A communication power supply battery health state evaluation method, device and computer program product

Abstract

This invention discloses a method, device, and computer program product for assessing the health status of communication power supply batteries. The method includes: a battery acquisition terminal collecting voltage, current, and temperature operating data; an edge gateway performing data preprocessing; and battery anomaly identification and tiered data uploading based on KLD divergence. An equivalent circuit model is constructed in the cloud, and battery parameters are identified by constraining the parameter identification space with edge anomaly features. The battery health status is calculated based on the battery model parameters. The cloud platform generates a dynamic equalization and floating charge control strategy based on the battery health status and anomaly features, and distributes it to the intelligent edge gateway for execution. Simultaneously, the health benchmark distribution and parameter prior library are adaptively updated, and differentiated maintenance decisions are made by combining SOH, remaining service life, and voltage range. This invention significantly reduces data transmission volume, improves the accuracy of parameter identification and SOH assessment, overcomes model inaccuracies caused by battery characteristic drift, and achieves intelligent and efficient operation and maintenance of backup batteries for communication base stations.

Description

Technical Field

[0001] This invention relates to the field of communication power system technology, and specifically to a method, apparatus, and computer program product for assessing the health status of a communication power supply battery. Background Technology

[0002] With the large-scale deployment of 5G communication networks, the coverage and number of communication base stations are growing exponentially. Communication power supplies and their backup battery packs, as the core support for the stable operation of base stations, directly determine the service continuity and emergency response capabilities of the base stations, making them critical infrastructure for ensuring uninterrupted operation of communication networks. Among these, backup battery packs, as emergency backup units for communication power, bear the important responsibility of supplying power to the core equipment of the base station during mains power outages. Their State of Health (SOH) directly affects the effectiveness of emergency power supply to the base station. Therefore, achieving accurate and real-time assessment of the health status of backup battery packs has become one of the core requirements for ensuring the stable operation and maintenance of 5G base stations.

[0003] Currently, the maintenance of backup battery packs for communication base stations mainly relies on manual periodic inspections or simple voltage threshold alarms. This maintenance model has obvious limitations: on the one hand, manual inspections are inefficient, labor-intensive, and highly dependent on the experience level of the inspectors, making it impossible to achieve real-time monitoring of battery status; on the other hand, judging battery health status solely based on voltage parameters cannot effectively reflect core characteristics such as internal polarization effects, capacity decay, and internal resistance changes, making it difficult to accurately assess the true health status of the battery. This leads to frequent problems of "failure to identify degraded batteries in a timely manner" and "misjudgment and scrapping of healthy batteries," which not only increases maintenance costs but also poses a safety hazard of base station emergency power supply failure.

[0004] To address the shortcomings of traditional maintenance methods, existing technologies are gradually adopting parameter identification methods based on Equivalent Circuit Models (ECMs). By constructing an equivalent circuit model of the battery, key parameters such as resistance and capacitance are identified, thereby enabling the assessment of the battery's State of Health (SOH) and improving the accuracy of the assessment to some extent. However, in the practical engineering applications of large-scale 5G base station deployment, this type of technology still faces several significant technical bottlenecks, specifically as follows:

[0005] First, there is a significant contradiction between computing resources and the need for real-time assessment. ECM parameter identification relies on complex optimization algorithms (such as particle swarm optimization and Kalman filtering), which require extensive iterative calculations to optimize parameters. Deploying all computational tasks on low-power embedded devices at the base station edge is insufficient due to limitations in device computing power, failing to meet the real-time iteration requirements of the algorithm. Uploading all high-frequency sampling data to the cloud for centralized processing would consume a large amount of network bandwidth, increasing network transmission costs and generating significant transmission latency, thus failing to meet the engineering requirements for online real-time assessment of battery health status.

[0006] Second, the initialization of model parameters is somewhat arbitrary, affecting the convergence performance of the algorithm. Existing ECM parameter identification algorithms typically use random initialization to determine the initial parameters of the model. However, the physical magnitudes of various parameters inside the battery (such as ohmic internal resistance, polarization resistance, polarization capacitance, etc.) differ significantly. Randomly generated initial parameters are very likely to deviate from the optimal solution range, causing the algorithm to get stuck in a local optimum or converge at an abnormally slow speed, which cannot meet the timeliness requirements of online real-time evaluation of base station battery SOH.

[0007] Third, the existing evaluation system fails to effectively utilize historical data throughout the battery's lifecycle, resulting in insufficient robustness of the evaluation results. When conducting SOH (State of Health) assessments, the current system does not fully consider the gradual evolution of battery health status. Each assessment uses a "start from zero" parameter identification mode, lacking a "hot start" mechanism based on the battery's historical optimal parameters. This leads to significant fluctuations in the evaluation results and poor adaptability to gradual parameter changes during battery aging. Consequently, the robustness and stability of the evaluation results cannot meet the requirements of engineering applications. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a method, device and computer program product for assessing the health status of communication power supply batteries, so as to improve the real-time performance, accuracy and robustness of the assessment and meet the needs of efficient operation and maintenance of backup battery packs under the large-scale deployment of communication base stations.

[0009] To address the aforementioned technical problems, this invention provides a method for assessing the health status of a communication power supply battery, comprising:

[0010] Step S1: The battery acquisition terminal collects the voltage, current and temperature operating data of the communication power backup battery pack in real time and transmits them to the intelligent edge gateway.

[0011] Step S2: The intelligent edge gateway preprocesses the operating data, identifies abnormal states based on battery voltage distribution characteristics, and executes a hierarchical data upload strategy according to the abnormal identification results. Only when abnormal polarization characteristics of the battery are identified, the waveform data of the corresponding time period is uploaded to the cloud platform.

[0012] Step S3: After receiving the abnormal waveform data, the cloud platform constructs a battery equivalent circuit model, identifies the battery model parameters based on the parameter search space of the abnormal feature constraint parameter identification algorithm based on edge side identification, and calculates the battery health status based on the battery model parameters.

[0013] In step S4, the cloud platform generates a dynamic equalization and floating charging control strategy based on the calculated battery health status and abnormal characteristics, and then sends the control strategy to the intelligent edge gateway for execution.

