A multi-phase ups battery parameter collaborative monitoring and fault prediction system

Abstract

The application discloses a multi-phase UPS battery parameter cooperative monitoring and fault prediction system and relates to the technical field of batteries, comprising a dynamic synchronous acquisition subsystem and a coupling correction subsystem. A plurality of sensor measuring points are installed on the DC bus branch between the battery pack and the inverter to determine the transient pull characteristics corresponding to each phase after disturbance occurs. Through correlation coefficient analysis, the coupling degree between phases on the corresponding parameters in the transient pull characteristics is obtained, and the disturbance residual is obtained by using the CP decomposition method. Based on the disturbance residual, a multi-stage abnormality measurement sequence is constructed. Based on the normal coupling caused by the natural energy migration of the UPS, the multi-stage abnormality measurement sequence is dynamically corrected, and it is judged whether the backtracking instruction is triggered. If the backtracking instruction is triggered, the sampling period of the parameters of the corresponding phase is dynamically adjusted.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery, in particular to a multi-phase UPS battery parameter cooperative monitoring and fault prediction system. BACKGROUND

[0002] In the existing multi-phase UPS system, a single-point voltage or current sampling method combined with a fixed threshold judgment is generally used to detect the health status of the battery pack. Some systems introduce a simple statistical model based on the remaining capacity SOC and the health status SOH in order to realize online estimation of the battery status during UPS load operation. However, such methods often only make threshold judgments on a single electrical parameter, and it is difficult to comprehensively capture the transient pull characteristics caused by the UPS multi-phase inverter output under different load conversions. For example, under the condition of normal three-phase balanced load of the UPS, the voltage and current at the battery end may fluctuate greatly at the same time in a short time, but this is often due to the natural energy rebalancing of the UPS, and is not caused by abnormality of the battery degradation. However, the existing method lacks in-depth analysis of the coupling behavior of the UPS multi-phase, and is prone to false positives in this scenario, reducing the reliability of the monitoring. SUMMARY

[0003] In view of the above problems existing in the prior art, the present application provides a multi-phase UPS battery parameter cooperative monitoring and fault prediction system.

[0004] The present application provides a multi-phase UPS battery parameter cooperative monitoring and fault prediction system, which comprises:

[0005] A dynamic synchronous acquisition subsystem is installed on the DC bus branch between the battery pack and the inverter to determine the transient pull characteristics of each phase after the disturbance occurs. Through correlation coefficient analysis, the coupling degree between each phase on the corresponding parameters in the transient pull characteristics is obtained, and the disturbance residual is obtained by using the CP decomposition method.

[0006] A coupling correction subsystem is based on the disturbance residual to construct a multi-stage abnormality measurement sequence. Based on the normal coupling caused by the natural energy migration of the UPS multi-phase, the multi-stage abnormality measurement sequence is dynamically corrected, and it is judged whether the backtracking instruction is triggered. If the backtracking instruction is triggered, the sampling period of the corresponding phase parameter is dynamically adjusted.

[0007] Optionally, the dynamic synchronous acquisition subsystem comprises an acquisition module, an offset module and a decomposition extraction module.

[0008] The collection module records the transient characteristics corresponding to each phase after the disturbance occurs, including the battery output voltage, battery output current and complex impedance corresponding to each phase after the disturbance occurs, by connecting a programmable electronic load in parallel with the multi-phase output end of the UPS system and installing multiple sensor measuring points on the DC bus branch between the battery pack and the inverter when the multi-phase UPS system is under normal load conditions.

[0009] Optionally, the offset module determines the transient pull characteristics corresponding to each phase after the disturbance occurs, including the transient drop amplitude of the battery voltage corresponding to each phase, the current peak value change and the impedance trajectory, based on the transient characteristics, and tensorizes the transient pull characteristics corresponding to each phase to form a third-order tensor. Based on the third-order tensor, the coupling degree between phases on the corresponding parameters in the transient pull characteristics is obtained through correlation coefficient analysis. The coupling degree is used to capture the cross disturbance of the transient pull characteristics between different phases, and the coupling degree between phases on the corresponding parameters in the transient pull characteristics is used as a supplementary attribute of the third-order tensor.

[0010] Optionally, the decomposition and extraction module decomposes the third-order tensor into a low-rank ground state and a disturbance residual using a CP decomposition method, wherein the low-rank ground state represents the cooperative behavior mode under normal load conversion of the multi-phase UPS, and the disturbance residual is used to explicitly represent abnormal single-phase or cross-phase degradation signs.

[0011] Optionally, the coupling correction subsystem includes an anomaly identification module, a correction module, a backtracking module and a correction module.

[0012] The anomaly identification module obtains a disturbance residual vector by expanding the disturbance residual in the time dimension, analyzes the disturbance residual at different time points based on the disturbance residual vector, and obtains the abnormal strength at each time point by calculating the Frobenius norm of the disturbance residual vector. In order to capture the abnormal accumulation trend in different time periods, a sliding window is set, and the average abnormal strength is calculated in each sliding window to form a multi-stage abnormality metric sequence.

[0013] Optionally, the correction module calculates the coupling degree between the phases of the UPS when facing the same parameters in the same sliding window, and obtains the average value to form a phase linkage index, in order to further reduce the risk of misjudging normal coupling as abnormal due to natural energy migration of the UPS multi-phase. Based on the numerical value of the phase linkage index, the average abnormal strength in the corresponding sliding window is corrected in stages to obtain the corrected average abnormal strength, and the multi-stage abnormality metric sequence is updated.

