage mutation point time, the response delay of the intersection of the two, and performs normalization processing, to generate a disturbance interval feature parameter set;

[0019] The staggered disturbance clustering sub-module adopts a spectral clustering algorithm to cluster and divide the voltage and current mutation point positions generated in the alternate power supply process according to the disturbance interval feature parameter set, extracts the overlapping section with a disturbance intensity greater than the average disturbance threshold in the clustering result, and obtains the disturbance peak clustering distribution information;

[0020] The access time adjustment sub-module calls the original access time point of each device in the load delay item set according to the disturbance peak clustering distribution information, maps it with the disturbance duration of the corresponding overlapping section, performs time point backward offset processing, and generates a load delay adjustment sequence.

[0021] As a further scheme of the application, the scheduling screening module comprises:

[0022] The power feature calculation sub-module obtains the current data and apparent power data of the air conditioner group of the data center in the steady-state operation period of the flywheel power supply, respectively calculates the current mean and apparent power output standard deviation in the time period, and forms a power supply side reference comparison item, to generate air conditioner group operation feature parameters;

[0023] The device matching comparison sub-module extracts the starting current demand and operating power factor of the current to-be-accessed operation and maintenance type device according to the air conditioner group operation feature parameters, respectively calculates the deviation value with the power supply side reference comparison item, and performs interval intersection determination on the deviation value set of all devices, to obtain a device operation adaptation interval;

[0024] The access load screening submodule is used for screening operation and maintenance type devices in the adaptive interval of the device, extracting device access attributes, arranging the unique identification and operation parameters of each device, and generating an operation and maintenance device advance scheduling list.

[0025] As a further scheme of the application, the time reconstruction module comprises:

[0026] The time offset identification submodule is used for extracting the time offset value and original access time of each load based on the load delay adjustment sequence and the operation and maintenance device advance scheduling list, classifying the forward offset and reverse offset respectively, and generating load access offset structure data.

[0027] The access time rearrangement submodule is used for replacing the original access timeline of the current main power supply channel as a whole according to the load access offset structure data, moving the delay items backward and moving the advance items forward, rearranging the access positions according to the adjusted time sequence, and obtaining a time axis sequence rearrangement result.

[0028] The time index generation submodule is used for extracting the new access time point and relative position number in the rearrangement list corresponding to each load, integrating the load unique identification and time coordinate information, and generating a load control time mapping index table.

[0029] As a further scheme of the application, the instruction update module comprises:

[0030] The time node extraction submodule is used for extracting the adjusted start time field, constructing an update sequence according to the time sequence and binding the relay action sequence number, generating a control instruction time sorting result, and calling all load identification and corresponding access time information in the load control time mapping index table.

[0031] The control logic rearrangement submodule is used for replacing the relay trigger logic sequence in the main current channel shared by the flywheel and the battery as a whole according to the control instruction time sorting result, mapping the original relay instruction and the updated time sequence execution, and rearranging the control trigger order according to the time sequence, to generate a relay response rearrangement queue.

[0032] The trigger flow generation submodule is used for writing the relay action and corresponding time sequence into the execution list of the control channel contact synchronously, combining the adjusted control trigger behaviors into a unified task flow, generating a trigger control scheduling instruction flow, and coordinating the load access time sequence of the flywheel energy storage and the battery in the hybrid UPS structure, to realize the power supply trigger sequence control in the main current channel of the data center.

[0033] Compared with the prior art, the application has the following advantages and positive effects:

[0034] In the application, by acquiring the current speed and output current of the flywheel, the slope change trend analysis is performed, the impact characteristics generated when the load starts are recognized in advance, and the load target that the power supply capacity is difficult to cover is screened out, and the transient instability risk caused by large current impact is avoided. Cross analysis of the original access timing of the load and the power disturbance response of the power supply side is performed, the disturbance intensity distribution in alternating power supply is identified and the load delay access node is matched, so that the load access behavior with concentrated timing and intense current fluctuation is orderly expanded in the time domain. The power supply parameters of the air conditioner group are extracted by the flywheel in the steady state running period as the reference, and the interval adaptive judgment is performed on the running characteristics of the to-be-accessed operation and maintenance equipment, the target with large power deviation is eliminated, the equipment with high stability is preferentially scheduled, and the balance and controllability of the overall scheduling response are improved. After the access time of all loads is processed in forward and reverse directions, the time axis structure is rearranged, the timing reconstruction of different types of equipment is realized according to the dynamic load intensity, the current peak value at the load access moment is reduced, and the instruction control precision of the relay action in the power supply path is improved. The new access time of each load is mapped to a unified scheduling trigger flow, the control action is output synchronously through the relay contact, and the trigger response jitter caused by control delay in the power supply link is reduced. The fine-grained time control of the cooperative access of multiple types of equipment is realized under the background of dynamic load, the stability and flexibility of the main power supply channel in the high-frequency disturbance scene are improved, and the cooperative efficiency of the flywheel and the battery hybrid power supply is enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 The system flowchart of the application is shown in the figure;

[0036] Figure 2 The flowchart of the load identification module of the application is shown in the figure;

[0037] Figure 3 The flowchart of the disturbance correction module of the application is shown in the figure;

[0038] Figure 4 The flowchart of the scheduling screening module of the application is shown in the figure;

[0039] Figure 5 The flowchart of the time reconstruction module of the application is shown in the figure;

[0040] Figure 6 The flowchart of the instruction updating module of the application is shown in the figure. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical scheme and advantages of the application clearer and more understandable, the application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application and do not limit the application.

