UPS battery health degree prediction method and system based on multi-source data

Abstract

The invention provides a UPS battery health degree prediction method and system based on multi-source data, relates to the technical field of battery health degree prediction, and establishes a corresponding relation group of health degree and multi-source parameters (such as charging power and discharging temperature curves) by constructing a battery health degree comparison experiment. Based on health degree classification and feature analysis, forming a health degree corresponding relation subset, determining a mapping parameter range of each parameter, and combining the mapping parameter ranges into a range group; and performing data increment by using the range group to generate a first battery application state parameter group, and constructing a parameter library through a verification experiment. And optimizing the parameter library based on the comprehensive anastomosis parameter, shrinking the mapping parameter range, and eliminating distorted data. And finally, analyzing real-time parameters by using the updated parameter library, and predicting the health degree of the battery. Through multi-source data integration, feature extraction and dynamic optimization, the prediction precision and robustness are improved, and reliable support is provided for UPS system maintenance.

Description

Technical Field

[0001] The present invention relates to the technical field of battery health prediction, and in particular to a UPS battery health prediction method and system based on multi-source data. Background Art

[0002] With the widespread deployment of uninterruptible power supply (UPS) systems in critical infrastructure such as communication networks, industrial automation, and emergency power supply, the health of batteries, as core components of UPS systems, directly affects the reliability and service life of the systems. Traditional battery health prediction methods often rely on single indicators (such as internal resistance changes) or simple statistical models, which cannot fully reflect the degradation mechanism of batteries under complex operating conditions, resulting in insufficient prediction accuracy, especially under dynamic loads and environmental changes. In recent years, advances in multi-source data processing and intelligent algorithms have opened up new paths for battery health prediction. By integrating multi-dimensional parameters such as charging power curves, discharging power curves, charging temperature change curves, and discharging temperature change curves, it is possible to more accurately characterize the operating characteristics and degradation patterns of batteries. Summary of the Invention

[0003] The purpose of the present invention is to provide a health prediction method and system that can accurately detect the health of batteries in a UPS system.

[0004] The present invention discloses a UPS battery health prediction method based on multi-source data, comprising: Step S100, constructing a battery health comparison experiment to determine a number of health correspondence relationship groups associated with battery health and battery application status parameters; Step S200: Classifying the health correspondence groups based on the battery health to obtain a plurality of health correspondence sets, performing feature analysis on the battery application status parameters in each health correspondence group in the health correspondence set, and further classifying health correspondence groups with identical feature expressions to obtain a plurality of health correspondence subsets; Step S300, determining a mapping parameter range for each type of battery application status parameter in each health correspondence subset, and combining the mapping parameter ranges of different types of battery application status parameters into a mapping parameter range group; Step S400: Based on the mapping parameter range groups corresponding to each health correspondence set, data increment is performed to construct several first battery application status parameter groups that conform to the mapping parameter range groups, and several new battery health comparison experiments are constructed to obtain several verification health correspondence relationship groups. Step S500: Combining the first battery application state parameter groups into a battery application state parameter library, and performing a verification test on the battery application state parameter library using a plurality of verification battery application state parameter groups. Analyzing the comprehensive matching parameters between the retrieved matching battery application state parameter groups and the verification battery application state parameter groups, and based on the comprehensive matching parameters, narrowing the mapping parameter range in the mapping parameter range group corresponding to the matching battery application state parameter groups, and eliminating the corresponding first battery application state parameter groups, thereby updating the battery application state parameter library. Step S600 : Analyze the real-time battery application status parameter group using the updated battery application status parameter library to determine the real-time battery health.

[0005] In the embodiment disclosed in the present invention, the types of battery application status parameters include: charging power curve, discharging power curve, charging temperature change curve, and temperature change curve.

[0006] In an embodiment disclosed in the present invention, a method for performing feature analysis on battery application status parameters in each health correspondence group in a health correspondence set includes: Step S201: aligning the charging power curve with the charging temperature change curve, and aligning the discharging power curve with the discharging temperature change curve according to the time correspondence principle; Step S202: Determine the changing trend of each curve, intercept the curve segments whose rising trend duration exceeds a preset value and whose average rising amplitude is greater than or equal to the preset value, and record them as rising curve segments; intercept the curve segments whose falling trend duration exceeds a preset value and whose average falling amplitude is greater than or equal to the preset value, and record them as falling curve segments; intercept the remaining curve segments, and record them as stable curve segments; Step S203, determining the time lengths and average ordinate values of the ascending curve segment, the descending curve segment, and the steady curve segment, and sorting them by size to obtain a curve segment parameter group; Step S204 : classifying the health correspondence groups based on the consistency between the curve segment parameter groups to obtain a number of health degree correspondence subsets.