[0014] Preferably, in step S2, the identification of abnormal states based on battery voltage distribution characteristics specifically includes:

[0015] The intelligent edge gateway extracts the preprocessed voltage sequence through a FIFO sliding window and performs Z-Score standardization on the voltage sequence to eliminate the influence of voltage amplitude at different charging and discharging stages.

[0016] A continuous probability distribution model of voltage sequence is constructed by using Gaussian kernel density estimation combined with Silverman empirical rule to adaptively calculate bandwidth.

[0017] The current probability distribution model is compared with the locally stored health baseline distribution model, and the Kullback-Leibler divergence is calculated as the basis for judging battery polarization anomalies and consistency degradation.

[0018] Preferably, the hierarchical data upload strategy is an edge-cloud collaborative event-driven mechanism:

[0019] When the KLD divergence is lower than the preset low threshold, it enters silent mode, only storing the timestamp, voltage extreme value, and divergence value locally, without uploading the original waveform;

[0020] When the KLD divergence is higher than the preset low threshold, the trigger mode is entered, the complete voltage, current and temperature waveforms before and after the anomaly are frozen for a preset period of time, and after adding the abnormal polarization tag, they are uploaded to the cloud platform through the MQTT high priority channel.

[0021] Preferably, step S3 specifically includes:

[0022] A second-order Thevenin equivalent circuit model is established, and the discretized state-space equations of the model are constructed based on Kirchhoff's voltage law.

[0023] Using KLD divergence as a constraint, a divergence-parameter space mapping mechanism is constructed to dynamically constrain the search interval of the Ohmic internal resistance of the particle swarm optimization (PSO) algorithm.

[0024] Using the root mean square error (RMSE) between the model output voltage and the measured voltage as the fitness function, the global optimal parameters are output through PSO iterative optimization.

[0025] The state of health (SOH) of the battery is calculated using a combined capacity-internal resistance criterion.

[0026] Preferably, the calculation of battery health state (SOH) using the capacity-internal resistance joint criterion specifically includes: using the ratio of the actual full-charge capacity to the rated capacity obtained by the ampere-hour integration method as the main criterion for SOH, and using the internal resistance attenuation ratio as the auxiliary criterion for SOH, and comprehensively outputting the battery health state (SOH).

[0027] Preferably, in step S4, the dynamic equalization and floating charge control strategy is a two-dimensional adaptive control process:

[0028] The float charge voltage is adaptively adjusted according to the SOH value. When the SOH value is lower than the preset aging threshold, the float charge voltage is linearly reduced as the SOH value decreases.

[0029] Based on the KLD divergence combined with the hysteresis comparison mechanism, the triggering, holding and exit control of equalization charging is performed, and the voltage drops back to the adaptive float charge voltage after equalization is completed.

[0030] Preferably, the control strategy is implemented using closed-loop smooth adjustment: the cloud platform sends instructions via the MQTT protocol, and the intelligent edge gateway sends adjustment instructions to the switching power supply via the RS485 bus to smoothly adjust the output voltage at a preset rate and limit the charging current at a preset rate.

[0031] Preferably, the method further includes:

[0032] Step S5: Post-evaluate the effect of the control strategy execution, and dynamically correct the cloud-based health benchmark model and parameter identification prior library based on the evaluation results, forming a closed-loop feedback mechanism of strategy execution - effect evaluation - model correction.

[0033] Preferably, step S5 specifically includes:

[0034] Collect voltage and current response data before and after equalization charging / float charging, and calculate the degree of improvement in battery consistency;

[0035] When the control strategy is effective and the change in SOH exceeds the preset threshold, the health baseline distribution model is dynamically updated using a rolling time-domain update rule.

[0036] The optimal model parameters obtained are stored in the aging parameter prior library, and a hot start initialization mechanism for the particle swarm algorithm is constructed to realize the evolution of the parameter identification prior library.

[0037] Preferably, the warm-start initialization mechanism specifically includes:

[0038] The particle swarm optimization algorithm uses the historical optimal parameter set in the prior library as the mean vector to generate multidimensional Gaussian distributed initial particles.

[0039] A scale-adaptive diagonal covariance matrix is ​​constructed to automatically match the perturbation intensity of each parameter dimension with the physical quantity.

[0040] Preferably, the method further includes:

[0041] Step S6: Based on the battery health status (SOH), remaining service life (RUL), and individual cell voltage range (ΔV), perform differentiated maintenance decisions:

[0042] When the State of Health (SOH) is below the retirement threshold or the Rullow is below the preset time limit, a battery replacement / scrap work order is generated.

[0043] When the SOH is higher than the decommissioning threshold but the voltage range exceeds the warning threshold, active balancing maintenance is triggered.

[0044] When both SOH and voltage range meet normal conditions, maintain standard float charging and extend the inspection cycle.

[0045] The present invention also provides a communication power supply battery health status assessment device, comprising:

[0046] One or more processors;

[0047] Memory;

[0048] One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to perform the communication power supply battery health status assessment method.

[0049] The present invention also provides a computer program product, including computer instructions, which instruct a computer device to perform operations corresponding to the communication power battery health status assessment method.

[0050] The beneficial effects of this invention are as follows: This invention achieves lightweight anomaly identification at the edge and high-precision modeling in the cloud through an edge-cloud collaborative architecture; it significantly reduces bandwidth consumption by adopting a hierarchical data upload mechanism based on KLD divergence; it effectively improves convergence speed and identification accuracy by utilizing anomaly feature constraints on the parameter identification space, avoiding the shortcomings of traditional algorithms that easily get trapped in local optima; it improves the accuracy of SOH assessment by using a joint criterion of capacity and internal resistance, and alleviates battery aging and consistency degradation problems by combining an adaptive dynamic equalization float charging strategy, extending battery life; it constructs a closed-loop feedback mechanism of strategy execution-effect evaluation-model correction, adaptively updating the health benchmark distribution and parameter prior library, overcoming the model inaccuracy problem caused by the drift of battery life cycle characteristics; it further improves real-time performance through PSO hot start and scale adaptive initialization, and achieves differentiated intelligent maintenance decisions based on SOH, RUL, and voltage range, significantly reducing manual inspection costs, effectively avoiding the problem of failure to detect bad batteries in time and the mistaken scrapping of good batteries, and comprehensively improving the operational stability and maintenance efficiency of the communication base station backup power system. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the 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.