[0042] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0043] Please refer to Figure 1 A flywheel energy storage and battery hybrid UPS system for a data center comprises:

[0044] The load identification module obtains the current speed and current output current of the flywheel and performs synchronous slope analysis to extract the starting current threshold of the current load to be connected, screens the devices whose starting current threshold is greater than the flywheel power supply limit, and marks the load delay item set;

[0045] The disturbance correction module obtains the flywheel output power change and the battery voltage response change trajectory and performs clustering division, performs backward processing on the original access time point of the devices in the load delay item set according to the division result, and generates a load delay adjustment sequence;

[0046] The scheduling screening module obtains the current mean and apparent power output standard deviation of the air conditioning group in the steady-state running cycle of the flywheel power supply in the data center, compares it with the current operation and maintenance type device to be connected, screens the operation and maintenance device early scheduling list;

[0047] The time reconstruction module maps and reorganizes the time axis of the load in the main power supply channel of the data center based on the load delay adjustment sequence and the operation and maintenance device early scheduling list, and forms a load control time mapping index table;

[0048] The instruction updating module updates the trigger control process in the main current channel of the data center combined with the flywheel and the battery through the load control time mapping index table, and forms a trigger control scheduling instruction stream;

[0049] The load delay item set includes devices whose starting current threshold exceeds the flywheel power supply limit, to-be-connected loads that need to be delayed to be connected, and load targets identified through synchronization slope analysis, the load delay adjustment sequence includes an offset value of an original connection time point, a disturbance grouping identifier corresponding to a cluster division, and a load mapping item corresponding to a power disturbance peak, the operation and maintenance device advance scheduling list includes devices that meet the current mean value requirement, devices that meet the apparent power output standard deviation condition, and operation and maintenance devices that are consistent with the air conditioner group operation characteristics, the load control time mapping index table includes a connection time update item in the operation and maintenance device advance scheduling list, a connection time adjustment item in the load delay adjustment sequence, and a time axis arrangement item after unified mapping, and the trigger control scheduling instruction stream includes a load connection sequence update instruction, a trigger control signal in the main current channel, and a scheduling trigger action corresponding to the time mapping index.

[0050] Referring to Figure 2 , the load identification module includes:

[0051] The rotation speed and current extraction submodule obtains continuous sampling data of the current rotation speed and output current of the flywheel, calculates the rotation speed change rate and current change rate in each sampling period, and generates a rotation speed and current joint change rate set;

[0052] The continuous sampling data of the current rotation speed and output current of the flywheel are obtained, which needs to be based on the 1-second detection period set in the operation process of the data center flywheel energy storage system, and isochronous sampling is performed at a fixed time interval of every 1 millisecond in the period, a total of 20 groups of instantaneous data of flywheel angular velocity and output current are collected. During the sampling process, the flywheel angular velocity is recorded as , wherein represents the flywheel angular velocity, and the unit is radian per minute (rad / min), and the subscript represents the th sampling moment; the output current is denoted as , wherein represents the instantaneous output current value, and the unit is ampere (A). The change between every two adjacent sampling moments and is calculated to obtain the change rate in unit time. Let represent the index of the rotation speed change rate, corresponding to the sampling time point , the next moment is , and the rotation speed change rate calculation formula is:

[0053] ;

[0054] , wherein represents the unit time change rate of the flywheel angular velocity in the th group of data, represents the angular velocity value of the th sampling point, This represents the angular velocity value at the next instant. The sampling time interval is constant at 1ms.

[0055] set up The index represents the rate of change of current, corresponding to the sampling time point. The next moment is The formula for calculating the rate of change of current is:

[0056] ;

[0057] in Indicates the first The rate of change of output current per unit time in the data set. Indicates the first Current values ​​at each sampling point This indicates the current value at the next moment.

[0058] Taking actual sampling data as an example, if the flywheel angular velocity is at the 6th sampling time... Group 7 is ,but:

[0059] ;

[0060] If the current is at the corresponding time ,but:

[0061] ;

[0062] The rate of change was calculated sequentially for all 19 adjacent sampling points, forming a result from... The resulting data set is denoted as This set is used to describe the instantaneous synergistic relationship between flywheel speed fluctuations and current response, serving as the input data source for subsequent trend modeling and anomaly identification, and ultimately generating a set of joint speed and current change rates.

[0063] The slope mutation determination submodule is based on the set of joint change rates of rotation speed and current. It uses a local weighted regression algorithm to fit the trend, performs continuous periodic slope gradient calculation on the fitted residual sequence, and determines whether the residual gradient rises continuously and exceeds the average background fluctuation amplitude of the trend segment, thereby generating the trend slope mutation identification interval.

[0064] Based on the set of combined speed and current change rates ,in, Indicates the first The rate of change of flywheel speed within each sampling interval, in rad / min / ms. Indicates the first The rate of change of current within each sampling interval, in A / ms, is given by... ,by As the independent variable, Using as the dependent variable, a locally weighted regression model is constructed, with the regression function in the form of:

[0065] ;

[0066] in, The fitted result represents the first... Predicted current change rate for each sample point This indicates the bias term, that is, in Predicted rate of change of current at time The trend slope coefficient represents the magnitude of the current response change corresponding to a unit rate of change in rotational speed. Indicates the first Rate of change of rotational speed at each sample point.