[0014] Optionally, the backtracking module is configured to set an abnormal threshold value, and if the corrected average abnormal intensity of a plurality of sliding windows in the updated multi-stage abnormality metric sequence exceeds the abnormal threshold value, a backtracking instruction is triggered to backtrack the fluctuation of the updated multi-stage abnormality metric sequence to the decomposition extraction module, and the decomposition rank value of the original CP decomposition is adaptively increased by 1 each time the backtracking is performed, so that the decomposed low-rank ground state is closer to the current actual operating state of the UPS system, until the backtracking instruction is not triggered, the backtracking action is stopped, and the final low-rank ground state and disturbance residual error are obtained.

[0015] Optionally, the correction module is configured to, based on the final disturbance residual error, take an average absolute residual error of each parameter of each phase in a time dimension to form a sensitivity matrix, each element in the sensitivity matrix refers to an average absolute residual error of each parameter of a corresponding phase in a time dimension, each element in the sensitivity matrix is used to reflect an average abnormal concentration of a corresponding parameter of a corresponding phase, and if a corresponding element in the sensitivity matrix exceeds a preset sensitivity threshold value, a sampling period is dynamically adjusted, otherwise no dynamic adjustment is performed.

[0016] Optionally, the dynamic adjustment of the sampling period is performed in the following manner: , wherein, is a sampling period of the i-th parameter of the j-th phase, is a reference sampling period, is an adjustment coefficient, is a reference sampling period, is an adjustment coefficient, is an average absolute residual error of the i-th parameter of the j-th phase in a time dimension. The beneficial effects of the present application are as follows:

[0017]

[0018] ​​The application realizes dynamic synchronous collection of key parameters of a multi-phase UPS battery by arranging a multi-point sensor measuring point on a DC bus branch between a battery pack and a UPS inverter, and artificially introduces a controllable perturbation in UPS operation, so that the transient pull characteristics corresponding to each phase can be accurately extracted under the condition of multi-phase load disturbance. By using the correlation coefficient analysis method in the transient pull characteristics, the parameter linkage relationship between each phase under the natural coupling state of the UPS multi-phase is captured, and the multi-dimensional disturbance residual is obtained by combining CP decomposition, which enhances the dynamic description ability of the coupling behavior of the UPS multi-phase operation. Further, through the coupling correction subsystem, a multi-stage abnormality measurement sequence is constructed based on the obtained disturbance residual, and the coupling mode of the UPS in normal multi-phase energy migration is used to dynamically correct the abnormality measurement sequence, which can effectively filter false abnormal signals caused by normal load migration of the UPS, further reducing the false alarm probability. At the same time, by monitoring whether the corrected abnormality measurement sequence continuously exceeds the set threshold and triggering the backtracking mechanism, the sampling period of the corresponding phase and parameter is automatically and dynamically adjusted, so that the system sampling resources can be concentrated on sensitive or potentially degraded phases and parameters, realizing adaptive optimization of the collection strategy and improving the flexibility of monitoring. Therefore, compared with the existing single-point voltage threshold monitoring or single SOC estimation method, the application can further improve the health monitoring accuracy and predictive diagnosis ability of the multi-phase battery under the natural energy migration scene of the UPS multi-phase, reduce the misjudgment and resource waste caused by unidentified multi-phase coupling, and improve the operation safety and scientificity of maintenance decision of the UPS power supply system.

[0019] The application can dynamically judge the multi-dimensional abnormal signs of the UPS multi-phase battery after transient disturbance in stages and from multiple angles by sequentially arranging an abnormality recognition module, a correction module, a backtracking module and a correction module in the coupling correction subsystem. First, the abnormality recognition module expands the disturbance residual along the time dimension, extracts the disturbance residual vector, and obtains the abnormal strength at each time point by using the Frobenius norm calculation method, so as to form a multi-stage abnormality measurement sequence in a sliding window and accurately reflect the abnormal accumulation of the UPS multi-phase battery in different time periods. Further, the correction module is used to statistically analyze the coupling degree of each phase of the UPS under the same parameter in the same sliding window, and the mean value is obtained to form a phase linkage index, and the abnormality measurement sequence is corrected in stages based on the index, so as to dynamically adaptively filter the normal coupling behavior caused by the natural energy migration of the UPS multi-phase, reduce the probability of misjudging the normal cooperative behavior as a degraded abnormality, thereby not only enhancing the pertinence of abnormal detection, but also effectively avoiding false alarms caused by the normal load migration coupling of the UPS, and improving the stability of the UPS monitoring system and the reliability of the battery health state judgment. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application.

[0021] Figure 1 A module diagram of a multi-phase UPS battery parameter cooperative monitoring and fault prediction system of the application. DETAILED DESCRIPTION

[0022] In order to make the objectives, technical solutions and advantages of the application clearer, the technical solutions in the application will be clearly and completely described below in combination with the drawings in the application. Obviously, the described embodiments are some embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the protection scope of the application.

[0023] As shown in Figure 1 the embodiment of the application, a multi-phase UPS battery parameter cooperative monitoring and fault prediction system is provided, which comprises,

[0024] a dynamic synchronous acquisition subsystem, a plurality of sensor measuring points are installed on the DC bus branches between the battery pack and the inverter to determine the transient pull characteristics corresponding to each phase after the disturbance occurs, the coupling degrees between the phases on the corresponding parameters in the transient pull characteristics are obtained through correlation coefficient analysis, and the disturbance residual is obtained by using CP decomposition.

[0025] a coupling correction subsystem, based on the disturbance residual, a multi-stage abnormality measurement sequence is constructed, based on the normal coupling caused by the natural energy migration of the UPS, the multi-stage abnormality measurement sequence is dynamically corrected, and it is judged whether the backtracking instruction is triggered. If the backtracking instruction is triggered, the sampling period of the parameters of the corresponding phase is dynamically adjusted.

[0026] UPS refers to the entire uninterruptible power supply device system, which usually includes:

[0027] a rectifier (AC input is rectified into DC);

[0028] a DC bus (connecting the battery and the rectifier / inverter);

[0029] a battery pack (energy storage);

[0030] an inverter (DC is inverted into AC again);

[0031] an output filter, a bypass switch, a monitoring module, etc.

