Backup power source health condition analysis method, system and equipment based on AI deep learning and battery big data processing and medium

Abstract

The invention discloses a backup power source health condition analysis method, system and device based on AI deep learning and battery big data processing and a medium, and belongs to the technical field of backup power source health condition analysis. Synchronously recording the timestamps and the device identifiers of the parameters in the acquisition process; the internal resistance of the battery is measured, when the internal resistance of the storage battery is too small, the detection method is switched, and internal resistance data are calibrated in combination with discharge current; establishing a backup power supply health criterion comprising a plurality of criterion types; constructing a battery health state evaluation model and a fault prediction model based on a deep learning algorithm, dynamically evaluating a battery health level and predicting a fault occurrence period; and comparing the acquired data with a criterion threshold in real time, and triggering an alarm mechanism when any criterion is detected to meet an abnormal condition. According to the invention, differential monitoring and health management of the storage battery, the lithium battery and the super-capacitor backup power supply are realized.

Description

Technical Field

[0001] This invention relates to the field of backup power health status analysis technology, specifically to backup power health status analysis methods, systems, devices, and media based on AI deep learning and battery big data processing. Background Technology

[0002] Electricity is a fundamental industry of the national economy, and the stability and reliability of power distribution networks directly affect economic development and residents' lives. As key equipment ensuring the continuous operation of the distribution network, the health status and maintenance quality of backup power supplies are of paramount importance. Currently, distribution network operation and maintenance are transforming towards digitalization and intelligence, and traditional backup power supply operation and maintenance models are no longer sufficient to meet the demands for efficient and precise operation and maintenance.

[0003] Current backup power supply maintenance relies primarily on manual inspections and reactive repairs, lacking real-time monitoring and intelligent predictive capabilities. Monitoring data is fragmented, lacking a unified integration and analysis platform, making it impossible to trace the battery's status and predict trends throughout its entire lifecycle.

[0004] The lack of accurate online internal resistance detection technology is a significant issue, as internal resistance is a core indicator of battery health. Most existing online monitoring devices only monitor voltage, ignoring changes in internal resistance. The few devices that do offer internal resistance testing typically employ a single AC injection method or a high-current instantaneous discharge method, which are susceptible to line noise interference and have poor measurement accuracy when the battery is in a float charge state, making it difficult to accurately reflect the inflection point of internal battery degradation. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention provides a method, system, device and medium for analyzing the health status of backup power sources based on AI deep learning and battery big data processing.

[0006] Therefore, the technical problem solved by this invention is: how to address the current shortcomings in backup power supply operation and maintenance, which rely primarily on manual inspections and reactive repairs, lacking real-time monitoring and intelligent predictive capabilities. Monitoring data is fragmented, lacking a unified integration and analysis platform, making it impossible to trace the battery's entire lifecycle status and predict trends.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: a backup power supply health status analysis method based on AI deep learning and battery big data processing, comprising: real-time acquisition of operating parameters of individual batteries and battery packs through the collaborative work of a sampling board and sensors, synchronously recording the timestamps and device identifiers of each parameter during the acquisition process; measuring the battery internal resistance using a detection method, calculating the battery internal resistance based on the real-time changes in discharge voltage and current, switching the detection method when the battery internal resistance is too low, and calibrating the internal resistance data in conjunction with the discharge current; establishing backup power supply health criteria including multiple criterion types; constructing a battery health status assessment model and a fault prediction model to dynamically assess the battery health status and predict the timing of fault occurrence; real-time comparison of the acquired data with the criterion thresholds, triggering an alarm mechanism when any criterion meets the abnormal conditions, and synchronously recording the abnormal parameters, occurrence time, and device location information.

[0008] As a preferred embodiment of the backup power health status analysis method based on AI deep learning and battery big data processing described in this invention, the method involves: real-time acquisition of operating parameters of individual batteries and battery packs through the collaborative work of a sampling board and sensors, with the timestamps and device identifiers of each parameter recorded synchronously during the acquisition process. This includes: connecting the battery to the sensors; converting the battery data from analog to digital signals through the sampling board; achieving indiscriminate monitoring of individual batteries through the storage of digital signals; real-time acquisition of operating parameters of individual batteries and battery packs, comprehensively covering battery operating status characteristics; and automatically performing sensor zero drift compensation and temperature and humidity adaptive gain adjustment after the device is powered on, correcting parameters by comparing the detected ambient temperature and humidity with the battery operating temperature and combining pre-calibrated temperature change parameters.

[0009] As a preferred embodiment of the backup power health status analysis method based on AI deep learning and battery big data processing described in this invention, the method involves: measuring the battery internal resistance using a detection method; calculating the battery internal resistance based on the real-time changes in discharge voltage and current; switching the detection method when the battery internal resistance is too low; and calibrating the internal resistance data in conjunction with the discharge current. This includes: dividing the battery pack into multiple cycles; discharging sequentially according to the cycle order when measuring internal resistance; synchronously acquiring the battery discharge curve; calculating the initial battery internal resistance based on the collected discharge voltage drop data; analyzing and calculating the battery internal resistance based on the real-time changes in discharge voltage and current; automatically switching the detection method when the battery internal resistance is too low; controlling the discharge load to ensure the continuous stability of the discharge current; and calculating the average value of the collected data after multiple measurements to derive stable internal resistance data.

[0010] As a preferred embodiment of the backup power health status analysis method based on AI deep learning and battery big data processing described in this invention, the following steps are included: establishing backup power health criteria including multiple criteria types, including establishing health criteria for different types of backup power; the criteria types in the health criteria include criteria based on battery operating parameters; setting corresponding threshold ranges for each criterion type; and combining the health criteria to comprehensively evaluate the health status of the backup power.