[0067] The sample weighting during regression is determined by the Gaussian kernel function, and the weights are expressed as:

[0068] ;

[0069] in, For the first Each sample pair fits the center point The weight, For the first The rate of change of rotation speed for each sample The rotational speed change rate at the current fitted center point. This is the bandwidth coefficient, controlling the weight decay rate. The setting is based on the actual sampled data within the flywheel's variable response period. Variance setting Its standard deviation For example, in the sampled data The sample standard deviation is Then let .

[0070] Regression coefficient , Derived from the weighted least squares criterion, the method that minimizes the weighted sum of squared residuals is:

[0071] ;

[0072] in, Let be the objective function. For the number of regression samples, For the first The rate of change of current for each sample is regressed by solving the normal equations. , make Minimum.

[0073] Then the residual value of each fitting point is calculated:

[0074] ;

[0075] wherein, represents the residual value of the i-th sample point, measuring the fitting error, with the unit of A / ms.

[0076] All residual sequences are divided into continuous sliding detection windows, and the i-th sliding detection window corresponds to the residual value , and the residual slope is calculated as:

[0077] ;

[0078] wherein, is the residual slope of the i-th residual point in the i-th sliding detection window, represents the global index position of the i-th residual point in the i-th sliding detection window, represents the next residual point of , which is used to construct two data points in the calculation of the pair of residual slopes, is the difference value of the angular velocity change rate between the i-th residual point and the i-th residual point in the i-th sliding detection window. Then the average value of all residual slopes in the sliding detection window is taken as the residual increase rate of the sliding detection window : ; wherein,

[0079] represents the average residual growth rate in the i-th sliding detection window, reflecting the strength of the fitting residual change trend, is the number of calculable residual slopes in the sliding detection window.

[0080] ;

[0081] wherein, represents the average residual growth rate in the i-th sliding detection window, reflecting the strength of the fitting residual change trend, is the number of calculable residual slopes in the sliding detection window. The mutation judgment threshold

[0082] is set, which is set according to the average value of all in the historical steady-state running period, and is represented as:

[0083] ; wherein,

[0084] ​​​​​​represents the amplification coefficient, used to improve the mutation judgment sensitivity, the setting basis is the upper limit of the system fluctuation under non-mutation condition, if the empirical evaluation is 1.5 times of the value, then set , .

[0085] Take , set a group of actual sampling data as: , the residual value .

[0086] The corresponding residual slope is calculated as follows:

[0087] ;

[0088] ;

[0089] ;

[0090] .

[0091] The residual increment rate of the sliding detection window is:

[0092] ;

[0093] Since , the system records the sliding detection window number , the sample time interval in it, and the load number in the access queue in this time period to form the trend slope mutation identification interval.

[0094] The power supply limit screening sub-module extracts the starting current threshold of the load according to the to-be-accessed load identified in the trend slope mutation identification interval, and compares it with the current power supply limit value of the flywheel, screens the target load whose starting current threshold is greater than the power supply limit value, and generates a load delay item set;

[0095] According to the trend slope mutation, the identified to-be-connected load number in the interval is identified, and the starting current threshold information of the corresponding load is first called from the data center load management database. The information is usually defined in the factory configuration parameters of the equipment and has been recorded in the load management table. Taking three to-be-connected loads corresponding to the number of the mutation interval detected by a certain flywheel power supply section as an example, they are load A, load B and load C, and their corresponding starting current thresholds are 220 A, 180 A and 250 A, respectively, with the unit being ampere (A). At the same time, the running state monitoring data of the flywheel control unit is called to obtain the upper limit of the output capacity of the flywheel in the mutation identification period, i.e., the power supply limit value, which is affected by the remaining energy storage state, the current speed and the voltage steady state, and can fluctuate in real time. For example, the power supply limit is 200 A at the detection time. Then, the starting current threshold of each to-be-connected load is compared with the power supply limit value one by one to determine whether there is an exceeding limit. The load whose starting current threshold is greater than the power supply limit value will be considered as not directly connectable in the current period to prevent transient voltage drop caused by flywheel output load over-limit. For example, the starting current of 220 A of load A is higher than the power supply limit of 200 A, so it needs to be delayed to be connected; the starting current of 180 A of load B is lower than the limit value, so it can be connected without adjustment; the starting current of 250 A of load C also exceeds the power supply range, so it also needs to be delayed to be started. For all loads whose starting current thresholds are higher than the current flywheel power supply limit, their original starting times in the connection sequence are marked and unified into a delay processing list. The list not only retains the unique identifier and starting current value of the load, but also contains the original set connection time point and the associated number corresponding to the mutation identification sliding detection window, which serves as the basic data for subsequent disturbance avoidance and connection sequence reconstruction processing. Finally, the load delay item set is summarized and generated.

[0096] Referring to Figure 3 , the disturbance correction module comprises:

[0097] The power feature extraction submodule obtains the flywheel output power change and the battery voltage response change trajectory, extracts the flywheel power change rate, the battery voltage mutation point time, the response delay of the intersection of the two, and performs normalization processing to generate a disturbance interval feature parameter set.