[0032] In the embodiment of the present application, the dynamic synchronous acquisition subsystem can capture the transient pull behavior of the multi-phase UPS during the disturbance process. Specifically, the dynamic synchronous acquisition subsystem records the transient response of the battery output of each phase after the disturbance event, such as step load switching or inverter inter-phase power balance adjustment, by deploying multiple voltage, current and impedance sensor measurement points on the DC bus branch between the battery pack and the UPS inverter. For example, when the UPS is under normal load, a programmable electronic load is connected in parallel at the output end to realize fast load disturbance, and the battery end voltage of phase A is observed to drop by 0.3V in a short time, while phases B and C simultaneously experience transient drops of 0.25V and 0.27V. At this time, the system can intuitively identify the transient pull characteristics in the normal energy balance process based on synchronous sampling data and phase comparison, avoiding misjudgment of the voltage drop of phase A as single-phase degradation.

[0033] During this process, the transient pull characteristics are used to represent the rapid shift behavior of the voltage, current and impedance of each phase battery output in a short period of load disturbance.

[0034] By coupling degree calculation and CP decomposition, the normal coupling mode in the multi-phase UPS system is effectively extracted and the abnormal disturbance residual is explicitly distinguished. Specifically, in the dynamic synchronous acquisition subsystem, the coupling degree between different phases under the same parameter can be obtained by using the correlation coefficient analysis of the transient pull characteristics, thereby quantifying the normal multi-phase energy transfer phenomenon. For example, after an automatic inverter balancing event, the Pearson correlation coefficient is calculated for the phase offset voltage waveform of phases A and B to obtain a coupling degree of 0.93, indicating that the process is mainly the natural load power transfer within the UPS. Subsequently, these inter-phase coupling degrees are used as tensor supplementary attributes, and the third-order tensor is decomposed into a low-rank ground state and a disturbance residual using CP decomposition, ensuring that the normal multi-phase cooperative behavior of the UPS is completely included in the ground state mode, and only the part that exceeds the normal coupling offset range is retained in the disturbance residual.

[0035] Here, CP decomposition is used to automatically extract the main cooperative behavior (low-rank ground state) of the UPS system from the multi-dimensional time series tensor, and the disturbance residual explicitly identifies the non-coupling behavior characteristics that may be related to degradation, loss and local faults.

[0036] The dynamic anomaly degree correction based on the natural multi-phase coupling of the UPS and the adaptive sampling adjustment are realized by the coupling correction subsystem, and the false positives are further reduced, and specifically, in the coupling correction subsystem, first, a multi-stage anomaly degree sequence is constructed based on the disturbance residual, and a phase linkage index is formed by using the average coupling degree of the different phases of the UPS under the same parameters, and the anomaly value is dynamically corrected. For example, when it is detected that the abnormal strength in the continuous sliding window is maintained at a high level, if the phase linkage index is still high (such as 0.89), it indicates that the multi-phase coupling of the UPS is still robust, and the system avoids false positives by correcting the anomaly degree. If the multi-stage anomaly degree still exceeds the threshold value after correction, a backtracking instruction is triggered, the CP decomposition rank value is dynamically increased (for example, from 3 to 4), and the tensor ground state is further fitted to the new normal coupling mode of the UPS, so as to reduce the disturbance residual and reduce the false positive rate.

[0037] Finally, based on the finally stabilized disturbance residual, the system takes the average absolute residual of each parameter of each phase in the time dimension to form a sensitivity matrix, which is used to guide the sampling period adjustment in real time. For example, if the matrix shows that the average residual of the phase shift voltage of phase A is 0.15 and the average residual of the phase shift voltage of phase C is only 0.03, the system automatically adjusts the sampling period of the phase shift voltage of phase A to 60% of the reference period, that is, higher frequency sampling is used to improve the monitoring resolution.

[0038] The coupling correction is used to ensure that the normal multi-phase load self-balancing of the UPS will not be incorrectly detected as an anomaly;

[0039] The backtracking instruction represents an automatic adjustment process of the tensor rank value based on the anomaly feedback, and forms a self-optimizing closed loop.

[0040] Therefore, the present application gradually realizes the complete closed loop process from anomaly feature detection to tensor model updating, and then to sampling resource focusing optimization by multi-dimensional and multi-phase data acquisition between the battery output and the UPS inverter, combined with correlation coefficient analysis, CP decomposition, anomaly correction and dynamic sampling adjustment, effectively overcomes the problems of easy false positives, easy omission and lack of adaptation to natural coupling behavior in the existing UPS multi-phase battery monitoring, and greatly improves the health prediction accuracy and operation safety of the multi-phase UPS battery.

[0041] Further, the dynamic synchronous acquisition subsystem includes an acquisition module, an offset module and a decomposition extraction module.

[0042] The collection module records the transient characteristics corresponding to each phase after the disturbance occurs, including the battery output voltage, battery output current and complex impedance corresponding to each phase after the disturbance occurs, by connecting a programmable electronic load in parallel with the multi-phase output end (close to the inverter output side) of the UPS system and installing multiple sensor measuring points on the DC bus branch between the battery pack and the inverter, when the multi-phase UPS system is in a normal load condition.

[0043] It should be noted that in many UPS system designs, the inverters of all phases are connected to the same battery bus, but even if the same battery bus is used, due to the slight difference in transient absorption current of inverters of different phases, this will cause a slight voltage and chemical potential difference in battery output when the load of three phases or multi-phase changes rapidly. Long-term accumulation of these small differences will cause some batteries to age faster and the load fluctuation of a single phase to be amplified, affecting the service life of the entire machine. Therefore, in UPS monitoring, it is necessary to collect data separately for each phase.