[0011] As a preferred embodiment of the backup power health status analysis method based on AI deep learning and battery big data processing described in this invention, the following steps are included: constructing a battery health status assessment model and a fault prediction model to dynamically assess battery health status and predict the timing of fault occurrence. This includes: transmitting data to a temporary database; automatically filtering and removing erroneous data; transmitting the remaining data to a core database; calculating battery operating data using a balancing algorithm; correcting parameters between the operating data and preset data; comparing the corrected data with standard battery operating data to obtain the final operating condition; issuing alarms for batteries with abnormal internal resistance and abnormal capacity; and when the calculated value does not match the actual value, entering actual data for further comparison and analysis, calculating the difference between the erroneous data and the actual data, adjusting the data based on the original preset data, recording the analysis results, and updating the preset database.

[0012] This preferred solution establishes a two-layer data screening mechanism consisting of a temporary database and a core database, combines a balancing algorithm to correct parameters of running data and preset data, and employs a self-learning update mechanism based on actual data feedback. This enables continuous optimization of the AI ​​model's prediction accuracy, effectively reduces the misjudgment rate caused by environmental differences, and ensures the accuracy of health status assessment under different operating conditions.

[0013] As a preferred embodiment of the backup power health status analysis method based on AI deep learning and battery big data processing described in this invention, the following steps are included: when the calculated value does not match the actual value, the actual data is entered for further comparison and judgment; the difference between the erroneous data and the actual data is calculated; adjustments are made based on the original set data; the judgment results are recorded and the preset database is updated. This includes: real-time comparison of the test values ​​of the operating conditions and the actual operating conditions; for alarm information indicating an error in the judgment that the calculated value has not been reached, the actual data is entered and the user clicks "return to verification"; the difference between the erroneous data referenced in the calculation and the entered actual data is calculated; adjustments are made based on the original set data; the judgment results are automatically recorded, and the results are input into the preset database to replace the original set data; when the same type of battery operating data reappears, new data is used for comparison and correction.

[0014] This preferred solution achieves adaptive learning and automatic parameter correction of the AI ​​model by calculating the difference between the erroneous data used in the calculation and the actual data entered, and automatically adjusting the preset database parameters. This avoids the recurrence of the same type of misjudgment and improves the long-term accuracy and environmental adaptability of the backup power health status assessment.

[0015] As a preferred embodiment of the backup power health status analysis method based on AI deep learning and battery big data processing described in this invention, the real-time comparison of collected data with criterion thresholds triggers an alarm mechanism when any criterion meets an abnormal condition, synchronously recording abnormal parameters, occurrence time, and device location information. This includes automatically issuing alarm prompts and battery replacement warnings when the model predicts abnormal battery health status, lifespan degradation approaching a critical value, or a fault occurring; continuously issuing alarm prompts on the monitoring platform, requiring manual confirmation from the operation and maintenance management platform and relevant personnel to reset the alarm information; developing personalized maintenance plans for each battery based on the battery health status assessment results; and encrypting the collected data, health status assessment results, and alarm data via a communication module before uploading them to the main station.

[0016] This preferred solution, through a mandatory confirmation mechanism that continuously alerts the monitoring platform and requires manual confirmation for recovery, combined with encrypted data upload and personalized maintenance plan generation, ensures that maintenance personnel do not miss critical alarm information, avoids the escalation of backup power failures due to untimely alarm handling, and achieves closed-loop management of alarm handling.

[0017] This invention provides a backup power health status analysis system based on AI deep learning and battery big data processing.

[0018] To address the aforementioned technical problems, the present invention provides the following technical solution: a backup power health status analysis system based on AI deep learning and battery big data processing, comprising:

[0019] The present invention provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the backup power health status analysis method based on AI deep learning and battery big data processing.

[0020] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the backup power health status analysis method based on AI deep learning and battery big data processing.

[0021] The beneficial effects of this invention are as follows: This invention addresses the practical problems of difficulty in accurately assessing the health status of backup power supplies in power distribution terminals, high misjudgment rates caused by the mixed use of criteria for different types of backup power supplies, and lack of targeted operation and maintenance. By establishing a full-process analysis architecture, it achieves differentiated and accurate monitoring and health management of three types of backup power supplies: storage batteries, lithium batteries, and supercapacitors. Among these features, an internal resistance detection mechanism that adaptively switches between DC discharge and ripple methods solves the problem of insufficient measurement accuracy for backup power supplies with low internal resistance. By establishing a health criterion model for different types of backup power supplies, differentiated thresholds are set for the voltage characteristics, internal resistance variation patterns, and capacity decay characteristics of batteries, lithium batteries, and supercapacitors, avoiding misjudgments and omissions caused by uniform criteria. A self-learning update mechanism based on deep learning algorithms automatically corrects preset parameters through feedback from actual backup power supply operation data, achieving continuous optimization of the health status assessment accuracy for different types of backup power supplies and effectively reducing the misjudgment rate caused by differences in backup power supply operating environments and equipment aging. A mandatory confirmation alarm mechanism ensures that maintenance personnel do not miss key backup power supply alarm information, preventing the expansion of backup power supply faults and power loss in distribution terminals due to untimely alarm handling. This achieves intelligent closed-loop management of the entire process from backup power supply data collection, health analysis, abnormal alarms to maintenance and handling, significantly improving the accuracy of backup power supply maintenance and power supply reliability in distribution terminals. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the overall process of a backup power health status analysis method based on AI deep learning and battery big data processing, as provided in one embodiment of the present invention. Detailed Implementation

[0024] To make the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a backup power health status analysis method based on AI deep learning and battery big data processing, including: S1. By working in conjunction with the sampling board and sensors, the operating parameters of individual cells and battery packs are collected in real time. During the collection process, the timestamps and device identifiers of each parameter are recorded simultaneously.

[0026] S2. The battery internal resistance is measured using a detection method. The battery internal resistance is calculated based on the real-time changes in discharge voltage and current. When the battery internal resistance is too low, the detection method is switched, and the internal resistance data is calibrated in conjunction with the discharge current.

[0027] S3. Establish backup power health criteria that include multiple criterion types.