[0052] Figure 1 This is a flowchart illustrating a method for assessing the health status of a communication power supply battery according to an embodiment of the present invention.

[0053] Figure 2 This is a schematic diagram of the implementation architecture of a communication power supply battery health status assessment method according to Embodiment 2 of the present invention.

[0054] Figure 3 This is a schematic diagram of the edge-side Kullback-Leibler divergence monitoring process in an embodiment of the present invention.

[0055] Figure 4 This is a schematic diagram of the cloud-based high-precision evaluation process in an embodiment of the present invention.

[0056] Figure 5 This is a schematic diagram of the dynamic equalization and floating charge control process in an embodiment of the present invention.

[0057] Figure 6 This is a diagram comparing the accuracy of RUL remaining lifetime prediction. Detailed Implementation

[0058] The following description of the embodiments is taken with reference to the accompanying drawings, which illustrate specific embodiments in which the invention can be implemented.

[0059] Please refer to Figure 1 As shown, this embodiment of the invention provides a method for assessing the health status of a communication power supply battery, including:

[0060] Step S1: The battery acquisition terminal collects the voltage, current and temperature operating data of the communication power backup battery pack in real time and transmits them to the intelligent edge gateway.

[0061] Step S2: The intelligent edge gateway preprocesses the operating data, identifies abnormal states based on battery voltage distribution characteristics, and executes a hierarchical data upload strategy according to the abnormal identification results. Only when abnormal polarization characteristics of the battery are identified, the waveform data of the corresponding time period is uploaded to the cloud platform.

[0062] Step S3: After receiving the abnormal waveform data, the cloud platform constructs a battery equivalent circuit model, identifies the battery model parameters based on the parameter search space of the abnormal feature constraint parameter identification algorithm based on edge side identification, and calculates the battery health status based on the battery model parameters.

[0063] In step S4, the cloud platform generates a dynamic equalization and floating charging control strategy based on the calculated battery health status and abnormal characteristics, and then sends the control strategy to the intelligent edge gateway for execution.

[0064] As can be seen from the above, the communication power supply battery health status assessment method of the present invention is based on an edge-cloud collaborative communication power management architecture consisting of a cloud platform, a smart edge gateway (FSU), and a bottom-level battery acquisition terminal, to realize the health status assessment and adaptive control of the communication power supply backup battery pack.

[0065] Specifically, please combine Figure 2 As shown, in step S1, the bottom-level battery acquisition terminal collects real-time voltage, current, and temperature operating data of the battery pack through sensors deployed on the backup battery pack of the communication power supply. The collected operating data is then transmitted to the intelligent edge gateway, completing the physical sensing and initial aggregation of the raw operating data, providing a foundation for subsequent SOH estimation and dynamic equalization and floating charging strategy generation. The specific process is as follows:

[0066] Battery pack module data acquisition: Acquiring individual cell voltages Total current of battery pack and battery surface temperature In this embodiment, the tested object is a 48V valve-regulated sealed lead-acid battery pack, with a number of individual cells... .

[0067] At each sampling time, the multidimensional data vector obtained by the system is represented as follows:

[0068]

[0069] In the formula, , D This represents the total number of feature dimensions. Based on the number of monitored cells and sensor configuration, the sampling frequency in this embodiment is set to 10Hz to capture transient polarization features.

[0070] The Smart Edge Gateway (FSU) reads the aforementioned vector data via the Modbus-RTU protocol and is responsible for data time alignment and encapsulation. The FSU communicates with the Smart Switching Power Supply (SMPS) via an RS485 bus to achieve voltage regulation control; and connects to the cloud platform via the MQTT protocol to upload data and receive policy commands.

[0071] Step S2 completes edge-side data preprocessing, status anomaly identification, and hierarchical data uploading.

[0072] After receiving the operational data transmitted from the underlying layer, the intelligent edge gateway first preprocesses the data, then completes the identification of abnormal states based on the battery voltage distribution characteristics, and finally executes a hierarchical data upload strategy based on the identification results.

[0073] Specifically, before uploading data to the cloud platform, the intelligent edge gateway needs to process the collected raw data vectors. Preprocessing is performed to eliminate the impact of environmental noise and transmission errors on subsequent feature extraction. The specific process includes:

[0074] 1. Data Integrity Verification and Missing Value Imputation: The system first checks the CRC checksum of received Modbus data frames and discards bad frames that fail the check. For sporadic data loss due to network fluctuations (i.e., discontinuous timestamps), linear interpolation is used for imputation. Assuming that in... If time-based data is missing, it will be reconstructed using data from the preceding and following times.

[0075]

[0076] In the formula, The target time when data loss occurs; and These are the two most recent valid sampling times before and after the missing time; Data values ​​to be filled in for reconstruction; and These represent the effective voltage or current data values ​​actually collected at the corresponding time.

[0077] 2. Outlier removal ( Guidelines):

[0078] To address erroneous sampled values ​​caused by transient interference, sensor drift, or communication anomalies in the voltage acquisition loop, the system employs a sliding window statistical method at the edge to remove outliers. Let the sliding window size be... Calculate the voltage sampling sequence within each window. mean and standard deviation When a sampling point meets the following condition: If the sampling point is not found to be an outlier or noise point, the system determines that the sampling point is an outlier or noise point. For detected outliers, the system uses median filtering to correct them: the median filtering method is used to filter out outliers before and after the sampling point. Using the voltage data of each valid sampling point as a window, calculate the median and use this median to replace the original outliers.

[0079] After median filtering, spike interference and abrupt changes can be effectively smoothed, ensuring the continuity and stability of the voltage data sequence and providing reliable data input for subsequent KLD feature modeling and PSO algorithm optimization.

[0080] 3. Digital Filtering and Smoothing: To extract the low-frequency voltage response reflecting the battery's chemical characteristics and filter out power frequency interference and high-frequency switching noise, this embodiment uses a finite impulse response (FIR) low-pass filter to smooth the data. The filter design employs the Hamming window function method, with a cutoff frequency of... Set to 0.5Hz. The voltage sequence after filtering. This will serve as valid input data for subsequent KLD divergence calculations and cloud-based PSO parameter identification.