[0044] The multi-phase output end of the UPS system refers to the AC output end after inversion of the entire UPS system, which is usually a multi-phase AC output;

[0045] The multiple sensor measuring points are used to capture the subtle changes in voltage, current and impedance over time in real time;

[0046] The programmable electronic load connected in parallel with the multi-phase output end (close to the inverter output side) of the UPS system is used to temporarily increase or decrease the load power at the normal output end of the UPS, so as to stimulate the dynamic response of the UPS and the battery;

[0047] The offset module determines the transient pull characteristics corresponding to each phase after the disturbance occurs according to the transient characteristics, including the transient drop amplitude of the battery voltage corresponding to each phase, the peak value change of the current and the impedance trajectory;

[0048] The transient drop amplitude of the battery voltage refers to the drop of the battery output voltage due to the rapid increase of the current after a sudden load change (such as step increase) at the UPS output end, which is recorded from the steady state to the transient state;

[0049] The peak value change of the battery current refers to the instantaneous jump of the battery output current from the original steady state current when the same load step event occurs, to record the difference between the instantaneous maximum value and the original value;

[0050] The impedance trajectory refers to the change of the complex impedance of the battery at multiple frequencies;

[0051] In a multi-phase UPS (for example, three-phase), each phase has a corresponding inverter bridge arm and filter unit. When the external load occurs a step change (for example, suddenly from 0% to 50%), or a harmonic disturbance occurs, the current pull (transient charge and discharge peak) of each phase of the inverter to the battery and the transient voltage drop (voltage drop amplitude) will be different, which is called transient pull characteristic.

[0052] The transient pull characteristics corresponding to each phase are tensorized to form a third-order tensor. Based on the third-order tensor, the coupling degree between each phase on the corresponding parameters in the transient pull characteristic is obtained through correlation coefficient analysis. The coupling degree is used to capture the cross disturbance of the transient pull characteristics between different phases, and the coupling degree between each phase on the corresponding parameters in the transient pull characteristic is used as a supplementary attribute of the third-order tensor. Here, the coupling degree is written into the supplementary attribute of the tensor, and is not directly added to the tensor value. In CP decomposition, the coupling degree is taken as one of the optimization objectives, for example, the phases with high coupling are decomposed into similar basis states.

[0053] In multi-phase UPS monitoring, the third-order tensor refers to that the sampling data has three dimensions: phase, parameter and time;

[0054] The coupling degree between each phase on the corresponding parameters in the transient pull characteristic is obtained in the following manner:

[0055] , wherein, is the coupling degree between phase A and phase B on the i-th parameter in the transient pull characteristic , is the correlation coefficient for capturing the linkage of two-phase time series, and are the i-th parameter values of phase A and phase B corresponding to the transient pull characteristic , are phase numbers; In the embodiment of the application, the acquisition module acquires the high time resolution data matrix of multi-phase and multi-electric parameters by connecting a programmable electronic load in parallel to the multi-phase output end (i.e. the multi-phase AC output after inverter output) of the UPS system when the UPS is in normal operation, for example, performing 10% load step change at the millisecond level, and arranging multiple sensor measuring points on the DC bus branch between the battery pack and the inverter to acquire the small fluctuations of the battery output voltage, current and complex impedance over time in real time.

[0056] In the embodiment of the application, the acquisition module acquires the high time resolution data matrix of multi-phase and multi-electric parameters by connecting a programmable electronic load in parallel to the multi-phase output end (i.e. the multi-phase AC output after inverter output) of the UPS system when the UPS is in normal operation, for example, performing 10% load step change at the millisecond level, and arranging multiple sensor measuring points on the DC bus branch between the battery pack and the inverter to acquire the small fluctuations of the battery output voltage, current and complex impedance over time in real time.

[0057] ​Here, the multi-phase output of the UPS is used to artificially introduce small perturbations to reveal the battery response, while the sensor points on the DC bus branches are used to capture the projection effects of these perturbations on the voltage, current, and impedance at the battery end.

[0058] After obtaining the perturbation data, the offset module extracts the transient pull characteristics corresponding to each phase within a short time window (e.g., 5 ms) after the perturbation occurs based on transient characteristic analysis.

[0059] The transient pull characteristics refer to the dynamic response shift of different phase batteries after the UPS load perturbation.

[0060] The offset module tensorizes these multi-phase, multi-parameter transient pull characteristics to form a third-order tensor, where the three dimensions correspond to phase, parameter, and time sequence, respectively, to comprehensively characterize the coupling behavior of multi-phase UPS batteries under perturbation. On this basis, the offset module analyzes the correlation coefficients of each parameter slice of the transient tensor, such as calculating the correlation coefficient of phase A and phase B impedance within the same transient pull window to be 0.92, indicating that there is strong coupling between the two phases; if the voltage transient correlation coefficient of phase A and phase C is only 0.48, it means that the linkage of the two phases under this parameter is weak. This coupling degree is used to capture the cross-response of different phases when facing the same perturbation, and can identify the linkage mode brought by the normal three-phase energy rebalancing of the UPS.

[0061] Preferably, the present application uses these coupling degree indicators as supplementary attributes of the third-order tensor, i.e., the transient coupling relationship of different phases under each parameter is saved in the tensor data structure, so that the subsequent CP decomposition can more effectively decompose the ground state and residual on the basis of preserving the coupling behavior characteristics of the UPS. Thus, by collecting high-precision, real-time perturbation response data in the acquisition module, the offset module extracts multi-parameter transient pull characteristics and analyzes the coupling degree, and the decomposition extraction module uses CP decomposition to explicitly extract the three-phase cooperative behavior ground state and abnormal disturbance residual of the UPS, the present application not only significantly improves the accuracy of describing the transient response of the battery caused by the UPS multi-phase load conversion, but also reduces the false positive rate caused by the natural energy migration of the UPS in subsequent abnormality measurement and backtracking correction. For example, the traditional alarm threshold method based on single-point voltage drop often causes false positives when the normal A and B phases of the UPS load migrate, while the joint analysis based on coupling degree and tensorization processing of the present application can correctly identify such natural coupling behavior and reduce false alarms.