[0028] S4. Construct a battery health status assessment model and a fault prediction model to dynamically assess the battery health status and predict the timing of fault occurrence.

[0029] S5. Real-time comparison of collected data with criterion thresholds. When any criterion is detected to meet abnormal conditions, an alarm mechanism is triggered, and abnormal parameters, occurrence time, and device location information are recorded synchronously.

[0030] It should be noted that during operation, the voltage, internal resistance, and capacity of backup power supplies (batteries, lithium batteries, supercapacitors) vary depending on factors such as ambient temperature changes, load fluctuations, and charge / discharge cycles. Traditional monitoring methods, which use uniform criteria, cannot accurately reflect the health status of different types of backup power supplies, leading to a high rate of false positives and false negatives. Furthermore, the complex operating environment of backup power supplies, including electromagnetic interference and temperature fluctuations, affects data acquisition accuracy. High or low temperatures significantly impact internal resistance measurements and can also damage backup power supplies due to capacity decay. Therefore, health monitoring and prediction of backup power supplies are crucial.

[0031] Therefore, to address the aforementioned issues in operation monitoring and health prediction, steps S1-S5 are used to collect backup power supply operating parameters in real time and perform adaptive calibration. This yields health criterion models and graded alarm mechanisms for different types of backup power supplies, enabling accurate calculation of internal resistance and capacity. Real-time monitoring of backup power supply health status is achieved, dynamically assessing backup power supply health levels and providing early warnings for abnormal backup power supply conditions. Simultaneously, based on deep learning algorithms and a self-learning update mechanism, accurate prediction of backup power supply failure timing is realized.

[0032] Example 2, an embodiment of the present invention, provides a backup power health status analysis method based on AI deep learning and battery big data processing, based on the previous embodiment, including: In this embodiment of the application, the sensor in step S1 can be a voltage sensor, a current sensor, or a temperature sensor. The battery is connected to the voltage sensor, current sensor, and temperature sensor. The battery data is converted from analog to digital by the sampling board, and the analog signal is converted into a digital signal. The core parameters such as the open circuit voltage, discharge voltage, discharge current, battery internal resistance, single cell temperature, and ambient temperature of the individual cells and battery pack are collected in real time, comprehensively covering the characteristics of the battery operating status.

[0033] In an alternative implementation, the sensor can also be a Hall sensor, which detects the current magnitude through the Hall effect to achieve non-contact current measurement and reduce physical interference to the battery pack.

[0034] In another alternative implementation, the sensor can also be an infrared temperature sensor, which collects the surface temperature distribution of the battery through non-contact infrared temperature measurement to achieve rapid scanning and monitoring of the battery temperature field.

[0035] This invention achieves accurate conversion of analog signals to digital signals through the coordinated operation of voltage sensors, current sensors, temperature sensors, and sampling boards. It ensures the real-time and accurate acquisition of core parameters such as open-circuit voltage, discharge voltage, discharge current, battery internal resistance, cell temperature, and ambient temperature, comprehensively covering the characteristics of battery operating status and providing reliable basic data for subsequent internal resistance calculation and health status analysis.

[0036] Furthermore, in step S1, the sampling board works in conjunction with the sensor to collect the operating parameters of individual cells and battery packs in real time. During the collection process, the timestamps and device identifiers of each parameter are recorded synchronously, including the following steps A1-A4: A1. The battery is connected to the sensor, and the sampling board performs analog-to-digital conversion on the battery data, converting the analog signal into a digital signal.

[0037] A2. By storing digital signals, non-discriminatory monitoring of individual batteries can be achieved.

[0038] A3. Real-time acquisition of operating parameters of individual cells and battery packs, comprehensively covering battery operating status characteristics.

[0039] A4. After the device is powered on, it automatically performs sensor zero drift compensation and temperature and humidity adaptive gain adjustment. It corrects parameters by detecting ambient temperature and humidity and comparing them with the battery operating temperature, combined with pre-calibrated temperature change parameters.

[0040] In this embodiment of the application, the operating parameters in step A3 can be open circuit voltage, discharge voltage, discharge current, battery internal resistance, single cell temperature, and ambient temperature. The sampling board works in conjunction with voltage sensors, current sensors, and temperature sensors to collect the open circuit voltage, discharge voltage, discharge current, battery internal resistance, single cell temperature, and ambient temperature of individual cells and battery packs in real time. During the collection process, the timestamps and device identifiers of each parameter are recorded simultaneously to comprehensively cover the characteristics of battery operating status.

[0041] In one alternative implementation, the operating parameters can also be charging voltage, charging current, battery capacity, state of charge (SOC), and state of health (SOH). By expanding the range of collected parameters, the operating status of the backup power supply can be reflected more comprehensively.

[0042] In another alternative implementation, the operating parameters can also be the battery's internal temperature gradient, voltage fluctuation amplitude, current response speed, and self-discharge rate. By collecting more refined dynamic parameters, minute abnormal changes in the backup power supply can be captured.

[0043] This invention achieves comprehensive coverage and data traceability of backup power supply operating status by real-time acquisition of core operating parameters such as open-circuit voltage, discharge voltage, discharge current, battery internal resistance, cell temperature, and ambient temperature, and simultaneously records the timestamp and device identifier of each parameter. This provides multi-dimensional, high-precision, and traceable basic data support for subsequent internal resistance calibration, health criterion comparison, and AI model analysis.

[0044] Specifically, the battery is connected to voltage, current, and temperature sensors. A sampling board converts the battery data from analog to digital signals. By storing these digital signals, indiscriminate monitoring of individual 12V batteries is achieved. Because the signal is converted to digital, the modular design allows for arbitrary selection of installation locations. Real-time monitoring can be completed simply by connecting the measurement signal lines to the corresponding interfaces on the module.

[0045] It collects key operating parameters of individual cells and battery packs in real time, including: open circuit voltage, discharge voltage, discharge current, battery internal resistance, cell temperature, ambient temperature, etc., comprehensively covering the characteristics of battery operating status.