[0081] As shown above, this embodiment constructs a complete link from underlying physical sensing to cloud data aggregation. Through high-frequency sampling and real-time preprocessing at the edge, the system can effectively filter out on-site power frequency interference and random noise, ensuring the accuracy of the voltage sequence uploaded to the cloud platform. With current sequence It possesses high signal-to-noise ratio (SNR) and time synchronization. This not only provides a reliable data foundation for subsequent small feature extraction based on Kullback-Leibler divergence (KLD, relative entropy), but also provides the necessary input guarantee for the convergence accuracy of complex model parameter identification (SOH estimation) in the cloud, achieving a dual improvement in data quality and algorithm performance.

[0082] As a hub connecting physical devices and the cloud, the intelligent edge gateway addresses the bandwidth pressure and cloud computing power waste caused by massive, high-frequency data uploads. This embodiment innovatively proposes an edge-side feature initial screening mechanism based on Kullback-Leibler divergence. This mechanism does not directly upload the original voltage and current waveforms; instead, it uses statistical methods to quantify the degree of consistency deviation of the battery pack, triggering data backhaul only when abnormal polarization features are detected. Please refer to... Figure 3 As shown, the specific implementation process is as follows:

[0083] Step S21: Voltage sequence truncation and normalization based on sliding window.

[0084] The FSU maintains a first-in-first-out (FIFO) data queue of length W in local memory. For any single battery cell, the current moment is captured in real time. Previous voltage observation sequence .

[0085] To eliminate the influence of voltage absolute value differences at different charging and discharging stages (such as float charging at 53.6V and equalizing charging at 56.1V) on the distribution pattern, the system first performs... Standardization process yields dimensionless standard sequences. :

[0086]

[0087] In the formula: Indicates a single cell The first within the time window The voltage values ​​of each original sampling point; This represents the arithmetic mean of the voltage sequence within that time window; This represents the standard deviation of the voltage series within that time window; This represents the voltage characteristic value after standardization, and its physical meaning is to reflect the degree to which the voltage deviates from the mean.

[0088] Step S22: Adaptive kernel density estimation constructs a probability model.

[0089] In order to accurately describe the probability distribution of the standardized sequence (especially to capture subtle polarization bimodal features), this embodiment abandons the traditional histogram statistical method (because it is sensitive to interval division) and instead uses the Gaussian kernel density estimation (KDE) method to construct a continuous probability density function (PDF).

[0090] Let the standardized sequence sample set be Then the sequence at any eigenvalue probability density estimate at The calculation formula is:

[0091]

[0092] In the formula: The total number of samples within the sliding window (in this example, the window length is 10 minutes, and the sampling rate is...). , ); h The bandwidth parameter controls the smoothness of the distribution curve. In this embodiment, the Silverman rule is used to adaptively calculate the optimal bandwidth. (in (sample standard deviation); For the kernel function, this embodiment uses the standard normal distribution function: , is the independent variable of the probability density function, representing the range of values ​​for the normalized voltage.

[0093] Through the above calculations, the actual distribution function of the single-cell voltage at the current moment is obtained, denoted as: .

[0094] Step S23: Comparison of baseline distribution and calculation of KLD divergence.

[0095] FSU local storage has a set of health benchmark distribution models. This health baseline distribution model is based on the initial stage of battery pack deployment. Alternatively, a standard normal distribution (or a measured healthy distribution) can be generated by collecting and training data under the same operating conditions after the most recent in-depth maintenance.

[0096] In order to quantify the current state Compared with the baseline state To determine the information difference between the two, the system calculates their Kullback-Leibler divergence (relative entropy). In engineering implementation, this is calculated using a discretized integral form.

[0097]

[0098] In the formula: express The divergence value, in its physical sense, represents the amount of information loss in the current voltage distribution relative to a healthy baseline distribution. The higher the value, the worse the battery consistency and the more severe the internal polarization; The number of parts in the discretization interval; For the first The center point values ​​of each discrete interval; Let be the step size of the discrete interval; For a very small positive number, this embodiment takes... This is used to prevent calculation overflow errors caused by a denominator of zero.

[0099] Step S24, anomaly classification triggering and data upload strategy.

[0100] Based on real-time calculations Values ​​are used to implement a tiered data governance strategy to balance real-time performance with communication costs.

[0101] Strategy 1 (Silent Mode): When (A low threshold is set in this embodiment) When the battery pack is in a state of electrochemical stability, it is determined that it is in a stable state. Only one metadata log entry (including timestamp) is stored in the local database. Maximum voltage Minimum voltage and divergence value ), without uploading the original waveform data to the cloud.

[0102] Strategy 2 (Triggering Mode): When When an abnormal polarization trend or inconsistency degradation is detected in the battery pack, the FSU immediately triggers the event-driven upload mechanism:

[0103] 1. Data Freeze: Lock the current time frame. (5 minutes) until after (1 minute) complete original voltage Current and temperature Waveform data.

[0104] 2. Tag Attachment: Add a trigger reason tag to the data packet header. And the specific divergence calculation value.

[0105] 3. Priority Push: Through the high-priority channel of the MQTT protocol ( The data packet is pushed to the cloud platform for use in the complex model identification in the subsequent step S2.

[0106] Through the above steps, this embodiment achieves a highly efficient collaborative mode of lightweight computing on the terminal side and on-demand intervention on the cloud side, effectively filtering more than 90% of invalid and redundant data and significantly reducing the bandwidth consumption of the base station backhaul network.

[0107] Step S3 is used to achieve accurate SOH identification and algorithm optimization in the cloud.

[0108] Once the cloud platform receives the waveform data packet tagged with High_KLD uploaded from the edge side, it immediately starts the high-performance computing engine. This step aims to quickly and accurately invert the battery's internal state parameters and achieve online assessment of the state of health (SOH) by constructing an electrochemical equivalent circuit model and utilizing the search space of the algorithm based on edge features (KLD values) to dynamically constrain parameters. Please refer to... Figure 4 As shown, the specific process is as follows:

[0109] Step S31: Model the second-order RC equivalent circuit.

[0110] To balance computational complexity with the ability to describe polarization effects, a second-order Thevenin equivalent circuit model was first built in the cloud. This model consists of an ideal voltage source. (Open circuit voltage), one ohm internal resistance And two parallel RC circuits (polarization internal resistance) Polarized capacitors It is composed of series connections.