[0007] In an embodiment disclosed in the present invention, a method for determining a mapping parameter range of each type of battery application state parameter in each health correspondence subset includes: Step S301 : Overlap and compare the curves of the same type in the health correspondence subset, determine the curve coverage area between the curves, and identify the curve coverage area as the mapping parameter range.

[0008] In the embodiment disclosed in the present invention, a method for performing data increment based on a plurality of mapping parameter range groups corresponding to each health correspondence relationship set includes: Step S401: uniformly set a number of reference time points on the horizontal time axis of the curve coverage area, determine the vertical mapping line of the curve coverage area corresponding to each reference time point, and determine the center point of the vertical mapping line, which is recorded as the vertical center point; Step S402: Connect the longitudinal center points to obtain a reference curve, and analyze the boundary distance characteristics of the area covered by the curve relative to the reference curve. The boundary distance characteristics are the relative distances of the boundary nodes of different areas relative to the reference line. The relative distances are sorted from large to small, and several relative distances with the highest order are selected. The average of these distances is calculated and recorded as the average relative distance. In step S403, based on the preset distance interval to which the average relative distance belongs, the allowable fluctuation parameter of each curve is determined. Based on the allowable fluctuation parameter of each curve, the curve corresponding to the mapping parameter range group is randomly reconstructed several times to obtain a reconstructed curve, and the reconstructed curves of different types are combined to obtain the first battery application state parameter group.

[0009] In an embodiment disclosed in the present invention, a method for performing a plurality of random reconstructions of curves corresponding to a mapping parameter range group based on an allowable fluctuation parameter of each curve includes: Step S4031, determining the lengths of the longitudinal mapping lines corresponding to different time nodes in the curve coverage area, and determining the reconstruction difference amounts at different time nodes in combination with the allowable fluctuation parameter; Step S4032 : Based on the reconstruction difference, construct reconstructed curve nodes above and below the curve nodes corresponding to the time nodes, and connect the reconstructed curve nodes to obtain a reconstructed curve.

[0010] In an embodiment disclosed in the present invention, the method for determining the reconstruction difference includes: Step S40311, based on the preset fluctuation parameter interval to which the allowed fluctuation parameter belongs, determine the corresponding length scale coefficient, calculate the product of the length scale coefficient and the length of the longitudinal mapping line, obtain the maximum reconstructed difference, and randomly generate several reconstructed difference amounts within a value less than or equal to the maximum reconstructed difference amount.

[0011] In an embodiment disclosed in the present invention, a method for analyzing the comprehensive matching parameters of the retrieved matching battery application state parameter group and the verification battery application state parameter group includes: Step S501: Overlap and compare the corresponding curves in the matching battery application state parameter set and the verification battery application state parameter set. If the coverage area between each relative curve is less than or equal to a preset value, it is determined that the matching battery application state parameter set and the verification battery application state parameter set match, and the battery health difference between the two is calculated; Step S502, determining a comprehensive matching parameter based on the number of verification comparisons and the corresponding health difference amount each time; ; Where W is the comprehensive matching parameter, K is the matching parameter adjustment coefficient, b is the matching parameter adjustment constant, and n is the total number of verification comparisons. is the difference in health during the i-th verification comparison, is the health difference judgment function, based on The health difference interval to which it belongs outputs the corresponding sub-parameter.

[0012] In an embodiment disclosed in the present invention, a method for narrowing the mapping parameter range in the mapping parameter range group corresponding to the matching battery application state parameter group based on the comprehensive matching parameter includes: Step S503: determining a reduction ratio of the mapping parameter range based on the preset matching parameter interval to which the comprehensive matching parameter belongs, and moving the upper and lower boundaries of the mapping parameter range inward based on the reduction ratio of the mapping parameter range.