[0046] After power-on, the device automatically performs sensor zero-drift compensation and temperature and humidity adaptive gain adjustment. This is mainly achieved by comparing the ambient temperature with the battery operating temperature and correcting parameters using pre-calibrated temperature change parameters. This ensures data acquisition accuracy under different environmental conditions, as shown below: in, This is the relative humidity value after temperature compensation. This is the original relative humidity value. This is the primary temperature compensation coefficient. This is the original temperature value. This is the secondary temperature compensation coefficient. The reference temperature value is used; the compensation coefficient is set based on historical data.

[0047] The raw data is denoised, primarily to remove outliers and missing values ​​from abnormal data generated by external switching actions, such as abnormal voltage increases and current surges caused by electromagnetic interference. The average value of multiple data acquisitions is used to optimize the data calculation, and the preset quantity, voltage, and current methods are further corrected based on the calculation results to ensure the reliability of the basic data for analysis. This is represented as follows: in, For a moment Battery terminal voltage, This is the initial voltage value. This is the steady-state voltage value. This is the average internal resistance value. This is the equivalent capacitance.

[0048] Real-time monitoring and recording of internal resistance change rate, voltage stability index, charge-discharge cycle life, temperature response coefficient, etc. during operation are performed. Abrupt changes affecting the data are eliminated, and continuous or intermittent changes are specially calibrated and stored. The calibrated values ​​are automatically extracted as key features of battery health status, providing effective data for the model as input criteria.

[0049] The pre-processed data is uploaded to the cloud platform in real time via wireless transmission such as Wi-Fi / 5G, while critical data is cached locally to achieve dual data backup and resume transmission from interruption.

[0050] In this embodiment, the detection method in step S2 can be either DC discharge or ripple method. The battery pack is divided into multiple cycles, and the internal resistance is measured sequentially according to the cycle order. The discharge curve of each battery is collected synchronously. The initial internal resistance of the battery is calculated based on the collected data of the discharge voltage drop. The internal resistance of the battery is calculated by analyzing the real-time changes in the collected discharge voltage and current. When the internal resistance of the battery is too low, the software automatically switches to ripple method detection. The difference between the initial voltage value and the voltage after discharge is compared, and the internal resistance data is calibrated in combination with the discharge current to ensure that the calculated internal resistance of the battery is the actual internal resistance of the battery in operation.

[0051] In an alternative implementation, the detection method can also be the AC impedance method, which involves applying a small-amplitude AC signal to the battery and measuring the battery's impedance response at different frequencies to obtain the battery's internal resistance characteristics.

[0052] In another alternative implementation, the detection method can also be the open-circuit voltage recovery method, which involves measuring the voltage recovery curve during the resting period after battery discharge and calculating the battery's internal resistance based on the voltage recovery rate.

[0053] This invention solves the problem of insufficient measurement accuracy of backup power supplies with low internal resistance by using an adaptive switching mechanism between DC discharge method and ripple method. When the battery internal resistance is too low, the detection method is automatically switched, and the internal resistance data is calibrated by combining the discharge current, which ensures the accuracy of internal resistance measurement of backup power supplies with different internal resistance ranges and provides a reliable internal resistance data basis for the health status assessment of backup power supplies.

[0054] Furthermore, in step S2, a detection method is used to measure the battery's internal resistance. The battery's internal resistance is calculated based on the real-time changes in discharge voltage and current. When the battery's internal resistance is too low, the detection method is switched, and the internal resistance data is calibrated in conjunction with the discharge current, including the following steps B1-B4: B1. Divide the battery pack into multiple cycles, and discharge them sequentially according to the cycle order when measuring the internal resistance, and collect the discharge curve of the battery simultaneously.

[0055] B2. Calculate the initial internal resistance of the battery based on the data collected from the discharge voltage drop.

[0056] B3. Collect and analyze the real-time changes in discharge voltage and current to calculate the battery's internal resistance. When the battery's internal resistance is too low, automatically switch the detection method.

[0057] B4. Control the discharge load to ensure the continuous stability of the discharge current. After multiple measurements, the processor calculates the average value of the collected data and derives the stable internal resistance data.

[0058] Specifically, a combined detection technique is adopted, with DC discharge method as the main method and ripple method as the auxiliary method.

[0059] Since each battery is equipped with an independent voltage detection line, the battery pack can be divided into multiple cycles through program control. When measuring internal resistance, the batteries are discharged sequentially according to the cycle order. Simultaneously, the discharge curve of each battery (or each group) is collected according to the battery (or group) allocated by the program control. Then, the initial internal resistance of the battery (or each group) is calculated by collecting data through the discharge voltage drop of the battery (or group). in, The initial internal resistance of the battery. This is the battery terminal voltage before discharge begins. This represents the battery terminal voltage during the discharge process. This is the discharge current value.

[0060] The battery internal resistance is calculated by analyzing the real-time changes in the collected discharge voltage and current. When the battery internal resistance is too low (10 milliohms), the system automatically switches to ripple detection. The difference between the initial voltage value and the voltage after discharge is compared, and the internal resistance data is calibrated in conjunction with the discharge current to ensure that the calculated battery internal resistance value is the actual operating battery internal resistance.

[0061] By controlling the parallel resistance of the discharge load, the continuous stability of the discharge current is ensured. After multiple measurements, the average value of the collected data is calculated by the processor, and the stable internal resistance data is derived as follows: in, This is the average internal resistance value. The initial internal resistance value obtained from the first measurement. For the first The initial internal resistance value obtained from this measurement For the number of times measured.

[0062] Furthermore, in step S3, backup power health criteria including multiple criterion types are established, including the following steps C1-C4: C1. Establish health criteria for different types of backup power supplies.

[0063] C2. The types of criteria in the health criteria include those based on battery operating parameters.

[0064] C3. Set the corresponding threshold range for each criterion type.

[0065] C4. The combined application of health criteria comprehensively assesses the health status of backup power supplies.