[0111] According to Kirchhoff's Voltage Law (KVL), the discretized state-space equations of this model can be described as follows:

[0112]

[0113] In the formula: For a moment The observed terminal voltage; For a moment The load current (negative for charging, positive for discharging). It is the open-circuit voltage, which is a nonlinear function of the state of charge (SOC). The internal resistance in ohms reflects the instantaneous voltage drop characteristics of the battery. and These are the electrochemical polarization voltage and the concentration polarization voltage, respectively. , and , These are the resistance and capacitance parameters corresponding to electrochemical polarization and concentration polarization, respectively. Their product determines the polarization response speed on a short timescale. The sampling time interval is 0.1s in this example.

[0114] Step S32: Optimize adaptive particle swarm (PSO) parameters based on KLD feature constraints.

[0115] In order to accurately identify the key parameters (especially the ohmic resistance) of the above second-order RC model from the observation data and polarization resistance In this embodiment, the particle swarm optimization algorithm (PSO) is used for iterative optimization.

[0116] Here, the present invention improves upon the inherent defects of the traditional PSO algorithm in battery parameter identification:

[0117] In conventional PSO applications, due to the lack of prior knowledge about the current aging state of the battery, the algorithm must be implemented in the global parameter space (e.g., Blindly searching within the entire domain results in a huge search space, requiring numerous iterations to approximate the true value, leading to excessive computation time and failing to meet online real-time requirements. Furthermore, the battery model is nonlinear, exhibiting multiple local minima. Global search can easily cause the particle swarm to converge prematurely to erroneous local optima, resulting in the identified internal resistance value deviating significantly from the true physical value. To address the two major drawbacks of traditional methods—slow convergence and susceptibility to local optima—this embodiment utilizes KLD divergence values ​​uploaded from the edge side. As a powerful indicator of prior state. This directly reflects the degree of electrochemical polarization and consistency deviation within the battery. Based on this, this invention constructs a "divergence-space mapping mechanism" to dynamically compress the search range of particles, achieving targeted detection.

[0118] Let the set of parameters to be identified be Among them, the internal resistance of the ohm search range It is no longer a fixed global range, but rather based on... Values ​​are set dynamically:

[0119]

[0120] In the formula: This is the battery's nominal factory internal resistance (reference value). The stability threshold is set to 0.02 in this embodiment. The aging gain coefficient (taken in this embodiment) Specifically, for the first run without historical data, the algorithm defaults to... Perform a global search.

[0121] when At this point, it indicates that the battery is in a healthy and stable state, with minimal change in internal resistance. At this time, the search range is strongly compressed into a narrow band near the nominal value. This greatly reduces invalid searches and significantly improves convergence speed. When When the voltage reaches a certain level, it indicates that the internal polarization of the battery is intensifying, and the internal resistance will inevitably increase significantly. At this time, the algorithm automatically shifts the entire search interval to the high-resistance region and appropriately widens the range to cover possible aging conditions. This effectively avoids wasting computational power in the low-resistance region (invalid region) and prevents the trap of local optima caused by not being able to find the true value.

[0122] Through the above mechanism, this invention compresses the search volume of the high-dimensional parameter space by more than 80%, which improves the convergence speed of the PSO algorithm by 3-5 times and the identification accuracy (RMSE) is better than the traditional global search method, realizing fast and accurate modeling of massive battery data in the cloud.

[0123] Step S33: Construction and iterative solution of the fitness function.

[0124] After determining the search space, the PSO algorithm is initialized. Particles ( M =50), each particle represents a set of potential parameter solutions. Define the fitness function. The root mean square error (RMSE) between the model output voltage and the actual observed voltage:

[0125]

[0126] In the formula: L The length of the data window; This represents the actual voltage value collected. Predict voltage values ​​for the model under the current parameters.

[0127] The algorithm updates the particle velocity iteratively. and position, up to the fitness function Less than the preset error limit Or it may reach the maximum number of iterations. Finally, the globally optimal parameter solution is output. This includes the optimal internal resistance estimate at the current moment. .

[0128] Step S34: Combine SOH estimation.

[0129] Based on the parameters identified in step S23, and with corrections using the Ah-Integration method, the actual maximum usable capacity of the current battery is calculated. Finally, the State of Health (SOH) of the battery is defined based on capacity decay:

[0130]

[0131] In the formula: This is the current estimated full charge capacity; The nominal rated capacity of the battery is 500Ah.

[0132] Simultaneously, the system will also calculate the internal resistance type SOH as an auxiliary criterion:

[0133]

[0134] In the formula: The internal resistance of the new battery; Internal resistance at the end of life (taken in this embodiment) ).

[0135] Final output The value will serve as the core decision-making basis for generating the dynamic charging and discharging strategy in the subsequent step S4.

[0136] Step S4 is used to generate and execute the dynamic equalization and equalization charging strategy.

[0137] Based on the battery health status identified in step S3 and the divergence index calculated in step S2 The cloud-based decision engine uses serial hierarchical control logic to generate the optimal charging and discharging strategy. This step aims to solve the problems of overcharging and water loss in aging batteries and the expansion of individual cell inconsistencies caused by traditional fixed-voltage float charging, achieving precise maintenance with a tailored approach for each site. Figure 5 As shown, the specific process is as follows:

[0138] Step S41, aging adaptive float charge voltage setting based on SOH.

[0139] Traditional communication power supplies typically use a fixed float charge voltage (such as a 53.5V / 48V system). However, as the battery ages (SOH decreases), its gas evolution potential decreases, and a fixed high-voltage float charge will accelerate electrolyte water loss and grid corrosion.

[0140] This embodiment constructs a SOH-voltage negative feedback regulation model. The system first determines whether the current SOH value is lower than the aging threshold. (This embodiment uses 80%)

[0141] Scenario A (Healthy Period): If The battery is in a stable service life, maintaining the standard float charge voltage command:

[0142]

[0143] Scenario B (Aging Period): If As the battery enters its aging and degradation phase, the system implements a linear voltage reduction protection strategy. Float charge voltage setpoint. It decreases dynamically as SOH decreases, and the calculation formula is:

[0144]

[0145] In the formula: Standard float charge voltage (53.6V); Aging compensation coefficient (range of values) (In this embodiment, 0.05 is used). This is the aging threshold.