[0013] In the embodiments disclosed in the present invention, a UPS battery health prediction system based on multi-source data is also disclosed, including: The first module is used to build a battery health comparison experiment and determine several health correspondence relationship groups associated with battery health and battery application status parameters; The second module is used to classify the health correspondence relationship groups based on the battery health to obtain a number of health correspondence relationship sets, perform feature analysis on the battery application status parameters in each health correspondence relationship group in the health correspondence relationship set, and further classify the health correspondence relationship groups with the same feature performance to obtain a number of health correspondence relationship subsets; The third module is configured to determine a mapping parameter range for each type of battery application status parameter in each health correspondence relationship subset, and combine the mapping parameter ranges of different types of battery application status parameters into a mapping parameter range group; The fourth module is used to perform data increment based on the mapping parameter range groups corresponding to each health correspondence set, construct several first battery application status parameter groups that conform to the mapping parameter range groups, and newly construct several battery health comparison experiments to obtain several health correspondence relationship groups for verification; a fifth module, configured to construct a battery application state parameter library by combining the first battery application state parameter groups, and perform verification testing on the battery application state parameter library using a plurality of verification battery application state parameter groups, analyze the comprehensive matching parameters between the retrieved matching battery application state parameter groups and the verification battery application state parameter groups, and based on the comprehensive matching parameters, narrow the mapping parameter range in the mapping parameter range group corresponding to the matching battery application state parameter groups, and eliminate the corresponding first battery application state parameter groups, thereby updating the battery application state parameter library; The sixth module is used to analyze the real-time battery application status parameter group using the updated battery application status parameter library to determine the real-time battery health.

[0014] The present invention proposes a UPS battery health prediction method and system based on multi-source data, which relates to the technical field of battery health prediction. By constructing a battery health comparison experiment, a correspondence group between health and multi-source parameters (such as charging power and discharge temperature curve) is established. Based on health classification and feature analysis, a subset of health correspondence is formed, and the mapping parameter range of each parameter is determined and combined into a range group. The range group is used to increment data to generate a first battery application status parameter group, and a parameter library is constructed through verification experiments. The parameter library is optimized based on comprehensive matching parameters, the mapping parameter range is narrowed, and distorted data is eliminated. Finally, the updated parameter library is used to analyze real-time parameters and predict battery health. The present invention improves prediction accuracy and robustness through multi-source data integration, feature extraction and dynamic optimization, providing reliable support for UPS system maintenance.

[0015] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a diagram of the steps of a UPS battery health prediction method based on multi-source data disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0018] The following will be combined with the accompanying drawings and specific embodiments to clearly and completely describe the technical solutions of the present invention. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and cannot be understood as limiting the scope of protection of the present invention. Those skilled in the art in this field can make some non-essential improvements and adjustments based on the content of the present invention described below. In the present invention, unless otherwise clearly specified and limited, the technical terms used in the present invention should have the common meanings understood by those skilled in the art of the present invention.

[0019] Example:

[0020] The purpose of the present invention is to provide a health prediction method and system that can accurately detect the health of batteries in a UPS system.

[0021] The present invention discloses a UPS battery health prediction method based on multi-source data. Figure 1 ,include: Step S100 : constructing a battery health comparison experiment to determine a number of health correspondence relationship groups associated with battery health and battery application status parameters.

[0022] The principle of step S100 is to experimentally establish a correlation between battery health and multiple application status parameters, providing foundational data for subsequent modeling. Specifically, a series of battery testing experiments with controlled variables are designed to collect operating data under different operating conditions, such as charging power curves, discharging power curves, charging temperature curves, and discharging temperature curves. Battery health indicators (such as remaining capacity or internal resistance) are also measured simultaneously. By analyzing this data, the corresponding relationship between battery health and application status parameters under each set of experimental conditions is determined, forming a health correspondence group. For example, if, in one set of experiments, the battery operates under a specific charging power curve (with power increasing slowly over time) and its health is characterized by a remaining capacity of 90%, then this charging power curve and a health of 90% constitute a correspondence group. Through multiple sets of experiments, a sufficient number of correspondence groups are accumulated to provide data support for subsequent classification and feature analysis. The essence of this process is to quantify complex battery operating conditions into analyzable, correlated datasets, laying the foundation for subsequent pattern recognition and prediction.

[0023] In step S200, the health correspondence groups are classified based on the battery health to obtain several health correspondence sets, and characteristic analysis is performed on the battery application status parameters in each health correspondence group in the health correspondence set. The health correspondence groups with the same characteristic performance are further classified to obtain several health correspondence subsets.