[0098] On the basis of the set of load delay items, the delay load device number and the original access period recorded therein are called one by one, the corresponding power supply channel of each delay load is located in the flywheel-battery hybrid power supply architecture, and the power output within 0.5 seconds before and after the access moment and the battery voltage response track are included in the joint analysis window. The flywheel output power data is obtained by multiplying the bus voltage and current measurement values, for example, if the delay before access time of a certain delay load L1 is 2.65 seconds, the observation joint analysis window starting point is set to 2.15 seconds, the sampling interval is 2 ms, and the flywheel power value sequence measured in the joint analysis window is 5.6 kW, 5.1 kW, 4.4 kW, 3.9 kW, 3.2 kW, and 2.6 kW. According to the adjacent point difference, the power change rate is calculated, and the sudden drop point occurs at the 5th sampling point, i.e. 2.23 seconds. The battery voltage response value sequence is 232V, 229V, 227V, 220V, 211V, and 204V. Comparative analysis shows that the maximum voltage drop occurs at the 6th point, i.e. 2.24 seconds, and the point is extracted as the voltage mutation reference point. Then, the time difference between the power mutation point and the voltage mutation point is calculated to obtain the response delay time. In this example, the response delay is 1 sampling period, i.e. 2 ms, which will be one of the subsequent disturbance characteristics. Then, the maximum value of the flywheel power change rate and the voltage drop amplitude in this section are calculated, for example, the flywheel power decreases by 2.4 kW from 2.21 seconds to 2.23 seconds, and the voltage decreases by 25 V from 2.22 seconds to 2.24 seconds. The subsequent 3 index values are normalized, wherein the normalization reference value is set to the flywheel rated maximum power change amplitude 8.0 kW, the voltage change maximum drop amplitude 60 V, and the response delay maximum observation tolerance 20 ms. After normalization, the power change rate is 0.30, the voltage drop is 0.42, and the response delay is 0.10. The combination of the three values forms the feature vector of a single disturbance sample. The power-voltage trajectory information in the power supply channel of each delay load is collected according to the processing logic, the corresponding feature vectors are extracted and arranged in order, and the normalized feature data sequence of the disturbance interval is constructed. Finally, the disturbance interval characteristic parameter set is obtained.

[0099] The interleaved disturbance clustering submodule clusters the voltage and current mutation points generated in the alternate power supply process according to the disturbance interval characteristic parameter set, extracts the overlapping sections with disturbance intensity greater than the average disturbance threshold in the clustering result, and obtains the disturbance peak clustering distribution information.

[0100] Based on the disturbance interval characteristic parameter set, each disturbance sample in the set is composed of three-dimensional normalized values, which are flywheel power change rate, battery voltage drop amplitude and response delay time. These values together constitute the feature vector of the sample. In order to identify the correlation trend between the samples, the similarity between any two samples is calculated first. The Euclidean distance is used to measure the difference between the two samples in the three-dimensional disturbance characteristic space. Then the distance value is substituted into the similarity calculation formula to represent the closeness of the two samples in the disturbance behavior. The similarity calculation is expressed as follows:

[0101] ;

[0102] wherein, , represent the disturbance feature vectors of the first and the second sample, represents the Euclidean distance between them, and the unit is a dimensionless normalized value, is the similarity distance scale coefficient, which is used to control the sensitivity of the distance difference to the similarity, and the value is set according to the average value of the standard deviation of all sample pairs. For example, if the average value of the standard deviation of all sample pairs is 0.48, then , is the similarity of the first and the second sample.

[0103] Take sample 1: and sample 2: as an example for calculation. Set sample 1 as the first sample and sample 2 as the second sample:

[0104] ;

[0105] .

[0106] The similarity values between all disturbance samples are obtained by calculating the similarity between the rest of the sample pairs in the same way. After that, the samples are classified according to their correlation strength.

[0107] Then, in order to identify different types of disturbance trends, spectral clustering is performed on all samples. The specific way is as follows: a connection graph is constructed based on the similarity information between all disturbance characteristic samples. Eigenvalue decomposition is performed on the graph, the first few feature vectors are extracted to form a new expression, and then the expression is subjected to mean clustering. The number of clusters is set according to the system operation disturbance classification principle, which is usually set to 3, corresponding to light disturbance, moderate disturbance and strong disturbance regions, i.e. 3.

[0108] There are 5 disturbance samples, sample 1: , sample 2: , sample 3: , sample 4: , sample 5: .

[0109] The square of the Euclidean distance between sample 1 and sample 2 is calculated as:

[0110] ;

[0111] Substitute the similarity formula:

[0112] .

[0113] After calculating the similarity of all samples in turn, the connection relationship between the samples is constructed, and after feature decomposition, it is mapped to a low-dimensional space and divided into three groups of disturbance types using the K-means algorithm. After clustering, the mean value of the disturbance intensity of each group of samples is calculated. If the mean value of the flywheel power change rate normalized value of a certain class of samples is more than 1.3 times the overall mean value of this item of all samples , then this class is defined as a high-intensity disturbance class, for example, if , the threshold is , and if the mean value of a certain class of samples is 0.72, then this class is selected as a strong disturbance section. Finally, the original time stamp, sequence position and associated load number of all samples belonging to the strong disturbance class are extracted, and the start and end positions, concentrated distribution section and corresponding power supply period are recorded to form the disturbance peak clustering distribution information.