[0146] For example, when the SOH drops to 60%, the float charge voltage is automatically adjusted to... This strategy significantly reduces the risk of float current and thermal runaway while ensuring full charging.

[0147] Step S42, equalization trigger determination (equalization control) based on KLD characteristics.

[0148] After determining the basic float charge voltage Then, the system further considers consistency metrics. Determine whether Boost Charge needs to be initiated. To prevent control oscillations at the critical point, this embodiment introduces a hysteresis comparison mechanism:

[0149] Triggering condition: If the current ( The equalization charge trigger threshold is set to 0.08 in this embodiment. If the battery pack's internal consistency is severely deteriorated, the system overrides the instruction in step S41, forcibly generating an equalization charge instruction. The target voltage is set as follows:

[0150]

[0151] In the formula: The target voltage for equalization charging.

[0152] Conditions for maintaining: If The current charging status will remain unchanged.

[0153] Exit Conditions: Only when ( The equalization charge exit threshold is set to 0.05 in this embodiment, and the rate of change of the charging current tail meets the following conditions. When equalization is complete, the system cancels the equalization charging command, and the voltage drops back to the level determined in step S31. In addition, if the equalization charging time exceeds the preset safety time limit (such as 10 hours), the system will also force the equalization charging to exit and report the abnormality.

[0154] In the formula: The current change rate cutoff threshold is set to [value] in this embodiment. When the current change is extremely small (tending to constant), it indicates that the battery is close to being fully charged. At this point, combined with... The descent condition indicates that the equilibrium process is complete.

[0155] Step S43: Issuance and execution of closed-loop instructions.

[0156] The cloud will encapsulate the final generated control parameters into a JSON format instruction package (e.g. The data is sent to the Smart Edge Gateway (FSU) via the MQTT protocol.

[0157] After receiving the command, the FSU performs protocol parsing and security verification (CRC), and then sends a remote adjustment command to the switching power supply (SMPS) via the RS485 bus. The specific execution logic is as follows:

[0158] 1. Voltage Regulation: The FSU writes the output voltage register (Register 0x1001) to the SMPS to smoothly regulate the bus voltage to... The adjustment rate is limited to To avoid electrical surges.

[0159] 2. Current Limit: Simultaneously write to the current limit register (Register 0x1002) to ensure that the charging current does not exceed [the limit value]. (i.e., 50A) to ensure battery safety.

[0160] Through the above steps, this embodiment achieves a leap from passive fixed maintenance to active adaptive management, which not only extends the life of aging batteries by reducing voltage, but also solves the problem of individual cell differences by dynamic equalization charging.

[0161] To overcome the model inaccuracy caused by electrochemical characteristic drift throughout the battery's life cycle, this embodiment constructs a closed-loop feedback mechanism of strategy execution, effect evaluation, and model correction. The system not only executes control commands unidirectionally but also dynamically updates the cloud-based health benchmark distribution model based on the actual battery response after command execution. And an aging parameter library, enabling adaptive evolution of the algorithm. Therefore, this invention also includes step S5, adaptive correction of model parameters and updating of the knowledge base. The specific process is as follows:

[0162] Step S51: Post-evaluation of the strategy implementation effect.

[0163] After the equalization charge or float charge command issued in step S3 is executed (i.e. At that moment, the cloud immediately initiates the effect evaluation program. The system collects the voltage response curve during command execution. and current curve Calculate the consistency improvement of this strategy execution. :

[0164]

[0165] In the formula: This represents the initial divergence value before the policy is executed (at the trigger moment); The strategy execution ends and is then paused. The final divergence value after 30 minutes.

[0166] like (Set an effective threshold of 15%), and the strategy is deemed effective, with significant improvement in battery consistency. If If the strategy is deemed ineffective, a warning is issued indicating that the battery may have irreversible physical damage (such as grid breakage or short circuit), requiring a manual maintenance alarm to be triggered.

[0167] Step S52, Health Benchmark Distribution Dynamic updates.

[0168] Traditional SOH estimation typically uses a fixed baseline from the factory. This ignores the relative health status of the battery at different aging stages. This embodiment employs a rolling time-domain update strategy to periodically correct the baseline distribution. .

[0169] When the following two conditions are met:

[0170] 1. The battery pack has just completed one full equalization charge cycle and ;

[0171] 2. Current estimated SOH value Compared to the last update The difference exceeds 5% (i.e. );

[0172] The system will display the voltage distribution at the current moment (the static steady state after the equalization charge is completed). The system accepts the new local baseline distribution and updates the cloud database. Before updating, the system first verifies... Statistical characteristics (such as kurtosis and skewness) are used to remove outliers caused by sampling anomalies. The following update formula is only executed after the data validity check has passed:

[0173]

[0174] In the formula: For the updated next-generation health baseline distribution model, Previous generation health baseline distribution model before update The forgetting factor is set to 0.2 in this embodiment.

[0175] This formula employs the concept of Exponentially Weighted Moving Average (EWMA), which retains the statistical characteristics of historical benchmarks while incorporating the latest health status features, thus... The calculations are always based on the optimal state at the current aging stage, rather than the unattainable factory state.

[0176] Step S53, Evolution of the Particle Swarm Optimization (PSO) Prior Library.

[0177] To address the computational time consumption caused by the need for a global blind search (cold start) at each startup in the traditional PSO algorithm, this embodiment constructs an aging parameter prior library and proposes a scale-adaptive Gaussian perturbation initialization strategy.

[0178] The system will identify the optimal parameter set successfully in step S3. The aging parameter prior library is stored in the cloud. During the next startup step S3 for parameter identification, the PSO algorithm no longer uses the traditional global uniform distribution initialization. Instead, it reads the historically optimal parameters from the prior library and uses these as the center to generate an initial particle swarm that follows a multidimensional Gaussian distribution.

[0179]

[0180] In the formula: For the first The initial position vector of each particle in generation 0; It is a mean vector, and its values ​​are the historical optimal parameter set in the prior library; To initialize the covariance matrix.