[0066] In this embodiment of the application, the criterion type in step C1 can be internal resistance criterion, voltage criterion, capacity criterion, or temperature criterion. Health criteria are established for different types of backup power supplies. Battery health criteria include: internal resistance criterion (static internal resistance greater than a set initial value or the difference in internal resistance between individual cells in the same group greater than a set threshold); voltage criterion (individual cell open-circuit voltage less than a set voltage value or group open-circuit voltage less than a set voltage value); capacity criterion (backup discharge time less than a set design value or annual capacity decay rate greater than a set value); and temperature criterion (charge / discharge temperature greater than a set high temperature threshold or less than a set high temperature threshold). The following criteria are used to determine the health of lithium batteries: a low-temperature threshold is set; the health criteria for lithium batteries include: a voltage criterion (single cell voltage greater than a set overcharge voltage or less than a set over-discharge voltage); an internal resistance criterion (operating internal resistance greater than a set initial value by a certain percentage); a capacity criterion (actual backup capacity less than a set rated capacity by a certain percentage); and a temperature criterion (charge / discharge temperature greater than a set high-temperature threshold or less than a set low-temperature threshold). The health criteria for supercapacitors include: a capacity criterion (actual capacity less than a set rated capacity by a certain percentage); an internal resistance criterion (equivalent series resistance greater than a set initial value by a certain percentage); and a voltage retention rate less than a set value after a set time of rest following charge / discharge at rated voltage.

[0067] In one alternative implementation, the criterion type can also be cycle life criterion, charge / discharge rate criterion, or self-discharge rate criterion. By expanding the criterion types, the health status and service life of the backup power supply can be assessed more comprehensively.

[0068] In another alternative implementation, the criterion type can also be voltage consistency criterion, temperature uniformity criterion, or internal resistance dispersion criterion. By evaluating the consistency between individual cells in the battery pack, potential health hazards caused by excessive differences between individual cells can be detected in a timely manner.

[0069] This invention establishes a categorized backup power supply health criterion model that includes internal resistance criteria, voltage criteria, capacity criteria, and temperature criteria. It sets differentiated threshold ranges for the voltage characteristics, internal resistance variation patterns, and capacity decay characteristics of batteries, lithium batteries, and supercapacitors, avoiding misjudgments and omissions caused by uniform criteria. This enables accurate assessment of the health status of different types of backup power supplies and provides accurate criteria for intelligent anomaly alarm triggering and operation and maintenance decisions.

[0070] Specifically, the health criteria for batteries are as follows: Internal resistance criteria: Static internal resistance > 150% of the initial value, or the difference in internal resistance between individual cells in the same group > 20%; Voltage criteria: Open circuit voltage of 12V cells < 12.0V, open circuit voltage of 24V group < 24.0V, discharge termination voltage < 10.5V (12V) / 21.0V (24V) respectively, or voltage fluctuation during charging and discharging > 0.5V (excluding interference from the start and stop of terminal equipment); Capacity criteria: Backup discharge time < 80% of the design value, or annual capacity decay rate > 5%; Temperature and operating condition criteria: Outdoor terminal battery charging and discharging temperature > 45℃ or < -10℃, temperature difference between cells in the same group > 5℃, or abnormal heating under float charging conditions.

[0071] Lithium battery health criteria: Voltage criteria: Single cell voltage > 3.65V (overcharge) or < 2.5V (over-discharge), voltage difference between single cells > 0.05V under static state, or voltage drop > 0.1V / 24 hours during terminal standby; Internal resistance criteria: Operating internal resistance > 120% of initial value, internal resistance dispersion of terminal batteries in batch deployment > 15%; Capacity criteria: Actual backup capacity < 80% of rated capacity, or capacity decay of a single charge / discharge cycle > 3%; Temperature and safety criteria: Charge / discharge temperature > 60℃ or < -20℃, bulging, smoke, or BMS reporting overcurrent / overtemperature faults; Operating condition compatibility criteria: Charging efficiency < 90% under normal AC power supply, or discharge capacity < 60% of rated value under low temperature (< -10℃) environment.

[0072] Supercapacitor health criteria: Capacity criterion: Actual capacity < 85% of rated capacity, or capacity decay > 10% after 1000 charge-discharge cycles; Internal resistance criterion: Equivalent series resistance (ESR) > 200% of initial value, or ESR difference > 30% among capacitors in the same batch; Voltage criterion: Voltage retention rate < 95% after 1 hour of standing at rated voltage, or overvoltage > 110% of rated value; Leakage current criterion: Leakage current > 150% of manufacturer's specified value at rated voltage, or excessive self-discharge during terminal standby; Operating condition compatibility criterion: Power supply voltage drop > 0.3V during terminal start-up and shutdown, or decreased charge-discharge response speed, and casing temperature > 50℃.

[0073] The threshold is set based on historical data and device parameters.

[0074] Furthermore, in step S4, a battery health status assessment model and a fault prediction model are constructed to dynamically assess the battery health status and predict the timing of fault occurrence, including the following steps D1-D5: D1. Transfer the data to a temporary database, automatically filter and remove erroneous data, and transfer the remaining data to the core database.

[0075] D2. Calculate the battery operating data using a balancing algorithm, and then correct the parameters by comparing the operating data with the preset data.

[0076] D3. The corrected data is compared with the standard battery operating data to obtain the final operating conditions.

[0077] D4. Issue an alarm for batteries with abnormal internal resistance and abnormal capacity.

[0078] D5. When the calculated value does not match the actual value, enter the actual data for comparison and analysis again, calculate the difference between the erroneous data and the actual data, adjust it in conjunction with the original set data, record the analysis results and update the preset database.

[0079] Furthermore, in step D5, when the calculated value does not match the actual value, the actual data is entered for further comparison and analysis. The difference between the erroneous data and the actual data is calculated, and adjustments are made based on the original set data. The analysis results are recorded and the preset database is updated, including the following steps D51-D56: D51. Compare the test values ​​of the operating conditions with the actual operating conditions in real time.