[0062] Further, the decomposition extraction module decomposes the third-order tensor into a low-rank ground state and a perturbation residual by using a CP decomposition method, wherein the low-rank ground state represents a cooperative behavior mode under normal load conversion of the multi-phase UPS, and the perturbation residual is used to explicitly represent abnormal single-phase or cross-phase degradation signs. The decomposition structure not only forms a normal coupling mode fingerprint of the current multi-phase UPS, but also provides a preliminary reference for later dynamic prediction.

[0063] The CP decomposition is to represent the tensor as: , wherein, is a third-order tensor, E is a decomposition rank value of the CP decomposition, r is a rank value number, is a mode vector of the rth group on the first dimension (for example, a phase of the UPS) of the tensor, is a mode vector of the rth group on the second dimension (for example, a parameter: voltage, current, impedance) of the tensor, is a mode vector of the rth group on the third dimension (for example, time) of the tensor, is an outer product of vectors, that is: A rank 1 tensor (a three-dimensional matrix) is generated, and an element of the rank 1 tensor is In other words, this is a three-vector outer product, which constitutes a “volume block”.

[0064] The formula actually represents that the third-order tensor is approximately represented as a superposition of E rank 1 tensors (that is, three-vector outer products), each of which is composed of a group of characteristic mode vectors in the phase dimension, the parameter dimension, and the time dimension. By accumulating the outer products of all these modes, the original tensor data can be well reconstructed.

[0065] The third-order tensor is decomposed into a low-rank ground state and a perturbation residual by using the CP decomposition method, and the specific steps are as follows:

[0066] (1) Initialization: first, set the decomposition rank value, for example, E=3 (indicating that 3 groups of main modes are used to describe the tensor), and randomly initialize three mode matrices corresponding to the three directions of phase, parameter, and time;

[0067] (2) Alternating Least Squares (ALS) loop update:

[0068] The mode matrix of the UPS acquisition parameter direction (voltage, current, impedance) and the mode matrix of the time direction are taken as known quantities, and then the optimal mode matrix of the UPS phase direction is obtained by using the least square method, so that the overall tensor reconstruction error is minimized, and the next round of the mode matrix of the UPS phase direction and the mode matrix of the UPS acquisition parameter direction (voltage, current, impedance) are taken as known quantities, and then the optimal mode matrix of the time direction is obtained by using the least square method, and the loop iteration is continued until convergence;

[0069] (3) After decomposition, three groups of vectors are: phase main mode, parameter main mode and time main mode, and the tensor reconstructed by combining them is the low-rank ground state;

[0070] (4) Calculate the perturbation residual: calculate the perturbation residual by subtracting the third-order tensor from the low-rank ground state.

[0071] Tensor decomposition can extract the main coordination mode (low-rank ground state) from a large amount of coupled data, and restore the complex data to: normal coupling behavior + abnormal perturbation residual;

[0072] In the embodiment of the application, the decomposition and extraction module further decomposes the third-order tensor into a low-rank ground state and a perturbation residual through CP decomposition, wherein: the low-rank ground state is used to describe the coordination behavior mode of the UPS multi-phase system under normal load migration or balanced distribution, for example, the natural energy rebalancing feature of the UPS internal three-phase inverter will be completely absorbed and represented by the low-rank ground state. The perturbation residual is used to explicitly reveal abnormal single-phase or cross-phase degradation signs beyond the normal UPS energy migration range, for example, if the internal resistance of the C-phase battery suddenly changes, causing the C-phase current to abnormally rise in a short time, this abnormality will mainly be reflected in the residual tensor. In the above manner, the application can effectively decouple the normal coordination energy migration mode of the UPS multi-phase system and the potential abnormal degradation signs, thereby improving the resolution of the UPS multi-phase battery health state monitoring. Compared with the traditional method of relying only on single-channel voltage and current fluctuations, the application can effectively avoid false positives under normal UPS three-phase migration conditions, and accurately identify and predict the early stage of UPS single-phase degradation, which helps to prolong the service life of the battery pack, reduce maintenance costs, and enhance the overall safety of the power supply system.

[0073] The third-order tensor is used to uniformly encode the response of the UPS multi-phase multi-parameter over time in a multi-dimensional space, capturing complex spatiotemporal coupling modes.

[0074] CP decomposition is used to separate the tensor into a normal ground state and an abnormal residual, i.e., to decompose complex behavior into normal coordination and super-limit abnormality.

[0075] The low-rank ground state is used to express the natural coupling behavior of the UPS normal load conversion, avoiding misjudgment. The perturbation residual is used to highlight the potential degradation points that exceed the normal UPS self-balancing capability.

[0076] Further, the coupling correction subsystem includes an abnormality identification module, a correction module, a backtracking module and a correction module.

[0077] The abnormality identification module obtains a disturbance residual vector by unfolding the disturbance residual in the time dimension, analyzes the disturbance residual at different time points based on the disturbance residual vector, obtains the abnormality strength at each time point by calculating the Frobenius norm of the disturbance residual vector, sets a sliding window, and calculates the average abnormality strength on each sliding window to form a multi-stage abnormality measurement sequence.

[0078] The multi-stage abnormality measurement sequence is used for describing the time trend of abnormality evolution.