[0181] Because the battery model parameters vary greatly in physical magnitude, using a uniform fixed variance would lead to divergence in the search for small-scale parameters or stagnation in the search for large-scale parameters. Therefore, this embodiment constructs a scale-adaptive diagonal covariance matrix to ensure that the perturbation strength of each dimension automatically matches its physical magnitude. The construction process is as follows:

[0182] First, define the disturbance intensity coefficient. (In this embodiment, we take 0.1); Next, we calculate the independent standard deviation of each parameter dimension. :

[0183]

[0184] Finally, construct the diagonal matrix. :

[0185]

[0186] In the formula: This represents the covariance matrix initialized for the particle swarm. Indicates the diagonalization operator; Indicates the first Initialized variance for each parameter dimension; Indicates the deflection intensity coefficient; For the first The historical best value of each parameter to be identified This represents the total number of parameters.

[0187] Through this memory replay mechanism, the algorithm can start searching directly near the historical optimal solution, avoiding the long iteration time caused by cold start. This makes the subsequent SOH estimation speed faster and faster with the number of uses, further improving the real-time response capability of the system.

[0188] To further verify the advantages of the method of the present invention over traditional detection algorithms in battery state of health (SOH) estimation and remaining useful life (RUL) prediction, the standard PSO algorithm (without prior constraints) and the traditional voltage threshold method were compared with the method of the present invention (improved PSO based on KLD constraints).

[0189] The experiment selected a 48V / 500Ah valve-regulated lead-acid battery pack for accelerated aging cycle testing, and selected RMSE (root mean square error) and MAE (mean absolute error) as evaluation indicators.

[0190] 1. Comparison of SOH estimation accuracy

[0191] Table 1 shows the tracking and estimation performance of the three methods for SOH over the entire battery lifespan (0-600 cycles):

[0192] Table 1. Comparison of SOH estimation accuracy of different algorithms

[0193]

[0194] Analysis of the above experimental results leads to the conclusion that the improved PSO method of this invention is significantly superior to the traditional method in terms of RMSE and MAE indicators, indicating that this invention has extremely high accuracy and stability in the state identification of the core component (battery) of communication power supply.

[0195] 2. Comparison of RUL remaining life predictions

[0196] Further verification of the ability of each algorithm to predict the "remaining usable days" of the battery is shown in Table 2. Figure 6 As shown:

[0197] Table 2 Comparison of RUL prediction errors of different algorithms

[0198]

[0199] Based on the aforementioned high-precision prediction results, the system achieves refined management of the equipment. Therefore, this invention also includes step S6, which combines the SOH, RUL, and consistency indicators output by the model to formulate differentiated maintenance decisions. The specific process is as follows:

[0200] Based on the model's output of the predicted remaining useful life Current health status and battery pack voltage range Managers can assess the current status of the equipment.

[0201] The specific process for maintaining decision-making is as follows:

[0202] SOH threshold setting: (Retirement threshold) (Attention threshold).

[0203] Voltage range threshold setting: (Severe imbalance) (Slight imbalance).

[0204] 1. Replacement / Scrap Decision: or If the system determines that the battery has reached the end of its lifespan, it will automatically generate a replacement recommendation work order, suggesting that the entire battery pack be replaced.

[0205] 2. Balanced maintenance decision: when At that time, If a consistency problem is detected in the battery pack, the system automatically issues an active balancing command, triggering the equalization charging strategy described in step S3.

[0206] 3. Normal operation decision: When and In this way, normal floating charging operation can be maintained, and the on-site inspection cycle can be postponed, thereby reducing operation and maintenance costs.

[0207] The goal of building the aforementioned decision support system is to achieve intelligent full lifecycle management of equipment, ensure the safe and efficient operation of communication power supply equipment, and maximize the extension of equipment lifespan.

[0208] It should be noted that the edge-cloud collaborative architecture of this invention can be extended to distributed energy systems in various scenarios such as new energy, transportation, industry, and the Internet of Things. The system preprocesses and analyzes fault trends of multi-source operating data at edge nodes, uploading key features and model parameters to the cloud platform. The cloud continuously updates the battery health model based on big data analysis and optimization algorithms, forming unified monitoring and decision support for cross-regional equipment. Data synchronization and policy closed-loop are achieved through 5G or industrial gateways, allowing field equipment to adaptively adjust charging voltage, load distribution, and maintenance cycles, thereby improving the intelligence and reliability of the overall energy system and achieving predictive maintenance and energy efficiency optimization.

[0209] Corresponding to the communication power supply battery health status assessment method in Embodiment 1 of the present invention, Embodiment 2 of the present invention also provides a communication power supply battery health status assessment device, comprising:

[0210] One or more processors;

[0211] Memory;

[0212] One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to perform the communication power supply battery health status assessment method.

[0213] Corresponding to the communication power supply battery health status assessment method in Embodiment 1 of the present invention, Embodiment 3 of the present invention also provides a computer program product, including computer instructions, which instruct a computer device to perform the operation corresponding to the communication power supply battery health status assessment method.

[0214] Preferably, the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor can be any conventional processor. The processor is the control center of the device, connecting various parts of the device through various interfaces and lines.

[0215] The memory mainly includes a program storage area and a data storage area. The program storage area can store the operating system, applications required for at least one function, etc., while the data storage area can store related data, etc. Furthermore, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard drive, a SmartMedia Card (SMC), a Secure Digital (SD) card, and a Flash Card, or other volatile solid-state storage devices.

[0216] It should be noted that the above-mentioned devices may include, but are not limited to, processors and memory, as will be understood by those skilled in the art.

[0217] For the working principle and process of the above embodiments, please refer to the description of Embodiment 1 of the present invention, which will not be repeated here.

[0218] Compared with existing technologies, this invention has the following significant advantages: It achieves lightweight anomaly identification at the edge and high-precision modeling in the cloud through an edge-cloud collaborative architecture; it significantly reduces bandwidth usage by employing a hierarchical data upload mechanism based on KLD divergence; it effectively improves convergence speed and identification accuracy by utilizing anomaly feature constraints on the parameter identification space, avoiding the shortcomings of traditional algorithms that easily get trapped in local optima; it improves the accuracy of SOH assessment by using a joint criterion of capacity and internal resistance, and alleviates battery aging and consistency degradation problems by combining an adaptive dynamic equalization float charging strategy, extending battery life; it constructs a closed-loop feedback mechanism of strategy execution-effect evaluation-model correction, adaptively updating the health benchmark distribution and parameter prior library, overcoming the model inaccuracy problem caused by the drift of battery life cycle characteristics; it further improves real-time performance through PSO hot start and scale adaptive initialization, and achieves differentiated intelligent maintenance decisions based on SOH, RUL, and voltage range, significantly reducing manual inspection costs, effectively avoiding the problem of failure to detect bad batteries in time and the mistaken scrapping of good batteries, and comprehensively improving the operational stability and maintenance efficiency of the communication base station backup power system.