[0024] The principle behind step S200 is to cluster and perform feature analysis on health correspondence groups, grouping the data to uncover hidden patterns and thereby improve the efficiency and accuracy of subsequent modeling. First, the correspondence groups are categorized into several health correspondence sets based on battery health (e.g., remaining capacity range). For example, correspondence groups with health levels between 80% and 90% are grouped together to form a health correspondence set. Next, feature analysis is performed on the battery application state parameters (e.g., charging power curves) within each health correspondence set to extract key features, such as the curve's upward and downward trends, and the duration and amplitude of plateaus. Based on these features, correspondence groups with similar characteristics are further categorized into health correspondence subsets. For example, if the charging power curves of two correspondence groups both exhibit a short, rapid rise followed by a long period of plateauing, and their health levels are similar, they are classified into the same subset. The essence of this process is to simplify complex datasets into more representative subsets through dimensionality reduction and pattern recognition, facilitating subsequent parameter range determination and data increment.

[0025] Step S300 : determining a mapping parameter range of each type of battery application status parameter in each health correspondence subset, and combining mapping parameter ranges of different types of battery application status parameters into a mapping parameter range group.

[0026] The principle of step S300 is to analyze the overlap of parameter curves in the subset to determine the range of variation for each parameter at different time points (i.e., mapping parameter ranges). These ranges are then combined into multidimensional range groups to provide constraints for data generation. Specifically, the overlap of curves of the same type (e.g., charging power curves) in the subset is compared to identify the overlap region of all curves at each time point, defining this as the mapping parameter range. For example, if the power values of all charging power curves in the subset at the 5th minute are within [120W, 140W], then the mapping parameter range for that time point is [120W, 140W]. Subsequently, mapping parameter ranges for different types of parameters (e.g., charging power and discharge temperature) are combined into a mapping parameter range group, such as {charging power at the 5th minute: [120W, 140W], discharge temperature at the 5th minute: [25°C, 28°C]}. The essence of this process is to quantify the allowable variation range of parameters based on the curve overlap characteristics, ensuring that subsequently generated data conforms to the operating rules of the subset and providing a reliable basis for data increment and verification.

[0027] In step S400, data increment is performed based on the mapping parameter range groups corresponding to each health correspondence set, and several first battery application status parameter groups that conform to the mapping parameter range groups are constructed. Several new battery health comparison experiments are also constructed to obtain several health correspondence groups for verification.

[0028] The principle of step S400 is to generate virtual data by mapping parameter ranges, expand the dataset, and verify its validity to enhance the adaptability of the model. Based on the mapped parameter range group, random reconstruction is used to generate curves that meet the range constraints. For example, power values are randomly generated within the power range [120W, 140W] at the 5th minute to construct the first battery application state parameter group. These generated curves are optimized based on their overlap characteristics (such as boundary distance and allowable fluctuation parameters) to ensure closeness to real operating data. Subsequently, through simulation or experimentation, these parameter groups are used to conduct new health comparison experiments to determine their corresponding health, forming a health correspondence group for verification. For example, after generating a set of charging power curves, if the health is experimentally verified to be 87%, a verification correspondence group is formed. This process constrains data generation by mapping parameter ranges (based on curve overlap) to ensure data representativeness. Verification experiments also confirm the reliability of the generated data, supporting the construction of the parameter library.

[0029] Step S500, the first battery application state parameter group is combined to form a battery application state parameter library, and a plurality of verification battery application state parameter groups are used to perform verification testing on the battery application state parameter library, and the comprehensive matching parameters between the retrieved matching battery application state parameter group and the verification battery application state parameter group are analyzed, and based on the comprehensive matching parameters, the mapping parameter range in the mapping parameter range group corresponding to the matching battery application state parameter group is narrowed, and the corresponding first battery application state parameter group is eliminated, and the battery application state parameter library is updated.

[0030] The principle of step S500 is to optimize the parameter library through validation testing, assess data quality based on comprehensive fit parameters, and narrow the mapping parameter range to improve prediction accuracy. First, the first battery application state parameter set is integrated into the parameter library. Then, testing is performed using the verification parameter set. A comprehensive fit parameter is calculated by comparing curve overlap and health difference, assessing the degree of fit between the matching and verification parameter sets. For example, if the matching charging power curve and the verification curve overlap at all time points and the health difference is less than 1%, the fit parameter is high. A low comprehensive fit parameter indicates that the matching parameter set is distorted, and the corresponding mapping parameter range needs to be narrowed (for example, narrowing the power range at the fifth minute from [120W, 140W] to [125W, 135W]). Parameter sets outside the new range are removed, and the parameter library is updated. This process triggers range narrowing by low fit parameters, optimizes curve overlap, ensures more reliable parameter library data, and improves prediction model accuracy.