[0079] The correction module calculates the coupling degree between each phase of the UPS in the same sliding window when facing the same parameters, and obtains the average value to form a phase linkage index, such as: if it is a three-phase UPS (A phase, B phase, C phase), the coupling degree between the A phase and the B phase, the coupling degree between the A phase and the C phase, and the coupling degree between the B phase and the C phase in the same sliding window when facing the same parameters are obtained, and the average value is obtained to obtain the phase linkage index; based on the value of the phase linkage index, the average abnormality strength on the corresponding sliding window is corrected in stages to obtain the corrected average abnormality strength, and the multi-stage abnormality measurement sequence is updated, further avoiding that the normal coupling behavior is misjudged as degradation.

[0080] The phase linkage index is used for dynamically quantifying the closeness of multi-phase coupling, so as to avoid misjudging the normal coupling behavior as abnormality.

[0081] The corrected average abnormality strength is obtained by the following formula: , wherein, is the corrected average abnormality strength on the i-th sliding window, is the average abnormality strength on the i-th sliding window, is the average abnormality strength on the i-th sliding window, is the average abnormality strength on the i-th sliding window, is a coefficient for adjusting the amplification degree, used for amplifying or reducing the influence degree of coupling correction, and is a positive value (such as 0.5-2), is the phase linkage index; if the phase linkage index tends to 1, it indicates that the multi-phase coupling is very close, and it is likely that the residual comes from normal migration, so it will be corrected and weakened, otherwise, the abnormality measurement will be amplified.

[0082] In the embodiment of the application, the coupling correction subsystem is further used for performing deep and multi-stage dynamic abnormality correction analysis on the disturbance residual extracted in the process of monitoring the UPS multi-phase battery. The coupling correction subsystem includes an abnormality identification module, a correction module, a backtracking module and a correction module, which cooperate with each other to form a closed-loop optimization logic chain.

[0083] Specifically, the anomaly identification module first expands the perturbation residual tensor in the time dimension to obtain a perturbation residual vector evolving over time. For example, when the UPS system collects multiple time series slice data within a 30-second monitoring period, the module converts the tensor residual slice at each time into a vector and calculates its Frobenius norm to measure the abnormal deviation amplitude of all phases and parameters at that time point. The Frobenius norm is used as a unified scalar measurement of multi-dimensional residual to reflect the overall abnormal strength of the UPS multi-phase battery at that time point.

[0084] To capture the abnormal accumulation trend in different time periods, the anomaly identification module further uses a sliding window to locally calculate the abnormal strength values, for example, a sliding window of length 5 can be selected, and the average abnormal strength is calculated window by window to form a multi-stage abnormality measurement sequence, which helps to identify potential long-term deviation patterns of the UPS. However, to avoid misjudging short-term coordinated fluctuations caused by normal energy migration of the UPS multi-phase battery as degradation signals, the correction module will analyze the correlation coefficients of the fluctuation sequences of the UPS multi-phase (such as phase A, phase B, and phase C) facing the same parameters (such as voltage or impedance) within the same sliding window, obtain the coupling degree between, for example, phase A and phase B, phase A and phase C, and phase B and phase C, and calculate the average value to form a so-called phase linkage index, which quantifies the degree of coordination between the UPS multi-phase within this window segment. The closer the value is to 1, the more consistent the fluctuations of the UPS three-phase are on this parameter, and the more likely it is a normal energy migration behavior.

[0085] Based on this, the correction module dynamically corrects the average abnormal strength within the sliding window according to the phase linkage index, for example, when the linkage index is close to 1, the abnormal value amplitude of the window is reduced through a correction formula, and vice versa if the linkage index is low, the abnormal value is retained or even amplified, thereby obtaining a more reasonable corrected multi-stage abnormality measurement sequence. This mechanism can effectively avoid false abnormalities caused by misjudging normal multi-phase energy balance adjustment of the UPS as degradation. Further, the backtracking module will monitor the updated multi-stage abnormality measurement sequence, and when it detects that the corrected abnormal strength of, for example, three consecutive sliding windows exceeds a preset threshold, it will trigger the backtracking mechanism and automatically return to the decomposition extraction module by adaptively increasing the decomposition rank value of the original CP decomposition by 1 to enhance the representation ability of the tensor ground state. The ground state can more fully absorb the current natural multi-phase coupling characteristics of the UPS, further reduce the residual amplitude, and significantly reduce the false alarm risk.

[0086] Therefore, through the above abnormality identification, correction based on the phase linkage index, and threshold-driven backtracking decomposition optimization process, the application can dynamically and accurately distinguish between the natural coupling behavior of the UPS multi-phase and potential degradation abnormalities, avoiding the frequent false positives caused by the traditional monitoring method without considering the three-phase cooperation, and timely capturing the abnormal evolution trend that really needs to be focused on, greatly improving the predictive maintenance value of the UPS battery and the power supply safety guarantee capability.

[0087] The disturbance residual vector is used to quantify the overall abnormal deviation of the UPS multi-phase multi-parameter at each time point.

[0088] The Frobenius norm compresses the multi-dimensional disturbance vector into a single scalar, facilitating analysis on the time series.

[0089] The sliding window is used to extract the multi-stage abnormal accumulation trend to form an abnormality measurement sequence.

[0090] The phase linkage index measures the multi-phase synchronism through the average coupling degree of the same parameter of each phase of the UPS.

[0091] The correction module dynamically adjusts the abnormal sequence based on the phase linkage index to prevent false positives.

[0092] The backtracking module automatically increases the tensor decomposition rank value when the abnormal value is continuously high to dynamically optimize the UPS coupling ground state description.