[0219] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for evaluating a state of health of a communication power supply battery, characterized by, include: Step S1: The battery acquisition terminal collects the voltage, current and temperature operating data of the communication power backup battery pack in real time and transmits them to the intelligent edge gateway. Step S2: The intelligent edge gateway preprocesses the operating data, identifies abnormal states based on battery voltage distribution characteristics, and executes a hierarchical data upload strategy according to the abnormal identification results. Only when abnormal polarization characteristics of the battery are identified, the waveform data of the corresponding time period is uploaded to the cloud platform. Step S3: After receiving the abnormal waveform data, the cloud platform constructs a battery equivalent circuit model, identifies the battery model parameters based on the parameter search space of the abnormal feature constraint parameter identification algorithm based on edge side identification, and calculates the battery health status based on the battery model parameters. Step S4: The cloud platform generates a dynamic equalization and floating charging control strategy based on the calculated battery health status and abnormal characteristics, and sends the control strategy to the intelligent edge gateway for execution. Step S3 specifically includes: A second-order Thevenin equivalent circuit model is established, and the discretized state-space equations of the model are constructed based on Kirchhoff's voltage law. Using KLD divergence as a constraint, a divergence-parameter space mapping mechanism is constructed to dynamically constrain the search interval of the Ohmic internal resistance of the particle swarm optimization (PSO) algorithm. Using the root mean square error (RMSE) between the model output voltage and the measured voltage as the fitness function, the global optimal parameters are output through PSO iterative optimization. The state of health (SOH) of the battery is calculated using a combined capacity-internal resistance criterion.

2. The method of claim 1, wherein, In step S2, the identification of abnormal states based on battery voltage distribution characteristics specifically includes: The intelligent edge gateway extracts the preprocessed voltage sequence through a FIFO sliding window and performs Z-Score standardization on the voltage sequence to eliminate the influence of voltage amplitude at different charging and discharging stages. A continuous probability distribution model of voltage sequence is constructed by using Gaussian kernel density estimation combined with Silverman empirical rule to adaptively calculate bandwidth. The current probability distribution model is compared with the locally stored health baseline distribution model, and the Kullback-Leibler divergence is calculated as the basis for judging battery polarization anomalies and consistency degradation.

3. The method according to claim 2, characterized in that, The hierarchical data upload strategy is an edge-cloud collaborative event-driven mechanism. When the KLD divergence is lower than the preset low threshold, it enters silent mode, only storing the timestamp, voltage extreme value, and divergence value locally, without uploading the original waveform; When the KLD divergence is higher than the preset low threshold, the trigger mode is entered, the complete voltage, current and temperature waveforms before and after the anomaly are frozen for a preset period of time, and after adding the abnormal polarization tag, they are uploaded to the cloud platform through the MQTT high priority channel.

4. The method of claim 1, wherein, The method of calculating the battery health state (SOH) using the capacity-internal resistance joint criterion specifically includes: using the ratio of the actual full-charge capacity to the rated capacity obtained by the ampere-hour integration method as the main criterion for SOH, and using the internal resistance attenuation ratio as the auxiliary criterion for SOH, and outputting the battery health state (SOH) in combination.

5. The method of claim 1, wherein, In step S4, the dynamic equalization and floating charge control strategy is a two-dimensional adaptive control process: The float charge voltage is adaptively adjusted according to the SOH value. When the SOH value is lower than the preset aging threshold, the float charge voltage is linearly reduced as the SOH value decreases. Based on the KLD divergence combined with the hysteresis comparison mechanism, the triggering, holding and exit control of equalization charging is performed, and the voltage drops back to the adaptive float charge voltage after equalization is completed.

6. The method of claim 5, wherein, The control strategy is implemented using closed-loop smooth adjustment: the cloud platform sends instructions via the MQTT protocol, and the intelligent edge gateway sends adjustment instructions to the switching power supply via the RS485 bus to smoothly adjust the output voltage at a preset rate and limit the charging current at a preset rate.

7. The method of claim 1, wherein, Also includes: Step S5: Post-evaluate the effect of the control strategy execution, and dynamically correct the cloud-based health benchmark model and parameter identification prior library based on the evaluation results, forming a closed-loop feedback mechanism of strategy execution - effect evaluation - model correction.

8. The method of claim 7, wherein, Step S5 specifically includes: Collect voltage and current response data before and after equalization charging / float charging, and calculate the degree of improvement in battery consistency; When the control strategy is effective and the change in SOH exceeds the preset threshold, the health baseline distribution model is dynamically updated using a rolling time-domain update rule. The optimal model parameters obtained are stored in the aging parameter prior library, and a hot start initialization mechanism for the particle swarm algorithm is constructed to realize the evolution of the parameter identification prior library.

9. The method of claim 8, wherein, The warm-start initialization mechanism is specifically as follows: The particle swarm optimization algorithm uses the historical optimal parameter set in the prior library as the mean vector to generate multidimensional Gaussian distributed initial particles. A scale-adaptive diagonal covariance matrix is ​​constructed to automatically match the perturbation intensity of each parameter dimension with the physical quantity.

10. The method according to claim 1, characterized in that, Also includes: Step S6: Based on the battery health status (SOH), remaining service life (RUL), and individual cell voltage range (ΔV), perform differentiated maintenance decisions: When the State of Health (SOH) is below the retirement threshold or the Rullow is below the preset time limit, a battery replacement / scrap work order is generated. When the SOH is higher than the decommissioning threshold but the voltage range exceeds the warning threshold, active balancing maintenance is triggered. When both SOH and voltage range meet normal conditions, maintain standard float charging and extend the inspection cycle.

11. A communication power supply battery state of health assessment device, characterized by, include: One or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to perform the communication power supply battery health status assessment method as described in any one of claims 1 to 10.

12. A computer program product, characterised in that, Includes computer instructions that instruct a computer device to perform an operation corresponding to the communication power battery health status assessment as described in any one of claims 1 to 10.

Images