[0114] The access time adjustment submodule calls the original access time point of each device in the load delay item set for the disturbance peak clustering distribution information, and maps it with the disturbance duration of the corresponding overlapping section to perform time point backward offset processing and generate a load delay adjustment sequence;

[0115] On the basis of the disturbance peak clustering distribution information, first, all disturbance samples marked as strong disturbance categories are retrieved, each of which contains the original time stamp, the disturbance sequence number it belongs to and the corresponding section start time , all in seconds (s), and the above two time values define the actual duration interval of the disturbance section. Then, perform time matching operation on each load to be accessed in the identified load delay item set, read the load number , original access time , where represents the th load object in the delay item set, in seconds (s). With all the time boundaries of high disturbance sections, adopt closed interval matching logic to determine whether there is a section So that the access time meets the conditions: If the condition is met, it means that the load hits a high-intensity disturbance section and needs to adjust the access time, otherwise the original planned access time remains unchanged. The adjustment range of the access time is determined by the delay amount according to the duration of the disturbance section The duration of the disturbance is recorded as , in seconds (s), and the system presets a fixed buffer delay value , which is used to ensure that the load avoids the impact of the tail of the disturbance, in seconds (s), so the new access time is set as: , where represents the access time after delay adjustment. To illustrate, if the original access time of the load is , it hits the time range of section D1 , the disturbance duration , the buffer value is set to , and the final offset is 0.17 seconds, resulting in a new access time of . If the load hits multiple disturbance sections at the same time, i.e. there are multiple so that falls into multiple intervals, then the sum of and corresponding to the last section (i.e. ) is selected as the offset basis to avoid interference conflicts between consecutive disturbances. After adjustment, all access times need to be verified again for overlap with other loads, with a minimum access time interval of . If multiple loads such as , are found to be adjusted to the same access time , then the offset of is added to the subsequent loads in the original priority order, updated to . In this way, all conflicting load access times are iteratively adjusted to ensure that the power supply system scheduling timing is reasonable and non-overlapping. Finally, each delay load adjustment record should contain the following fields: device number, original access time, hit disturbance section number, adjusted access time, access time offset, and all the above data are arranged in the original order of the load in the delay item set to form the final output result, i.e. the load delay adjustment sequence.

[0116] Please refer to Figure 4 , the scheduling screening module includes:

[0117] The power feature calculation sub-module obtains current data and apparent power data of the air conditioner group in a steady state running period powered by the flywheel, calculates the current mean value and apparent power output standard deviation in the time period respectively, and forms a power supply side reference comparison item to generate air conditioner group running feature parameters;

[0118] In the process of collecting the running information of the air conditioner group in the flywheel power supply channel by the system, first, the continuous running period of the data center in the steady state power supply state is determined, the detection period is generally set to 5 seconds, and isochronous sampling is performed at every 200 milliseconds, a total of 25 groups of sample data are obtained, the instantaneous current value and the apparent power value of the air conditioner group are recorded for each group of samples, the unit is ampere (A) and kilovolt-ampere (kVA) respectively, the current sequence obtained by sampling is 13.2A, 13.5A, 13.3A, 13.7A…, the average operation is performed on the sequence to obtain the current mean value of the air conditioner in the period, for example, the sum is 333A, and the mean value is 13.32A, at the same time, the standard deviation calculation is performed on the apparent power data sequence such as 3.9kVA, 4.0kVA, 3.8kVA, 4.1kVA…, if the standard deviation is 0.12kVA, it means that the power fluctuation degree is the value, the two values are combined to obtain the typical power supply behavior characteristics of the air conditioner group in the current period, which is used as the power supply side reference comparison item when the subsequent device is connected, and is correspondingly recorded as "power supply current reference mean value" and "power output standard deviation" two parameters, which are used to measure the consistency of the power supply load connection condition and the current running state, and finally generate the air conditioner group running feature parameters.

[0119] The device matching comparison sub-module extracts the starting current demand and running power factor of the current to-be-connected operation and maintenance type device according to the air conditioner group running feature parameters, calculates the deviation value with the power supply side reference comparison item respectively, and performs interval intersection determination on the deviation value set of all devices to obtain the device running adaptation interval.

[0120] For the operating characteristic parameters of the air conditioner group, the starting current demand value and the power factor value of all the operation and maintenance type devices to be connected are extracted, with the unit being ampere (A) and dimensionless power factor ratio value. Then, the numerical difference between the starting current of each device and the above-mentioned power supply current average value is calculated to obtain the current deviation between the starting stage and the power supply side current state of the current device, which is defined as the "starting current deviation". Similarly, the running power factor and the average power factor of the power supply side air conditioner group (estimated by the apparent power standard deviation and the equivalent real power) are processed by difference to define the "power factor deviation", for example, if the starting current of device X is 15.0A and the reference of the power supply side is 13.32A, the deviation is 1.68A; if the power factor of the device is 0.94 and the reference of the power supply side is 0.88, the deviation is 0.06. Then, the two deviations of all devices are summarized respectively to form two sets, a mapping diagram is established according to the numerical interval distribution, and interval intersection judgment is performed to determine which devices are within the compatible interval range of the system in terms of current deviation and power factor deviation, for example, the system is set to allow the current deviation not to exceed ±2.0A and the power factor deviation not to exceed ±0.08, then in the above example, device X belongs to the adaptation range, and all devices within the double-parameter intersection interval are filtered out by judging all devices in this way, and the index and parameter identification thereof are output, that is, the identification of the device operation adaptation interval is completed.