[0031] Step S600 : Analyze the real-time battery application status parameter group using the updated battery application status parameter library to determine the real-time battery health.

[0032] In the embodiment disclosed in the present invention, the types of battery application status parameters include: charging power curve, discharging power curve, charging temperature change curve, and temperature change curve.

[0033] In an embodiment disclosed in the present invention, a method for performing feature analysis on battery application status parameters in each health correspondence group in a health correspondence set includes: In step S201 , the charging power curve and the charging temperature change curve are aligned, and the discharging power curve and the discharging temperature change curve are aligned according to the time correspondence principle.

[0034] In step S202, the changing trend of each curve is determined, and the curve segments whose duration of the rising trend exceeds the preset value and whose average rising amplitude is greater than or equal to the preset value are intercepted and recorded as rising curve segments. The curve segments whose duration of the falling trend exceeds the preset value and whose average falling amplitude is greater than or equal to the preset value are intercepted and recorded as falling curve segments. The remaining curve segments are intercepted and recorded as stable curve segments.

[0035] Step S203 : determining the time lengths and average ordinate values of the ascending curve segment, the descending curve segment, and the steady curve segment, and sorting them by size to obtain a curve segment parameter group.

[0036] Step S204 : classifying the health correspondence groups based on the consistency between the curve segment parameter groups to obtain a number of health degree correspondence subsets.

[0037] In an embodiment disclosed in the present invention, a method for determining a mapping parameter range of each type of battery application state parameter in each health correspondence subset includes: Step S301 : Overlap and compare the curves of the same type in the health correspondence subset, determine the curve coverage area between the curves, and identify the curve coverage area as the mapping parameter range.

[0038] In the embodiment disclosed in the present invention, a method for performing data increment based on a plurality of mapping parameter range groups corresponding to each health correspondence relationship set includes: Step S401: uniformly set a number of reference time points on the horizontal time axis of the curve coverage area, determine the vertical mapping line of the curve coverage area corresponding to each reference time point, and determine the center point of the vertical mapping line, which is recorded as the vertical center point; Step S402: Connect the longitudinal center points to obtain a reference curve, and analyze the boundary distance characteristics of the area covered by the curve relative to the reference curve. The boundary distance characteristics are the relative distances of the boundary nodes of different areas relative to the reference line. The relative distances are sorted from large to small, and several relative distances with the highest order are selected. The average of these distances is calculated and recorded as the average relative distance. In step S403, based on the preset distance interval to which the average relative distance belongs, the allowable fluctuation parameter of each curve is determined. Based on the allowable fluctuation parameter of each curve, the curve corresponding to the mapping parameter range group is randomly reconstructed several times to obtain a reconstructed curve, and the reconstructed curves of different types are combined to obtain the first battery application state parameter group.

[0039] The principle of step S403 is to determine the fluctuation parameter based on the average relative distance, generate a reconstructed curve that conforms to the mapping parameter range, and construct the first battery application state parameter set. Specifically, based on the preset distance interval to which the average relative distance (e.g., 8W) falls, an allowable fluctuation parameter (e.g., ±5W) is determined, representing the variable range of the curve around the reference curve. Then, within the mapping parameter range (e.g., [120W, 140W] at the 5th minute), a random reconstruction is performed with the reference curve value as the center, combined with the fluctuation parameter, to generate a new curve. For example, if the reference value at the 5th minute is 130W and the fluctuation parameter is ±5W, then the randomly generated value is within [125W, 135W], and the values at each time node are connected to form a reconstructed curve. Reconstructed curves of different types (e.g., charging power and discharging temperature) are combined to form the first battery application state parameter set. This process ensures that the generated data closely follows the patterns of the coverage area. If the comprehensive fit parameter is low, it indicates that the reconstructed curve is distorted, and the mapping parameter range needs to be narrowed (e.g., [125W, 135W]) to constrain the fluctuation range and improve data authenticity and fit.

[0040] In an embodiment disclosed in the present invention, a method for performing a plurality of random reconstructions of curves corresponding to a mapping parameter range group based on an allowable fluctuation parameter of each curve includes: Step S4031 : determining the lengths of the longitudinal mapping lines corresponding to different time nodes in the curve coverage area, and determining the reconstruction difference amounts at different time nodes in combination with the allowable fluctuation parameter.