[0093] If the long-term sliding window abnormal value is increasing, it means that the current tensor decomposition ground state may not be suitable for the normal operation mode of the latest UPS (for example, the UPS has changed the working strategy or the load structure has changed), so the fluctuation of the multi-stage abnormal sequence needs to be backtracked to the tensor decomposition process;

[0094] Further, the backtracking module is pre-set with an abnormal threshold, and if the corrected average abnormal intensity of the updated multi-stage abnormality measurement sequence in the continuous multiple sliding windows exceeds the abnormal threshold, a backtracking instruction is triggered to backtrack the fluctuation of the updated multi-stage abnormality measurement sequence to the decomposition extraction module, and the decomposition rank value of the original CP decomposition is adaptively increased by 1 each time the backtracking is performed, so that the low-rank ground state after decomposition is closer to the current actual operation state of the UPS system, reducing the residual error amplitude and false positive rate, until the backtracking instruction is not triggered, the backtracking action is stopped, and the final low-rank ground state and disturbance residual are obtained.

[0095] By adaptively increasing the decomposition rank value, the decomposition ground state part is closer to the actual UPS normal coupling mode, gradually reducing the potential source of misjudgment. This way actively optimizes the coupling model through abnormal feedback, achieving dynamic self-learning.

[0096] In this multi-phase UPS battery monitoring system, because the load transients corresponding to each phase (e.g., phase A, phase B, and phase C) are different, the transient response of the battery providing energy to each phase is also different. In addition, different degrees of abnormal disturbances may be detected in each phase during coupled tensor decomposition. Therefore, the system calculates that certain phases are more worthy of focused monitoring (e.g., the tensor residual is largest in the direction of that phase), and thus allocates a higher sampling period to those phases. This achieves higher resolution monitoring of high-risk phases while ensuring global monitoring, continuously reducing the false positive rate and false negative rate of the entire system.

[0097] The correction module, based on the final perturbation residual, takes the average absolute residual of each parameter in each phase over time to form a sensitivity matrix. Each element in the sensitivity matrix refers to the average absolute residual of each parameter in the corresponding phase over time. Each element in the sensitivity matrix is ​​used to reflect the average anomaly concentration of the corresponding parameter in the corresponding phase. If the corresponding element in the sensitivity matrix exceeds the preset sensitivity threshold, the sampling period is dynamically adjusted; otherwise, no dynamic adjustment is performed.

[0098] The specific method for obtaining the average absolute residual over time for each parameter in each phase is as follows: ,in, For the first Phase 1 The parameters are the average absolute residuals over time, where N is the monitoring period. Number the time points, In the first At a certain point in time, the first Phase 1 The final perturbation residual (i.e., the abnormal residual value) of the parameter.

[0099] The method for dynamically adjusting the sampling period is as follows: ,in, For the first Phase 1 The sampling period of the parameters As the reference sampling period, This is an adjustment coefficient used to control the weight of the influence of parameter sensitivity on the sampling period (the value is generally around 0.5 to 2). For the first Phase 1 The parameters are averaged absolute residuals over the time dimension.

[0100] In this embodiment of the invention, by setting a backtracking module and a correction module in the coupled correction subsystem, the monitoring system for UPS multiphase battery parameters is equipped with the ability to dynamically self-optimize and adaptively focus sampling resources, which significantly improves the predictability and robustness of multiphase UPS battery collaborative health monitoring.

[0101] Specifically, the backtracking module first sets an abnormal threshold value for judging the maximum abnormal risk level that can be accepted in the UPS running state. By step-by-step detection on the updated multi-stage abnormality metric sequence (i.e. the corrected average abnormal intensity sequence on the sliding window), if the average abnormal intensity exceeds the set abnormal threshold value in a plurality of consecutive sliding windows (for example, 3 consecutive sliding windows), the backtracking instruction is triggered.

[0102] The plurality of consecutive sliding windows are used to avoid frequent triggering of backtracking due to single short-term fluctuations, and can more reliably identify the persistent abnormal deviation of the UPS multi-phase during load migration. For example, after the UPS switches to the energy-saving mode at night, it may be detected that the abnormality metric exceeds the threshold in the next 3 windows, at which time the backtracking action can be triggered.

[0103] After the backtracking is triggered, the backtracking module feeds back the multi-stage abnormal fluctuation information to the decomposition and extraction module, and adaptively increases the decomposition rank value by 1 based on the original tensor CP decomposition. The effect of increasing the decomposition rank value is that by introducing more principal coupling modes, the low-rank ground state after decomposition is closer to the new multi-phase energy migration mode of the UPS at this time, thereby reducing the residual amplitude and reducing the false positive rate caused by the migration of the normal running state. This process continues until the subsequent multi-stage abnormality metric sequence does not exceed the threshold in a plurality of consecutive sliding windows, indicating that the decomposition rank value is sufficient, and the backtracking action is automatically stopped, obtaining the low-rank ground state and abnormal residual that best fit the current actual operation of the UPS.

[0104] Further, the correction module generates a sensitivity matrix by taking the average absolute residual of each monitoring parameter (such as battery output voltage, current, impedance) of each phase in the time dimension based on the final disturbance residual tensor. For example, for the output voltage parameter of phase A battery, the absolute value of the residual sequence can be taken point by point and then averaged in the monitoring period to obtain a value reflecting the concentration of the parameter in the phase. Each element of the sensitivity matrix indicates the average abnormal intensity of the corresponding phase and the corresponding parameter, and if the value exceeds the preset sensitivity threshold, the mechanism of dynamically adjusting the sampling period is triggered, which automatically shortens the sampling period of the phase and the parameter, for example, when the sensitivity of the B-phase impedance is detected to be significantly increased, the sampling period is dynamically adjusted from 5ms to 2ms, so as to more densely collect data and timely capture the evolution process of potential abnormalities; if it does not exceed the threshold, the baseline sampling period is maintained and no dynamic adjustment is made.