[0080] D52. For alarm messages indicating that the calculated value has not actually been reached, enter the actual data and click "return to verify".

[0081] D53. Calculate the difference between the erroneous data used in the calculation and the actual data entered.

[0082] D54. Add or subtract data based on existing settings.

[0083] D55. Automatically record the analysis results and input the results into the preset database to replace the original set data.

[0084] D56. When the same type of battery operation data appears again, use the new data for comparison and correction.

[0085] Specifically, the data is mainly transmitted to a temporary database through the data acquisition module at the grassroots level. The data is then automatically filtered to remove erroneous data such as current, voltage, and temperature that occurred accidentally during multiple tests or that were caused by circuit breaker operation during the test. The remaining data is transmitted to the core database, where the battery operating data is calculated using a balancing algorithm. The operating data is then compared with the preset data to correct the parameters. Finally, the corrected data is compared with the standard battery operating data to obtain the final operating condition.

[0086] The battery health status assessment model is built upon a Long Short-Term Memory (LSTM) network, using average internal resistance as the core indicator. This primarily involves the processor filtering out numerical values ​​after analog-to-digital conversion. The input features of the deep learning model include time-series data on battery operating internal resistance, supplemented by operating voltage, current, and temperature parameters related to internal resistance calculation and calibration as auxiliary features. The input layer receives data with a length of [length missing]. Time series input: in, For the input feature set, for Input characteristics at any given time.

[0087] The feature extraction layer (LSTM layer) is used to extract long-term dependency features of battery operating parameters over time. Its state update formula is: in, Output the forget gate for the current time step. It is the Sigmoid activation function. Here is the forget gate weight matrix. This is the hidden layer output of the LSTM unit from the previous time step. Forget gate bias vector, The current time step is the input gate output. The input gate weight matrix, This is the input gate bias vector. The candidate cell state vector generated at the current time step. This is the weight matrix corresponding to the candidate unit state. This is the bias vector corresponding to the state of the candidate unit. This is the updated cell state vector at the current time step. This is the cell state vector updated in the previous time step. This is an element-wise multiplication operation. Output the gate value at the current time step. This is the output gate weight matrix. This is the output gate bias vector. The hidden layer output is the output of the LSTM unit at the current time step.

[0088] Mapping the LSTM output to the continuous health indicator SOH is represented as follows: in, For the first Constantly monitor the battery health status. For mapping weights, This is the mapping bias.

[0089] During the model input phase, physical constraints are introduced into the internal resistance data to eliminate single-point outliers caused by sampling jitter or circuit breaker operation, ensuring stable model input.

[0090] An improved adaptive Kalman filter algorithm is introduced to correct the remaining capacity, and its state transition model is as follows: in, For the first The estimated capacity covariance, For the first The predicted value of the capacity covariance of this estimate. Here is the state transition matrix. For the first The estimated capacity covariance, The process noise covariance matrix is... This is the transpose of the state transition matrix. For Kalman gain, For the observation matrix, For the transpose of the observation matrix, To measure the noise covariance, It is an identity matrix.

[0091] The battery health status assessment model is based on time-series input features constructed with internal resistance as the core. It analyzes the internal resistance change characteristics of the battery during continuous operation. The model outputs a health assessment result that characterizes the current operating status of the battery. The output includes at least a health status value reflecting the overall performance level of the battery and the remaining usable capacity after adaptive filtering correction, which is used to characterize the actual usability of the battery under the current operating conditions.

[0092] For the fault prediction model, the input includes a multi-dimensional feature set reflecting the long-term deterioration trend of the battery. The features include at least the difference between the initial internal resistance and the current internal resistance, which characterizes the degree of increase in the battery's internal resistance relative to the initial state; the change in internal resistance, which characterizes the rate of change of internal resistance over time; the battery health status value output by the health status assessment model, which reflects the overall performance level of the battery; the current available capacity, which characterizes the actual energy supply capacity of the battery; and the operating temperature parameter, which reflects the impact of environmental and operating conditions on the battery deterioration process. These features together constitute the input basis for fault prediction and life decay analysis.

[0093] The change curve features are extracted using a Convolutional Neural Network (CNN), and the lifespan decay trend function is output. And construct a monotonically increasing constraint: in, For the first Lifetime decay index It is a convolutional neural network. The initial average internal resistance, for Average internal resistance at any time for Average internal resistance at any time, for Average internal resistance at any time, for Average internal resistance at any given time.

[0094] The minimum safe running time output by the model is expressed as: in, For minimum safe operating time, The fault determination threshold is set based on historical data.

[0095] The fault prediction model is based on the difference between the initial internal resistance and the current internal resistance of the battery and its changing trend. It models and analyzes the battery's operational degradation process. The model outputs prediction results to characterize the future operational risks of the battery. The output includes at least the battery's lifespan degradation trend index and the corresponding minimum safe operating time, which is used to identify the battery's degradation inflection point and potential failure period in advance.

[0096] The training loss function is defined as: in, For loss function, For the first Real-time battery health status value. For the first Real-time lifespan degradation index To estimate the weighting coefficients of the error term for SOH, The weighting coefficients for the prediction error term are set based on historical data.

[0097] During model training, the constructed loss function is used as the optimization objective, and the Adam optimization algorithm is employed to iteratively update the trainable parameters in the model. These trainable parameters include the weights and biases of each network layer. In each training iteration, the learning rate of each parameter is adaptively adjusted based on the gradient information of the loss function with respect to the trainable parameters, thus completing the parameter update. The weight coefficients in the loss function... and The hyperparameters, which are pre-set based on historical data, remain unchanged during training. When the model training reaches the preset maximum number of iterations, or when the change in the loss function in adjacent iterations is less than a set threshold, the updating of the model parameters stops, and the model training is completed.