[0041] Step S4032 : Based on the reconstruction difference, construct reconstructed curve nodes above and below the curve nodes corresponding to the time nodes, and connect the reconstructed curve nodes to obtain a reconstructed curve.

[0042] In an embodiment disclosed in the present invention, the method for determining the reconstruction difference includes: Step S40311, based on the preset fluctuation parameter interval to which the allowed fluctuation parameter belongs, determine the corresponding length scale coefficient, calculate the product of the length scale coefficient and the length of the longitudinal mapping line, obtain the maximum reconstructed difference, and randomly generate several reconstructed difference amounts within a value less than or equal to the maximum reconstructed difference amount.

[0043] In an embodiment disclosed in the present invention, a method for analyzing the comprehensive matching parameters of the retrieved matching battery application state parameter group and the verification battery application state parameter group includes: In step S501, the corresponding curves in the matching battery application status parameter group and the verification battery application status parameter group are overlapped and compared. If the coverage area between each relative curve is less than or equal to a preset value, it is determined that the parameters between the matching battery application status parameter group and the verification battery application status parameter are matched, and the health difference between the two batteries is calculated.

[0044] Step S502: Determine the comprehensive matching parameter based on the number of verification comparisons and the corresponding health difference each time.

[0045] .

[0046] Where W is the comprehensive matching parameter, K is the matching parameter adjustment coefficient, b is the matching parameter adjustment constant, and n is the total number of verification comparisons. is the difference in health during the i-th verification comparison, is the health difference judgment function, based on The health difference interval to which it belongs outputs the corresponding sub-parameter.

[0047] In an embodiment disclosed in the present invention, a method for narrowing the mapping parameter range in the mapping parameter range group corresponding to the matching battery application state parameter group based on the comprehensive matching parameter includes: Step S503: determining a reduction ratio of the mapping parameter range based on the preset matching parameter interval to which the comprehensive matching parameter belongs, and moving the upper and lower boundaries of the mapping parameter range inward based on the reduction ratio of the mapping parameter range.

[0048] In the embodiments disclosed in the present invention, a UPS battery health prediction system based on multi-source data is also disclosed, including: The first module is used to build a battery health comparison experiment and determine several health correspondence relationship groups associated with battery health and battery application status parameters; The second module is used to classify the health correspondence relationship groups based on the battery health to obtain a number of health correspondence relationship sets, perform feature analysis on the battery application status parameters in each health correspondence relationship group in the health correspondence relationship set, and further classify the health correspondence relationship groups with the same feature performance to obtain a number of health correspondence relationship subsets; The third module is configured to determine a mapping parameter range for each type of battery application status parameter in each health correspondence relationship subset, and combine the mapping parameter ranges of different types of battery application status parameters into a mapping parameter range group; The fourth module is used to perform data increment based on the mapping parameter range groups corresponding to each health correspondence set, construct several first battery application status parameter groups that conform to the mapping parameter range groups, and newly construct several battery health comparison experiments to obtain several health correspondence relationship groups for verification; a fifth module, configured to construct a battery application state parameter library by combining the first battery application state parameter groups, and perform verification testing on the battery application state parameter library using a plurality of verification battery application state parameter groups, analyze the comprehensive matching parameters between the retrieved matching battery application state parameter groups and the verification battery application state parameter groups, and based on the comprehensive matching parameters, narrow the mapping parameter range in the mapping parameter range group corresponding to the matching battery application state parameter groups, and eliminate the corresponding first battery application state parameter groups, thereby updating the battery application state parameter library; The sixth module is used to analyze the real-time battery application status parameter group using the updated battery application status parameter library to determine the real-time battery health.

[0049] The present invention proposes a UPS battery health prediction method and system based on multi-source data, which relates to the technical field of battery health prediction. By constructing a battery health comparison experiment, a correspondence group between health and multi-source parameters (such as charging power and discharge temperature curve) is established. Based on health classification and feature analysis, a subset of health correspondence is formed, and the mapping parameter range of each parameter is determined and combined into a range group. The range group is used to increment data to generate a first battery application status parameter group, and a parameter library is constructed through verification experiments. The parameter library is optimized based on comprehensive matching parameters, the mapping parameter range is narrowed, and distorted data is eliminated. Finally, the updated parameter library is used to analyze real-time parameters and predict battery health. The present invention improves prediction accuracy and robustness through multi-source data integration, feature extraction and dynamic optimization, providing reliable support for UPS system maintenance.