[0121] The access load screening submodule extracts the device access attributes of the operation and maintenance type devices in the device operation adaptation interval, sorts the unique identification and operation parameters of each device, and generates an operation device advance scheduling list.

[0122] Among all the devices filtered out in the obtained device operation adaptation interval, the access attribute records of each device are further retrieved, including device number, access priority, installation location identification, access mode category (such as grid-connected / independent), average running time, historical current peak value, etc. Five basic information are matched and extracted through the association system operation and maintenance module database according to the device ID field, a separate access attribute list record is constructed for each device, and the starting current deviation and power factor deviation parameter fields calculated previously are mounted in the record to form a final operation parameter group. This data sorting is performed on all devices in turn to generate a device list for pre-scheduling preparation. Then, the devices are preliminarily sorted according to the device priority and refined according to the running parameter stability (such as the sum of the absolute values of the deviations from small to large), and finally a data structure sequence with unique identification, access parameters, and running evaluation fields is generated, which is output as the identification input of the scheduling system to form the operation and maintenance device advance scheduling list of this round of scheduling period.

[0123] Please refer to Figure 5 The time reconstruction module includes:

[0124] The time offset identification submodule is based on the load delay adjustment sequence and the operation and maintenance equipment advance scheduling list, extracts the time offset value and the original access time of each load, classifies the positive offset and the reverse offset respectively, and generates load access offset structure data;

[0125] On the basis of the load delay adjustment sequence and the operation and maintenance equipment advance scheduling list, each load item is traversed in turn, the time difference value between the original access time and the adjusted access time of the load is extracted, and the offset direction is determined according to the positive and negative of the difference value. If the adjustment time is later than the original access time, it is classified as positive offset, indicating that the load is delayed access; if the adjustment time is earlier than the original access time, it is classified as reverse offset, indicating that the load is accessed in advance. Taking load A as an example, its original access time is 5.00 seconds, and the adjusted time is 5.18 seconds, so the offset value is +0.18 seconds, which is classified into the positive offset set; the original time of load B is 6.50 seconds, and the adjustment is 6.42 seconds, so the offset value is -0.08 seconds, which is classified into the reverse offset set. After completing the classification of all loads, the positive offset loads are uniformly formed into a set of structure data, each of which contains device number, original time, offset value, and direction identifier; the reverse offset also forms a corresponding set, and the structure remains the same. The two sets together constitute the load access offset structure data.

[0126] The access time rearrangement submodule rearranges the original access time sequence of the current main power supply channel according to the load access offset structure data, moves the delayed items to the rear and moves the advanced items to the front, rearranges the access positions according to the adjusted time sequence, and obtains the time axis sequence rearrangement result;

[0127] According to the load access offset structure data, the original access time sequence on the current main power supply channel is once overall reconstructed. The operation process is as follows: first, rearrange all the positive offset loads according to the time point after the offset, remove their original positions in the original time sequence, and insert them into the new time sequence with the new access time; then process the reverse offset loads, extract them from the original time point, and insert them into the sequence with the access time after the advance. Taking three loads as an example, the original time of load C is 7.10 seconds, the offset is +0.15 seconds, and the updated time is 7.25 seconds; the original time of load D is 7.40 seconds, the offset is -0.20 seconds, and the updated time is 7.20 seconds; the original time of load E is 7.30 seconds, which remains unchanged, and the updated sequence becomes: D (7.20 seconds), C (7.25 seconds), E (7.30 seconds), and the new sequence is adjusted according to the time size: D (7.20 seconds), C (7.25 seconds), E (7.30 seconds). Therefore, after sorting, the insertion position of each load is rearranged according to the actual updated time, covering the original access table, and realizing the sequence rearrangement result of the access scheduling sequence on the time axis.

[0128] The time index generation submodule calls the timeline sequence rearrangement result, extracts the new access time point corresponding to each load and the relative position number in the rearranged list, integrates the load unique identifier and the time coordinate information, and generates a load control time mapping index table;

[0129] After completing the sequence rearrangement of the time series, the system further indexes the access time of each load to generate the corresponding time mapping structure. Each load records its updated access time point and calculates its relative position number in the new time series, for example, load X is the first access item in the updated sequence, load Y is the second, and load Z is the third, which correspond to their positions in the sorted timeline. Then, the unique identifier of each load is combined with its updated access time point and sequence number to form complete time-number mapping information. Taking load F as an example, its updated access time is 8.75 seconds, and it is located in the 5th position in the sorted list, so the record item is: load F, access time 8.75 seconds, position number 5. Perform this step on all loads, and the system finally outputs the number, time, and position number of all valid access loads, which constitute a complete load control time mapping index table.