[0098] Furthermore, in step S5, the collected data is compared with the criterion threshold in real time. When any criterion is detected to meet the abnormal condition, an alarm mechanism is triggered, and the abnormal parameters, occurrence time, and device location information are recorded synchronously, including the following steps E1-E4: E1. When the model predicts that the battery health status is abnormal, the lifespan is approaching the critical value, or a fault occurs, the system will automatically issue an alarm and a battery replacement warning.

[0099] E2. The monitoring platform continuously sends alarm prompts, and the operation and maintenance management platform and relevant personnel need to manually confirm whether the alarm information can be restored.

[0100] E3. Based on the battery health status assessment results, develop a personalized maintenance plan for each battery.

[0101] E4. The collected data, health status assessment results, and alarm data are encrypted and uploaded to the main station through the communication module.

[0102] Specifically, okay, I will explain the content of the S5 steps in a few paragraphs: S5 Step-by-Step Instructions The system continuously acquires pre-processed and AI model-analyzed data from the data acquisition module, including corrected internal resistance, voltage, capacity, and temperature values, as well as health status assessment results and fault prediction results. It then compares these real-time acquired data with health criterion models for different types of backup power supplies, retrieved from a pre-configured database, to determine whether the internal resistance is below the lower threshold or above the upper threshold, whether the voltage is overcharged or over-discharged, whether the capacity decay exceeds the set ratio, and whether the temperature exceeds the high temperature threshold or falls below the low temperature threshold.

[0103] When any criterion is detected to meet the abnormal condition, the system immediately marks the criterion as an abnormal state. When the model predicts that the battery health status is abnormal, the lifespan is approaching the critical value, or faults such as overvoltage or undervoltage occur, the system determines the alarm level (early warning, warning, emergency alarm) according to the severity of the abnormality, automatically triggers the alarm mechanism and issues alarm prompts and battery replacement warnings, and generates alarm information including the abnormal parameter name, the actual value of the abnormal parameter, the criterion threshold, and the degree of deviation.

[0104] Synchronously record abnormal parameters, occurrence time, and equipment location information. Record the current value, historical change curve, threshold exceedance, and change rate of abnormal parameters to the alarm database. Record the precise timestamp of the alarm occurrence. Record equipment location information such as the power distribution terminal equipment number, geographical coordinates, and the power supply line number to which it belongs. Link and store the abnormal parameters, occurrence time, and equipment location information to form a complete alarm record.

[0105] The monitoring platform continuously sends alarm notifications through various means, including flashing alarm messages, continuous alarm sound playback, periodic alarm pop-ups, and SMS or app push notifications. The maintenance management platform and relevant personnel must manually confirm whether the alarm information can be reset to ensure no critical alarm is missed. Based on the battery health status assessment results, the system develops a personalized maintenance plan for each battery, including maintenance time, maintenance content, and spare parts configuration suggestions. Maintenance personnel conduct on-site verification by retrieving the suggested messages during routine inspections.

[0106] The collected data, health status assessment results, and alarm data are packaged and encrypted using an encryption module. Supporting national cryptographic algorithms ensures data transmission security. The encrypted data is then uploaded to the distribution network automation master station via Wi-Fi or 5G communication. The master station's human-machine interface display module shows real-time abnormal alarm information, health status levels, abnormal parameter details, and device locations. Monitoring personnel can promptly issue inspection and handling instructions to on-site maintenance personnel based on the visualized data, forming a closed-loop management system for alarm handling, from alarm triggering, information recording, platform display, maintenance scheduling to handling feedback.

[0107] Example 3 is an embodiment of the present invention. This embodiment provides a backup power health status analysis system based on AI deep learning and battery big data processing, including a data acquisition module, an internal resistance detection and calibration module, a health criterion construction module, an analysis module, and an alarm triggering module.

[0108] The data acquisition module is used to work in conjunction with the sampling board and sensors to collect the operating parameters of individual cells and battery packs in real time. During the acquisition process, the timestamps and device identifiers of each parameter are recorded simultaneously.

[0109] The internal resistance detection and calibration module is used to measure the battery's internal resistance using detection methods. It calculates the battery's internal resistance based on the real-time changes in discharge voltage and current. When the battery's internal resistance is too low, it switches the detection method and calibrates the internal resistance data in conjunction with the discharge current.

[0110] The health criterion construction module is used to establish backup power health criteria that include multiple criterion types.

[0111] The analysis module is used to build battery health status assessment models and fault prediction models, dynamically assess battery health status and predict the timing of fault occurrence.

[0112] The alarm triggering module is used to compare the collected data with the criterion thresholds in real time. When any criterion is detected to meet the abnormal conditions, the system triggers the alarm mechanism and records the abnormal parameters, occurrence time and device location information simultaneously.

[0113] This embodiment also provides an electronic device applicable to the backup power health status analysis method based on AI deep learning and battery big data processing, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the backup power health status analysis method based on AI deep learning and battery big data processing proposed in the above embodiment.

[0114] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the backup power health status analysis method based on AI deep learning and battery big data processing as proposed in the above embodiments.

[0115] The storage medium proposed in this embodiment and the backup power health status analysis method based on AI deep learning and battery big data processing proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0116] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0117] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A backup power health status analysis method based on AI deep learning and battery big data processing, characterized by: include, By working in conjunction with the sampling board and sensors, the operating parameters of individual cells and battery packs are collected in real time, and the timestamps and device identifiers of each parameter are recorded synchronously during the collection process. The battery internal resistance is measured using a detection method. The battery internal resistance is calculated based on the real-time changes in discharge voltage and current. When the battery internal resistance is too low, the detection method is switched, and the internal resistance data is calibrated in conjunction with the discharge current. Establish backup power health criteria that include multiple criterion types; Construct a battery health status assessment model and a fault prediction model to dynamically assess battery health status and predict the timing of fault occurrence. The system compares the collected data with the threshold criteria in real time. When any criterion is detected to meet the abnormal conditions, an alarm mechanism is triggered, and the abnormal parameters, occurrence time, and device location information are recorded simultaneously.