[0050] Through the above description of the embodiments, those skilled in the art will clearly understand that the present invention can be implemented via hardware or via software combined with a necessary general-purpose hardware platform. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product. This software product can be stored on a non-volatile storage medium (such as a CD-ROM, USB flash drive, or external hard drive) and includes instructions for enabling a computer device (such as a personal computer, server, or network device) to execute the methods described in various implementation scenarios of the present invention.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A UPS battery health prediction method based on multi-source data, characterized in that: include: Step S100, constructing a battery health comparison experiment to determine a number of health correspondence relationship groups associated with battery health and battery application status parameters; Step S200: Classifying the health correspondence groups based on the battery health to obtain a plurality of health correspondence sets, performing feature analysis on the battery application status parameters in each health correspondence group in the health correspondence set, and further classifying health correspondence groups with identical feature expressions to obtain a plurality of health correspondence subsets; Step S300, determining a mapping parameter range for each type of battery application status parameter in each health correspondence subset, and combining the mapping parameter ranges of different types of battery application status parameters into a mapping parameter range group; Step S400: Based on the mapping parameter range groups corresponding to each health correspondence set, data increment is performed to construct several first battery application status parameter groups that conform to the mapping parameter range groups, and several new battery health comparison experiments are constructed to obtain several verification health correspondence relationship groups. Step S500: Combining the first battery application state parameter groups into a battery application state parameter library, and performing a verification test on the battery application state parameter library using a plurality of verification battery application state parameter groups. Analyzing the comprehensive matching parameters between the retrieved matching battery application state parameter groups and the verification battery application state parameter groups, and based on the comprehensive matching parameters, narrowing the mapping parameter range in the mapping parameter range group corresponding to the matching battery application state parameter groups, and eliminating the corresponding first battery application state parameter groups, thereby updating the battery application state parameter library. Step S600 : Analyze the real-time battery application status parameter group using the updated battery application status parameter library to determine the real-time battery health.

2. The UPS battery health prediction method based on multi-source data according to claim 1 is characterized in that: The types of battery application status parameters include: charging power curve, discharging power curve, charging temperature change curve and temperature change curve.

3. The UPS battery health prediction method based on multi-source data according to claim 2 is characterized in that: The method for performing feature analysis on the battery application status parameters in each health correspondence group in the health correspondence set includes: Step S201: aligning the charging power curve with the charging temperature change curve, and aligning the discharging power curve with the discharging temperature change curve according to the time correspondence principle; Step S202: Determine the changing trend of each curve, intercept the curve segments whose rising trend duration exceeds a preset value and whose average rising amplitude is greater than or equal to the preset value, and record them as rising curve segments; intercept the curve segments whose falling trend duration exceeds a preset value and whose average falling amplitude is greater than or equal to the preset value, and record them as falling curve segments; intercept the remaining curve segments, and record them as stable curve segments; Step S203, determining the time lengths and average ordinate values of the ascending curve segment, the descending curve segment, and the steady curve segment, and sorting them by size to obtain a curve segment parameter group; Step S204 : classifying the health correspondence groups based on the consistency between the curve segment parameter groups to obtain a number of health degree correspondence subsets.

4. The UPS battery health prediction method based on multi-source data according to claim 2, characterized in that: The method for determining the mapping parameter range of each type of battery application state parameter in each health correspondence subset includes: Step S301 : Overlap and compare the curves of the same type in the health correspondence subset, determine the curve coverage area between the curves, and identify the curve coverage area as the mapping parameter range.

5. The UPS battery health prediction method based on multi-source data according to claim 4 is characterized in that: Methods for performing data increment based on a plurality of mapping parameter range groups corresponding to each health correspondence relationship set include: Step S401: uniformly set a number of reference time points on the horizontal time axis of the curve coverage area, determine the vertical mapping line of the curve coverage area corresponding to each reference time point, and determine the center point of the vertical mapping line, which is recorded as the vertical center point; Step S402: Connect the longitudinal center points to obtain a reference curve, and analyze the boundary distance characteristics of the area covered by the curve relative to the reference curve. The boundary distance characteristics are the relative distances of the boundary nodes of different areas relative to the reference line. The relative distances are sorted from large to small, and several relative distances with the highest order are selected. The average of these distances is calculated and recorded as the average relative distance. In step S403, based on the preset distance interval to which the average relative distance belongs, the allowable fluctuation parameter of each curve is determined. Based on the allowable fluctuation parameter of each curve, the curve corresponding to the mapping parameter range group is randomly reconstructed several times to obtain a reconstructed curve, and the reconstructed curves of different types are combined to obtain the first battery application state parameter group.