[0130] Please refer to Figure 6 , the instruction update module includes:

[0131] The time node extraction submodule calls all load identifiers and corresponding access time information in the load control time mapping index table, extracts the adjusted start time field, constructs an update sequence according to time order and binds the relay action sequence number, and generates a control instruction time sorting result;

[0132] In the load control time mapping index table, read the load unique identifier and the corresponding new access time information in all record items in turn, take the adjusted start time field of each load as the scheduling reference point of the next stage of the control system, arrange them in ascending order according to the access time value, and construct a set of time update sequence. To illustrate, if there are five loads corresponding to access times of 7.20 seconds, 7.35 seconds, 7.60 seconds, 7.90 seconds, and 8.10 seconds, the system will sort them as follows: 7.20 seconds for the first position, 7.35 seconds for the second position, and so on to the fifth position. After sorting, generate the sequence number of the relay action for each load according to the sorted position, for example, the first load is bound to number 001, the second load is bound to number 002, and so on to the fifth load, which is bound to number 005. Each instruction records three key information fields: load number, access time, and relay instruction number, which constitute a standardized relay operation configuration item, and then form a complete control instruction time sorting result, which will be used to guide the control logic of the access of various loads in the flywheel and battery shared power supply path.

[0133] The control logic rearrangement submodule rearranges the relay trigger logic sequence in the main current channel shared by the flywheel and the battery according to the control instruction time sequence result, performs mapping between the original relay instruction and the updated time sequence, rearranges the control trigger sequence according to time, and generates a relay response rearrangement queue;

[0134] According to the control instruction time sequence result, the original relay control logic deployed in the main current channel shared by the flywheel and the battery is rearranged. First, all existing control instructions and their corresponding execution time points are extracted from the original relay logic, and are compared with the newly sorted instruction time one by one. If there is a difference between the original load relay command and the updated time, the original control entry is marked as invalid. Then, according to the updated instruction sequence, the corresponding relay logic queue is re-established to ensure that each relay action command is strictly covered and executed according to the new access time on the time axis. For example, if the original relay trigger of load A is 7.50 seconds and is now adjusted to 7.20 seconds, the system will cancel the control action at the original 7.50 second period and create a new relay instruction at 7.20 seconds, and the action number is still 001. For example, load B is changed from 8.00 seconds to 7.35 seconds, and the same operation is also completed. All the adjusted relay logic is recombined into a new trigger sequence with time as the main line. This sequence has the characteristics of no overlap, no jump and full update, and finally generates a complete relay response rearrangement queue.

[0135] The trigger flow generation submodule calls the relay response rearrangement queue, writes the relay action and corresponding time sequence into the execution list of the control channel contact synchronously, and combines the adjusted control trigger behavior into a unified task flow to generate a trigger control scheduling instruction flow. This flow is used to coordinate the load access timing of the flywheel energy storage and the battery in the hybrid UPS structure, and to realize the power supply trigger sequence control in the main current channel of the data center.

[0136] After the rearrangement of the relay action is completed, the system enters the trigger control generation phase, calls all relay action items in the entire relay response rearrangement queue and their binding time sequence, and writes each specific action into the relay contact execution list of the main control channel. The list structure is periodically read by the central control unit and the control command is issued. During this writing process, the system ensures that the action time field is synchronized with the clock of the actual control signal trigger to within millisecond level, for example, the first action is triggered at 7.20 seconds, the system will start preparing signal conversion at 7.199 seconds, and the relay is formally closed at 7.200 seconds. Subsequently, all relay operations and their corresponding execution time sequences are arranged into a standardized format and packaged as a unified scheduling task stream. This task stream records the logical order of all control events, the execution object, the command number, and the trigger time field, and is the main scheduling carrier called by the power control system in the running phase, and finally forms the trigger control scheduling instruction stream that can be issued to the execution layer. The instruction stream structure can be directly mapped to the dual-channel power supply regulation logic of the flywheel energy storage and battery in the hybrid UPS structure, and the trigger control nodes are triggered in the main current channel according to the scheduled order to drive the load access action to be sequentially developed in the unified time frame, and to fully cover the power sequence control process when multiple loads in the data center are concurrently switched.

[0137] The above is only a preferred embodiment of the present application, and does not limit the present application in other forms. Any skilled person in the art can modify or change the above disclosed technical content to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made to the above embodiments without departing from the technical solution content of the present application, according to the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. A flywheel energy storage and battery hybrid UPS system for a data center, characterized by, The system comprises: The load identification module obtains the current rotating speed and the current output current of the flywheel and performs synchronous slope analysis to extract the starting current threshold of the current to-be-connected load, screen the devices with starting current threshold greater than the flywheel power supply limit, and mark the load delay item set; The disturbance correction module obtains the flywheel output power change and the battery voltage response change trajectory and performs clustering division, performs backward processing on the original access time point of the devices in the load delay item set according to the division result, and generates a load delay adjustment sequence; The scheduling screening module obtains the current mean value and apparent power output standard deviation of the air conditioning group in the data center in the steady-state running cycle of the flywheel power supply, compares with the current to-be-connected operation and maintenance device, screens the operation and maintenance device early scheduling list; The time reconstruction module reconstructs the time axis of the load in the data center main power supply channel based on the load delay adjustment sequence and the operation and maintenance device early scheduling list, and forms a load control time mapping index table; The instruction updating module updates the trigger control process in the flywheel and battery combined data center main current channel through the load control time mapping index table, and forms a trigger control scheduling instruction stream; The load identification module comprises: The rotating speed and current extraction submodule obtains the continuous sampling data of the current rotating speed and the current output current of the flywheel, calculates the rotating speed change rate and the current change rate in each sampling period, and generates a rotating speed and current joint change rate set; The slope mutation judgment submodule performs trend fitting based on the rotating speed and current joint change rate set by using a local weighted regression algorithm, performs continuous period slope gradient calculation on the fitting residual sequence, and judges whether the residual gradient is continuously rising and exceeds the average background fluctuation amplitude of the trend section to generate a trend slope mutation identification interval; The power supply limit screening submodule extracts the starting current threshold of the load according to the to-be-connected load identified in the trend slope mutation identification interval, compares with the current power supply limit value, screens the target load with all starting current thresholds greater than the power supply limit value, and generates a load delay item set; The disturbance correction module comprises: The power feature extraction submodule obtains the flywheel output power change and the battery voltage response change trajectory, extracts the flywheel power change rate, the battery voltage mutation point time, the response delay of the intersection, and performs normalization processing to generate a disturbance interval feature parameter set; The staggered disturbance clustering submodule performs clustering division on the voltage and current mutation points generated in the alternate power supply process by using a spectral clustering algorithm according to the disturbance interval feature parameter set, extracts the overlapping section with disturbance intensity greater than the average disturbance threshold in the clustering result, and obtains the disturbance peak clustering distribution information; The access time adjustment submodule calls the original access time point of each device in the load delay item set according to the disturbance peak clustering distribution information, maps with the disturbance duration of the corresponding overlapping section, performs time point backward offset processing, and generates a load delay adjustment sequence.