[0105] In the above process, the abnormal threshold value is used to constrain the multi-stage abnormal risk trigger point;

[0106] The decomposition rank value is used to adjust the fitting degree of the tensor low-rank ground state to the UPS multi-phase cooperative behavior;

[0107] The sensitivity matrix is used to quantitatively reflect the abnormal concentration degree of each phase on each parameter;

[0108] Through the layer-by-layer progressive linkage of the backtracking module and the correction module, the application can timely absorb the natural migration when the normal working mode of the UPS multiphase system slightly migrates, and can realize higher-precision early abnormal capture through dynamic focusing of sampling resources when there is a real abnormal concentration, thereby significantly improving the health prediction accuracy of the UPS multiphase battery, and reducing false positives and waste of sampling resources. The backtracking module judges whether to trigger the adaptive tensor rank value through a multi-stage abnormal sliding sequence, thereby reducing false positives under natural migration; the correction module calculates the sensitivity matrix through the final residual error, dynamically adjusts the sampling period, and realizes self-focusing of monitoring resources.

[0109] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multi-phase UPS battery parameter collaborative monitoring and failure prediction system, characterized in that: comprises, The dynamic synchronous acquisition subsystem determines the transient pull characteristics corresponding to each phase after the disturbance occurs by installing multiple sensor measuring points on the DC bus branch between the battery pack and the inverter, obtains the coupling degree between each phase on the corresponding parameters in the transient pull characteristics through correlation coefficient analysis, and obtains the disturbance residual by using the CP decomposition method; The coupling correction subsystem constructs a multi-stage abnormality measurement sequence based on the disturbance residual, dynamically corrects the multi-stage abnormality measurement sequence based on the normal coupling caused by the multi-phase natural energy migration of the UPS, and judges whether to trigger a backtracking instruction. If the backtracking instruction is triggered, the sampling period of the parameters of the corresponding phase is dynamically adjusted. The coupling correction subsystem comprises an abnormality identification module, a correction module, a backtracking module, and a correction module. The correction module calculates the coupling degree between each phase of the UPS when facing the same parameters in the same sliding window, and obtains the average value to form a phase linkage index. Based on the value of the phase linkage index, the average abnormal strength in the corresponding sliding window is corrected in stages to obtain the corrected average abnormal strength, and the multi-stage abnormality measurement sequence is updated. The dynamic adjustment sampling period mode is: wherein, is the first is the first is the sampling period of the parameter, is the reference sampling period, is the adjustment coefficient, is the first is the first averages the absolute residual error of the parameter in the time dimension.

2. The multi-phase UPS battery parameter collaborative monitoring and failure prediction system of claim 1, wherein: The dynamic synchronous acquisition subsystem comprises an acquisition module, an offset module, and a decomposition and extraction module. The acquisition module records the transient characteristics corresponding to each phase after the disturbance occurs by connecting a programmable electronic load in parallel to the multi-phase output end of the UPS system and installing multiple sensor measuring points on the DC bus branch between the battery pack and the inverter under normal load conditions of the multi-phase UPS system, including the battery output voltage, battery output current, and complex impedance corresponding to each phase after the disturbance occurs.

3. The multi-phase UPS battery parameter collaborative monitoring and fault prediction system according to claim 2, characterized in that: The offset module determines the transient pull characteristics corresponding to each phase after the disturbance occurs according to the transient characteristics, including the transient drop amplitude of the battery voltage corresponding to each phase, the current peak change, and the impedance trajectory, and tensorizes the transient pull characteristics corresponding to each phase to form a third-order tensor. Based on the third-order tensor, the coupling degree between each phase on the corresponding parameters in the transient pull characteristics is obtained through correlation coefficient analysis. The coupling degree is used to capture the cross disturbance of the transient pull characteristics between different phases, and the coupling degree between each phase on the corresponding parameters in the transient pull characteristics is used as a supplementary attribute of the third-order tensor.

4. The multi-phase UPS battery parameter collaborative monitoring and fault prediction system according to claim 3, characterized in that: The decomposition and extraction module decomposes the third-order tensor into a low-rank ground state and a disturbance residual by using the CP decomposition method, wherein the low-rank ground state represents the collaborative behavior mode under normal load conversion of the multi-phase UPS, and the disturbance residual is used to explicitly represent abnormal single-phase or cross-phase degradation signs.

5. The multi-phase UPS battery parameter collaborative monitoring and fault prediction system according to claim 4, characterized in that: The anomaly identification module obtains a disturbance residual vector by unfolding the disturbance residual in the time dimension, analyzes the disturbance residual at different time points based on the disturbance residual vector, obtains the anomaly strength at each time point by calculating the Frobenius norm of the disturbance residual vector, sets a sliding window, and calculates the average anomaly strength on each sliding window to form a multi-stage anomaly metric sequence.

6. The multi-phase UPS battery parameter collaborative monitoring and fault prediction system according to claim 5, characterized in that: The backtracking module pre-sets an anomaly threshold, if the corrected average anomaly strength of a plurality of consecutive sliding windows in the updated multi-stage anomaly metric sequence exceeds the anomaly threshold, a backtracking instruction is triggered to backtrack the fluctuation of the updated multi-stage anomaly metric sequence to the decomposition and extraction module, and the decomposition rank value of the original CP decomposition is adaptively increased by 1 each time the backtracking is performed until the backtracking instruction is not triggered, and the backtracking action is stopped, to obtain the final low-rank ground state and disturbance residual.

7. The multi-phase UPS battery parameter collaborative monitoring and fault prediction system according to claim 6, characterized in that: The correction module takes the average absolute residual in the time dimension based on the final disturbance residual to form a sensitivity matrix, each element in the sensitivity matrix refers to the parameters of the corresponding phase, and each element in the sensitivity matrix is used to reflect the average anomaly concentration of the corresponding parameter of the corresponding phase, if the corresponding element in the sensitivity matrix exceeds the pre-set sensitivity threshold, the sampling period is dynamically adjusted, otherwise no dynamic adjustment is performed.

Images