6. The UPS battery health prediction method based on multi-source data according to claim 5, characterized in that: The method of performing a number of random reconstructions of the curves corresponding to the mapping parameter range group based on the allowed fluctuation parameter of each curve includes: Step S4031, determining the lengths of the longitudinal mapping lines corresponding to different time nodes in the curve coverage area, and determining the reconstruction difference amounts at different time nodes in combination with the allowable fluctuation parameter; Step S4032 : Based on the reconstruction difference, construct reconstructed curve nodes above and below the curve nodes corresponding to the time nodes, and connect the reconstructed curve nodes to obtain a reconstructed curve.

7. The UPS battery health prediction method based on multi-source data according to claim 6, characterized in that: Methods for determining the amount of reconstruction difference include: Step S40311, based on the preset fluctuation parameter interval to which the allowed fluctuation parameter belongs, determine the corresponding length scale coefficient, calculate the product of the length scale coefficient and the length of the longitudinal mapping line, obtain the maximum reconstructed difference, and randomly generate several reconstructed difference amounts within a value less than or equal to the maximum reconstructed difference amount.

8. The UPS battery health prediction method based on multi-source data according to claim 2, characterized in that: The method for analyzing the comprehensive matching parameters of the retrieved matching battery application state parameter group and the verification battery application state parameter group includes: Step S501: Overlap and compare the corresponding curves in the matching battery application state parameter set and the verification battery application state parameter set. If the coverage area between each relative curve is less than or equal to a preset value, it is determined that the matching battery application state parameter set and the verification battery application state parameter set match, and the battery health difference between the two is calculated; Step S502, determining a comprehensive matching parameter based on the number of verification comparisons and the corresponding health difference amount each time; ； Where W is the comprehensive matching parameter, K is the matching parameter adjustment coefficient, b is the matching parameter adjustment constant, and n is the total number of verification comparisons. is the difference in health during the i-th verification comparison, is the health difference judgment function, based on The health difference interval to which it belongs outputs the corresponding sub-parameter.

9. The UPS battery health prediction method based on multi-source data according to claim 1, characterized in that: The method for narrowing the mapping parameter range in the mapping parameter range group corresponding to the matching battery application state parameter group based on the comprehensive matching parameter includes: Step S503: determining a reduction ratio of the mapping parameter range based on the preset matching parameter interval to which the comprehensive matching parameter belongs, and moving the upper and lower boundaries of the mapping parameter range inward based on the reduction ratio of the mapping parameter range.

10. UPS battery health prediction system based on multi-source data, characterized by: A UPS battery health prediction method for executing any one of claims 1-9, comprising: The first module is used to build a battery health comparison experiment and determine several health correspondence relationship groups associated with battery health and battery application status parameters; The second module is used to classify the health correspondence relationship groups based on the battery health to obtain a number of health correspondence relationship sets, perform feature analysis on the battery application status parameters in each health correspondence relationship group in the health correspondence relationship set, and further classify the health correspondence relationship groups with the same feature performance to obtain a number of health correspondence relationship subsets; The third module is configured to determine a mapping parameter range for each type of battery application status parameter in each health correspondence relationship subset, and combine the mapping parameter ranges of different types of battery application status parameters into a mapping parameter range group; The fourth module is used to perform data increment based on the mapping parameter range groups corresponding to each health correspondence set, construct several first battery application status parameter groups that conform to the mapping parameter range groups, and newly construct several battery health comparison experiments to obtain several health correspondence relationship groups for verification; a fifth module, configured to construct a battery application state parameter library by combining the first battery application state parameter groups, and perform verification testing on the battery application state parameter library using a plurality of verification battery application state parameter groups, analyze the comprehensive matching parameters between the retrieved matching battery application state parameter groups and the verification battery application state parameter groups, and based on the comprehensive matching parameters, narrow the mapping parameter range in the mapping parameter range group corresponding to the matching battery application state parameter groups, and eliminate the corresponding first battery application state parameter groups, thereby updating the battery application state parameter library; The sixth module is used to analyze the real-time battery application status parameter group using the updated battery application status parameter library to determine the real-time battery health.

Images