2. The flywheel energy storage and battery hybrid UPS system for a data center of claim 1, wherein, The load delay item set comprises devices whose starting current threshold exceeds the flywheel power supply limit, to-be-connected loads that need to be delayed to be connected, and load targets identified through synchronization slope analysis, the load delay adjustment sequence comprises an offset value of an original connection time point, a disturbance grouping identifier corresponding to a cluster division, and a load mapping item corresponding to a power disturbance peak, the operation and maintenance device advance scheduling list comprises devices that meet the current mean value requirement, devices that meet the apparent power output standard deviation condition, and operation and maintenance devices consistent with the air conditioner group operation characteristics, the load regulation time mapping index table comprises a connection time update item in the operation and maintenance device advance scheduling list, a connection time adjustment item in the load delay adjustment sequence, and a time axis arrangement item after unified mapping, and the trigger control scheduling instruction stream comprises a load connection sequence update instruction, a trigger control signal in the main current channel, and a scheduling trigger action corresponding to the time mapping index.

3. The flywheel energy storage and battery hybrid UPS system for a data center of claim 1, wherein, The scheduling screening module comprises: The power characteristic calculation submodule obtains current data and apparent power data of the air conditioner group in the data center in a steady-state operation period of flywheel power supply, calculates the current mean value and the apparent power output standard deviation in the time period respectively, and forms a power supply side reference comparison item to generate air conditioner group operation characteristic parameters; The device matching comparison submodule extracts the starting current demand and operating power factor of the current to-be-connected operation and maintenance devices according to the air conditioner group operation characteristic parameters, calculates deviation values respectively with the power supply side reference comparison item, and performs interval intersection determination on the deviation value set of all devices to obtain a device operation adaptation interval; The connected load screening submodule extracts device connection attributes of the operation and maintenance devices in the device operation adaptation interval, sorts the unique identifier and operating parameters of each device, and generates an operation and maintenance device advance scheduling list.

4. The flywheel energy storage and battery hybrid UPS system for a data center of claim 3, wherein, The time reconstruction module comprises: The time offset identification submodule extracts the time offset value and the original connection time of each load based on the load delay adjustment sequence and the operation and maintenance device advance scheduling list, classifies the forward offset and the reverse offset respectively, and generates load connection offset structure data; The connection time rearrangement submodule replaces the original connection time line of the current main power supply channel as a whole according to the load connection offset structure data, moves the delay items backward and moves the advance items forward, rearranges the connection positions according to the adjusted time sequence, and obtains a time axis sequence rearrangement result; The time index generation submodule calls the time axis sequence rearrangement result, extracts the new connection time point corresponding to each load and the relative position number in the rearrangement column, integrates the load unique identifier and the time coordinate information, and generates a load regulation time mapping index table.

5. The flywheel energy storage and battery hybrid UPS system for a data center of claim 4, wherein, The instruction update module comprises: The time node extraction submodule calls all load identifiers and corresponding connection time information in the load regulation time mapping index table, extracts the adjusted starting time field, constructs an update sequence according to the time sequence and binds a relay action sequence number, and generates a control instruction time sorting result; The control logic rearrangement submodule rearranges the control logic according to the control instruction time sequence result, replaces the relay trigger logic sequence in the main current channel shared by the flywheel and the battery as a whole, maps the original relay instruction to the updated time sequence, rearranges the control trigger sequence according to the time sequence, and generates a relay response rearrangement queue. The trigger flow generation submodule calls the relay response rearrangement queue, writes the relay action and the corresponding time sequence into the execution list of the control channel contact synchronously, combines all the adjusted control trigger behaviors into a unified task flow, generates a trigger control scheduling instruction flow, and coordinates the load access time sequence of the flywheel energy storage and the battery in the hybrid UPS structure, so as to realize the power supply trigger sequence control in the main current channel of the data center.

Images