2. The backup power health status analysis method based on AI deep learning and battery big data processing as described in claim 1, characterized in that: The system utilizes a sampling plate and sensors to work together to collect real-time operating parameters of individual cells and battery packs. During the collection process, timestamps and device identifiers for each parameter are recorded simultaneously. The battery is connected to a sensor, and the sampling board performs analog-to-digital conversion on the battery data, converting the analog signal into a digital signal. By storing digital signals, indiscriminate monitoring of individual battery cells can be achieved. Real-time acquisition of operating parameters of individual cells and battery packs, comprehensively covering battery operating status characteristics; After the device is powered on, it automatically performs sensor zero drift compensation and temperature and humidity adaptive gain adjustment. It corrects parameters by detecting ambient temperature and humidity and comparing them with the battery operating temperature, combined with pre-calibrated temperature change parameters.

3. The backup power health status analysis method based on AI deep learning and battery big data processing as described in claim 2, characterized in that: The method employs a detection method to measure the battery's internal resistance, calculating the internal resistance based on real-time changes in discharge voltage and current. When the battery's internal resistance is too low, the detection method is switched, and the internal resistance data is calibrated using the discharge current. The battery pack is divided into multiple cycles, and the internal resistance is measured by discharging the batteries sequentially in the cycle order, while the discharge curves of the batteries are collected synchronously. Calculate the initial internal resistance of the battery based on the data collected from the discharge voltage drop. The system collects and analyzes the real-time changes in discharge voltage and current to calculate the battery's internal resistance. When the battery's internal resistance is too low, the detection method is automatically switched. The discharge load is controlled to ensure the continuous stability of the discharge current. After multiple measurements, the processor calculates the average value of the collected data and derives the stable internal resistance data.

4. The backup power health status analysis method based on AI deep learning and battery big data processing as described in claim 3, characterized in that: The establishment of backup power health criteria includes multiple criterion types, including: Establish health criteria for different types of backup power supplies; The types of criteria in the health criteria include those based on battery operating parameters; Set corresponding threshold ranges for each criterion type; The combined application of health criteria allows for a comprehensive assessment of the health status of backup power supplies.

5. The backup power health status analysis method based on AI deep learning and battery big data processing as described in claim 4, characterized in that: The construction of the battery health status assessment model and fault prediction model, which dynamically assesses the battery health status and predicts the timing of fault occurrence, includes: The data is transferred to a temporary database, where it is automatically filtered to remove erroneous data, and the remaining data is transferred to the core database. The battery operating data is calculated using a balancing algorithm, and the operating data is then compared with preset data to correct the parameters. The corrected data is compared with the standard battery operating data to obtain the final operating conditions; Alarms are issued for batteries with abnormal internal resistance and abnormal capacity. When the calculated value does not match the actual value, the actual data is entered for comparison and analysis again. The difference between the erroneous data and the actual data is calculated, and adjustments are made in conjunction with the original set data. The analysis results are recorded and the preset database is updated.

6. The backup power health status analysis method based on AI deep learning and battery big data processing as described in claim 5, characterized in that: When the calculated value differs from the actual value, the actual data is entered for further comparison and analysis. The difference between the erroneous data and the actual data is calculated, and adjustments are made based on the original set data. The analysis results are recorded and the preset database is updated. include, The test values ​​of the operating condition are compared with the actual operating conditions in real time. For alarm messages indicating an incorrect judgment that the calculated value has not actually been reached, enter the actual data and click "return to verify"; Calculate the difference between the incorrect data referenced in the calculation and the actual data entered. Add or subtract data based on existing settings; Automatically record the analysis results, and input the results into a preset database to replace the original set data; When the same type of battery's operating data reappears, the new data will be used for comparison and correction.

7. The backup power health status analysis method based on AI deep learning and battery big data processing as described in claim 6, characterized in that: The real-time comparison of collected data with threshold criteria triggers an alarm mechanism when any criterion meets an abnormal condition, simultaneously recording the abnormal parameters, occurrence time, and device location information. include, When the model predicts that the battery health is abnormal, the lifespan is approaching the critical value, or a fault has occurred, the system will automatically issue an alarm and a battery replacement warning. The monitoring platform continuously sends alarm prompts, and the operation and maintenance management platform and relevant personnel need to manually confirm whether the alarm can be reset. Based on the battery health status assessment results, a personalized maintenance plan is developed for each battery. The collected data, health status assessment results, and alarm data are encrypted and then uploaded to the main station via the communication module.

8. A backup power health status analysis system based on AI deep learning and battery big data processing, employing the backup power health status analysis method based on AI deep learning and battery big data processing as described in any one of claims 1 to 7, characterized in that, include: The module includes a data acquisition module, an internal resistance detection and calibration module, a health criterion construction module, an analysis module, and an alarm triggering module. The data acquisition module is used to work in conjunction with the sampling plate and the sensor to collect the operating parameters of individual cells and battery packs in real time. During the acquisition process, the timestamps and device identifiers of each parameter are recorded synchronously. The internal resistance detection and calibration module is used to measure the battery internal resistance using a detection method, calculate the battery internal resistance based on the real-time changes in discharge voltage and current, and switch the detection method when the battery internal resistance is too low, and calibrate the internal resistance data in conjunction with the discharge current. The health criterion construction module is used to establish backup power health criteria that include multiple criterion types; The analysis module is used to build a battery health status assessment model and a fault prediction model, dynamically assess the battery health status and predict the timing of fault occurrence. The alarm triggering module is used to compare the collected data with the criterion threshold in real time. When any criterion is detected to meet the abnormal condition, the system triggers the alarm mechanism and records the abnormal parameters, occurrence time and device location information in a synchronous manner.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the backup power health status analysis method based on AI deep learning and battery big data processing as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the backup power health status analysis method based on AI deep learning and battery big data processing as described in any one of claims 1 to 7.

Images